Daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology. https://github.com/andrewsu/ai-nuggets en-us AI Nuggets by the Su Lab Daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology. Curated for Alan Huebschen. Alan Huebschen false Today's lead is a bioRxiv preprint that spends most of its length demolishing its own results and is more useful for it. On a frozen benchmark of 586 compounds with paired alpha-2A and alpha-2C affinities, Glide SP docking correlates with measured selectivity at Spearman 0.071 (CI includes zero) and GNINA CNN rescoring at 0.188, while a Morgan-fingerprint random forest that never sees a receptor reaches 0.564 under the strictest split. Blind redocking of five alpha-2 co-crystals put 20/20 seeds within 2 angstroms, so pose recovery is not the problem. Five controls then reorder everything: 90% of compound pairs draw both affinities from the same source document (within-document error correlation 0.66), so measuring noise rather than propagating it raises the attainable ceiling from 0.704 to 0.897; a cluster-identity-only model explains half the variance on the standard D3/D2 benchmark and none on alpha-2, meaning the usual test case is substantially solvable by series recognition; a metadynamics protocol assigns +1.43 to +4.79 kcal/mol of apparent selectivity to nonselective eticlopride across nine basin definitions while a genuinely D3-selective antagonist reads near zero; a same-receptor contact-divergence floor (A2A vs A2A at 1.77-fold, 5-HT2B vs itself at 1.68) makes the alpha-2 result of 1.88-fold indistinguishable from crystallographic variation; and five trivial SMILES descriptors reach 0.645, 72% of the measured ceiling, so any structure-based method scored here competes first against molecular size. Residualize size out and the pose contact pattern falls to 0.037 while the fingerprint retains 0.258. Practical read for alpha-2A programs (clonidine, lofexidine) and for xylazine/medetomidine subtype attribution: static docking will not rank candidates. Also today: nine frozen EEG representations show ICC 0.08-0.76 against AUC varying a tenth as much, with alpha driving reliability and theta driving AD/FTD discrimination, BENDR collapsing from 0.48 to 0.13 across depth, and a non-EEG time-series model among the most reliable. ProteinDB brings absolute PaxDb protein abundances into PK-Sim; CYP agreement is good, BCRP and UGT1A10 diverge 170- and 300-fold across platforms, but 16 of 17 discordant proteins sit inside PaxDb inter-study variability, and the rifampicin-midazolam DDI AUC-ratio GMFE improves from 1.90 to 1.66. RELEASE, a pragmatic cluster-randomised trial across 26 Australian GP practices (n=483, mean 14.1 years of antidepressant use, recruited irrespective of intent to stop), reports 12-month cessation of 14.9% vs 8.6% (OR 1.95, p=0.050), 6-month cessation 11.7% vs 4.8%, >75% dose reduction 19.6% vs 9.9%, and no difference in depression, anxiety or withdrawal scores. Anodal left-DLPFC tDCS lowers self-reported distress in controls (d=-0.30) and open-label placebo lowers it under sham (d=-0.40), but the two do not add - a significant group-by-stimulation interaction that undercuts the dorsolateral expectancy account of open-label placebo. PSILAUT (67 adults, 37 autistic, 2/5 mg psilocybin crossover) finds no baseline group differences but opposite 5 mg responses: within-network frontoparietal and limbic decreases in non-autistic participants, increased DMN-FPN and dorsal-ventral attention integration in autistic ones, scaling continuously with Autism Quotient. A Molecular Psychiatry circuit paper separates diminished interest from anhedonia: chemogenetic control of vmPFC-to-posterior-piriform (not anterior) moves exploratory rearing without touching tail suspension, sucrose preference, locomotion or olfaction, works through Gria1/GluA1-dependent burst firing, degrades after two weeks of CUMS before other measures, and is rescued by reactivating tagged vmPFC ensembles at four weeks. Lithium, as a synaptic downscaling agent, rescues resting potential and AP waveform in human neurons carrying SNAP25 V48F/D166Y/L50S where levetiracetam and topiramate do nothing, and in a CRISPR L50S knock-in mouse reduces seizure frequency and improves long-term memory (but not sociability or grip strength). MALDI mass-spec imaging separates nandrolone, testosterone and trenbolone by tail-striatum dopamine/3-MT and phospholipid class. HealthBench-Psych extracts 610 mental-health conversations from HealthBench via blinded clinician review with concealed known-exclude controls; 20 models across three judges give a statistically tied top five (0.610-0.627), self-preference vanishing after severity correction, and refusals as the only clear differentiator. And a PNNL ablation on protein-characterization agents finds model choice dominant (Opus 92-94% vs o4-mini 40-50%), federation nearly free (-0.4 to -6.8 pp), and a PPO policy at 88% with zero token cost, fastest latency and consistency of exactly 1.000 - minus any reasoning trace. https://www.biorxiv.org/content/10.64898/2026.08.18.745649v1 2026-08-27-receptor-and-reason Thu, 27 Aug 2026 12:00:00 +0000 1232 Eleven items. Lead: an alpha-2 adrenergic subtype-selectivity benchmark where Glide SP scores 0.071, CNN rescoring 0.188, a receptor-blind fingerprint 0.564 and five trivial SMILES descriptors 0.645 (72% of a measured ceiling of 0.897, up from 0.704 propagated) - plus a cluster-identity null, a nonselective free-energy reference, and a same-receptor contact floor that each retract a conclusion. Also: EEG foundation-model reliability is unpredictable from discrimination or pretraining; ProteinDB brings absolute proteomics into PK-Sim; RELEASE moves antidepressant cessation to 14.9% vs 8.6% after a mean 14 years of use; DLPFC tDCS fails to augment open-label placebo; low-dose psilocybin integrates autistic brains and segregates non-autistic ones; a vmPFC-to-posterior-piriform burst-firing pathway carries diminished interest specifically; lithium rescues SNAP25 encephalopathy phenotypes by AMPAR downscaling; MALDI imaging separates three anabolic steroids neurochemically; HealthBench-Psych finds a flat frontier and refusal behavior as the differentiator; and a PPO router matches frontier LLMs at zero token cost and perfect consistency. false Daily roundup for August 26, 2026 — eleven segments. LEAD: a Clinical Pharmacology & Therapeutics simulation study from Leiden tests the assumption underneath every CNS occupancy projection — that unbound brain extracellular fluid concentration is the biophase. Using LeiCNS-PK3.0 (microvascular, brain ECF, brain ICF, brain cell membrane, lysosomal and four CSF compartments) extended with D2 binding kinetics, receptor internalization and recycling, and endogenous dopamine competition, they built population PK models for haloperidol, risperidone, aripiprazole, brexpiprazole, cariprazine, sulpiride, remoxipride and raclopride and compared predicted occupancy against PET occupancy in caudate, putamen and striatum from 66 participants (46 with schizophrenia) across eight studies. Brain cell membrane concentrations aligned better than brain ECF for five of eight compounds — and lipophilicity does not identify which five, since brexpiprazole, the most lipophilic compound in the set, was one of the two best described by ECF. Substituting a microsomal membrane partition coefficient for logP improved membrane-based predictions by up to 10 percentage points for haloperidol and sulpiride. Including receptor internalization helped remoxipride and haloperidol and hurt five others. Schizophrenia-associated changes in intracellular pH and membrane phospholipid content had negligible effect. Endogenous dopamine competition was negligible even at simulated phasic surges of 5,000% of baseline, because dopamine’s D2 Kd of ~6,369 nM is 33- to 296,000-fold weaker than the ligands tested — a modeled result that sits awkwardly beside the tracer-dose displacement literature. Raclopride, sulpiride and cariprazine still miss at RMSE 23.4–27.9%. Also: a Molecular Psychiatry study locating stress-induced cholinergic hyperactivity in the laterodorsal tegmental nucleus to loss of local somatostatin-interneuron inhibition rather than increased excitatory drive, with activation of those interneurons during chronic unpredictable mild stress preserving SST-to-ChAT transmission and limiting depression-like behavior — framed by the authors as preventive, with the caveat that reduced afferent input and cholinergic autoregulation of the same local circuit remain unexcluded; a preprint decoding ESM-2 embeddings through InterPLM sparse autoencoders for neuropeptide precursor discovery, where per-feature F1 scores for raw ESM-2 collapse degenerately around 0.66 while sparse features spread with many above 0.8, and logistic regression on the top 320 features reaches ~92% cross-validated accuracy and recovers 88.7% of annotated human neuropeptides (46 false positives) against DeepPeptide’s 78.0% (88 false positives) and NeuroPP’s near-total recall at 270 false positives, generalizing cleanly to C. elegans and Drosophila and only partly to mouse and zebrafish where the misses are large protein hormones the operational definition excludes; PandaDock, an open-source docking platform with torsion-tree flexibility, Monte Carlo plus L-BFGS refinement, closed-form rotational gradients through the SO(3) exponential map, an exact blocked-neighbor grid engine 5.6–9.7x faster than dense evaluation, and an SE(3)-equivariant GNN scorer trained on 741,706 co-folded complexes (34% rank-1 pose recovery within 2 angstroms, r=0.407 held out, 0.467 to crystal structures, 0.690 for a PDBbind-trained model) — and, unusually, a within-target 30-compound GABA-A series where the empirical scorer places 8th of 25 methods ahead of every Vina and Vinardo configuration while the GNN scores below Vina, with the authors stating plainly that the model is unsuitable for pose rescoring; Clearmind’s published US patent application on MDMA combined with palmitoylethanolamide for PTSD, anxiety and eating disorders, notable less for the filing than for what prompted it — FDA’s July 13, 2026 guidance ‘Psychedelic Drugs: Considerations for Clinical Investigations’, which explicitly folds MDMA into the psychedelic category as an entactogen and specifies trial design, conduct, safety monitoring and NDA-enabling expectations; FDA clearance of Roche’s Elecsys phospho-tau 217P plasma assay, developed with Eli Lilly, the first single-biomarker blood test supporting both rule-in and rule-out of amyloid pathology on the same validated cutoffs across primary and specialty care in patients 55 and older, running on more than 4,500 installed cobas analyzers, with a three-tier negative/intermediate/positive output, explicitly non-standalone and not validated for predicting future dementia or therapy monitoring; a Nature Mental Health binge-eating GWAS (39,279 cases, 1,227,436 controls, six loci) whose loci and genetic correlations sit with smoking, risk tolerance and problematic alcohol use rather than metabolic traits, where GWAS-by-subtraction leaves only 12% of binge-eating and 10% of anorexia genetic variance shared with BMI, and where the pharmacology-relevant results are uniformly negative — no significant drug sets for either phenotype, a single drug-class hit (antimigraine preparations for anorexia), no significant tissue or cell-type heritability enrichment, PRS explaining 0.32% and 2.32% of liability variance, and a 64% increase in effective anorexia sample size buying two new loci while three previously significant ones dropped out; a 61-sample Molecular Psychiatry gut-brain study pairing stool metagenomics with MRS in anterior cingulate, dlPFC and inferior occipital gyrus, where local FDR sits between 0.23 and 0.55 on essentially every association, several effects flip sign between regions, and the models predict pathway abundance from brain metabolites rather than the reverse; a 30-operator perturbation audit of medical chain-of-thought across 14 LLMs, four benchmarks and ~364,000 perturbed generations, whose joint chain-update x answer-flip analysis yields a 72.9% chain-decoupling rate on clinically destructive edits, with chain corruption leaving accuracy unchanged, CoT removal leaving accuracy unchanged, two board-certified clinicians confirming 98.5% of edits leave the gold answer defensible, and 17.3–33.3% of destructive flips judged clinically harmful; BenchBench-Protocol, 149 wet-lab protocol-modification tasks recovered from real scientist edits to published protocols across nine domains with weighted expert-reviewed rubrics, where Claude Opus 5 leads at 59.2% against 34.1–47.1% for eight other models, roughly a quarter of rubric credit is still unearned at best-of-ten, model ranking is unstable in k, and performance falls specifically on troubleshooting and adaptation — the tasks needing tacit expert judgment; and the compaction cliff, where a production compaction prompt preserves 53% of agent safety rules after one round and 10% after five because a safety constraint and an episodic log are compressed at the same rate, fixed by type-routed retention that preserves 2–4x more rules at every ratio with 96% recall over five rounds. https://doi.org/10.1002/cpt.70453 2026-08-26-receptor-and-reason Wed, 26 Aug 2026 12:00:00 +0000 1044 Wednesday, August 26, 2026, eleven segments. LEAD: for a receptor embedded in a membrane, the membrane may be the biophase. An extended CNS physiologically based pharmacokinetic model with D2 binding kinetics, internalization and endogenous dopamine competition, compared against PET occupancy in 66 people across eight studies, finds brain cell membrane concentrations align better with observed occupancy than brain extracellular fluid for five of eight antipsychotics — and lipophilicity does not tell you which five, since the most lipophilic compound in the set is one of the two best described by extracellular fluid. A membrane partition coefficient beats logP; internalization helps two drugs and hurts five; and simulated phasic dopamine at fifty times baseline moves occupancy not at all, because dopamine binds D2 orders of magnitude more weakly than the drugs. Then: chronic stress disinhibiting laterodorsal tegmental cholinergic neurons by suppressing local somatostatin interneuron input, with activation of those interneurons during stress protecting the synapse and the behavior; sparse autoencoder features from InterPLM outperforming dense ESM-2 embeddings and letting plain logistic regression on 320 features beat two dedicated neuropeptide predictors on the human secretome; PandaDock, an open-source docking platform whose most valuable page is the within-target GABA-A series where its own neural scorer loses to AutoDock Vina; FDA’s July psychedelic-trials guidance folding MDMA in as an entactogen, and the patent filings arriving in its wake; Roche’s pTau217 plasma assay cleared as the first single-biomarker test supporting both rule-in and rule-out on the same cutoffs, on already-installed instruments; the first powered binge-eating GWAS, whose six loci sit with impulse-control traits rather than metabolic ones, and whose drug-set, tissue-enrichment and polygenic-prediction results are all negative; a 61-sample gut-brain neurochemistry study where almost nothing survives correction and effects flip sign between regions; a perturbation audit finding that medical chain-of-thought is decoupled from the answer in 73% of clinically destructive edits, with clinician validation that the edits are real; BenchBench-Protocol, built from real wet-lab protocol diffs, where the ranking of models is not stable in the number of attempts; and the compaction cliff — safety rules surviving at 53% after one compaction round and 10% after five, because a penicillin allergy and a debugging trace are compressed at exactly the same rate. false Daily roundup for August 25, 2026 — twelve segments. LEAD: a Nature Medicine multiomic atlas of the adult human hippocampus in unmedicated major depressive disorder — single-nucleus RNA plus ATAC from the same nuclei, spatial transcriptomics at two resolutions and regional proteomics, 495,037 nuclei from 11 MDD and 19 controls after QC. Two modalities independently recover a neurogenic lineage from quiescent stem cells through activated stem cells, intermediate progenitors, doublecortin-positive neuroblasts and two immature granule cell states, with a bifurcation to mature granule cells or a persistently immature state and a reverse path implying granule cell dematuration (previously described in marmoset); the immature-marker cells sit in the granule cell layer rather than the subgranular zone. In MDD the lineage is halted rather than depleted: more quiescent stem cells, fewer neuroblasts, confirmed by fewer nestin/Ki67 subgranular zone cells and fewer doublecortin-positive cells, with the persistently immature population unchanged — the control that separates a jammed pipeline from tissue-wide degeneration. The earliest-rising module in the depressed trajectory is interferon signaling, in the compartment that should be dividing; HTR4 and HTR1E are down in intermediate progenitors, HTR7 in immature granule cells, and 5-HT2A/5-HT2C across spatial subfields, so the receptors an SSRI works through indirectly are themselves downregulated in the target cells; and a previously undescribed medium-spiny-neuron-like PENK-high GABAergic interneuron is the most dysregulated cell type in the dataset, with elevated immediate-early genes suggesting compensation for a stress-hyperactivated dentate gyrus. Also: a four-rat preprint where 10 mg/kg psilocybin leaves ripple rate untouched but lengthens sharp-wave ripples and lowers peak frequency in every animal, raises sharp-wave amplitude in three of four (implicating the CA3→CA1 limb), and produces immobility filled with high-voltage spindles instead of NREM — long ripples being what novelty and memory tasks produce and what optogenetic prolongation shows to improve memory, with the honest confound that the first saline session also had almost no sleep; a randomized double-blind placebo-controlled fMRI trial (75 enrolled, 62 analysed, healthy volunteers, placebo-placebo vs placebo-ketamine vs lamotrigine-ketamine) where a hybrid default-mode-plus-frontoparietal coactivation pattern tracks rumination at baseline, ketamine acutely cuts its occurrence in proportion to baseline rumination, the effect is gone at 24 hours, and lamotrigine attenuates only nominally; a Nature Neuroscience study where 14 days of isolation escalates alcohol drinking in male mice and reduces it in females, with BLA→mPFC projection neuron excitability rising in males and falling in females yet tracking intake in both, the projection encoding alcohol better than undifferentiated BLA activity, and chemogenetic drive or inhibition moving intake accordingly (rank predicted alcohol but not water intake); a $6.5M NIDA award moving PleoPharma's once-daily oral CB1-pathway candidate into phase 3 for cannabis withdrawal after a phase 2b that hit its primary endpoint with a dose-response signal, in an indication with no approved pharmacotherapy for 19 million affected Americans; FDA acceptance with priority review of Teva's NDA for ecopipam, a selective D1/D5 antagonist that would be Tourette syndrome's first novel mechanism in over 50 years, on a significant week-12 YGTSS total tic score improvement and a 53% relapse-risk reduction in randomized withdrawal — an enriched-responder design; triple-color spectrally resolved fiber photometry in dorsolateral striatum showing dopamine is uncorrelated with speed and turning (and even rises after locomotion stops) but predicts spiny projection neuron output relative to glutamate input with opposite signs across pathways inside a 200ms-before to 100ms-after window, with methylphenidate raising direct-pathway gain and reserpine reversing it, and an asymmetry consistent with D2's higher apparent affinity; a Nature Medicine study on the Lothian Birth Cohort 1936 (IQ at 11, testing from 70, postmortem donation, 866 with follow-up) where severe cognitive decline carries abnormally thick myelin on large-diameter axons, a shift toward smaller myelinated calibres and INCREASED oligodendrocyte density, independent of postmortem interval, brain pH and Alzheimer's or small-vessel pathology, traced to reduced oligodendrocyte NRF2 and reproduced by oligodendrocyte-specific NRF2 deletion in mice — giving dimethyl fumarate a defined cell type; GUIA, a research-agent network on the agent-to-agent protocol with 97 tools and ten agents including third-party Pathology-o3, ChemCrow, CRISPR-GPT and Eubiota, scoring 78% vs 56% on a 720-question biomedical benchmark but jumping 45% to 93% on a microbiome benchmark only when the specialist agent arrives (and only 'subtly' on MedQA where it was already strong), with case studies including a KCNK9 virtual screen to 19 hits, a survival-separating NSCLC spatial subtyping, and CASZ1 nominated as a novel androgen receptor coregulator then validated by knockdown, ChIP-qPCR and proliferation; K-Bench 01, built from 178 verbatim first-turn requests with attachments and no reference answers, run under nine frontier models in identical sandboxes for 1,602 runs and ~40,000 judgments by three identity-blinded judges with read access to the output files, where no model clears the 'a domain scientist would accept this with minor edits' anchor under all three judges, scientific accuracy trails communication by 1.11 points within every model, overclaiming is the top failure tag at 31.4%, and 47.9% of runs finish with no file on disk — vendor-authored, with LLM judges; and MIND, a retrieval-grounded model trained only on curated patient-education material, which answered 38 of 50 escitalopram questions and swept an automated rubric on accuracy, clarity, completeness, nuance, safety and referral appropriateness, only for ten board-certified psychiatrists to rate ChatGPT more accurate (79% vs 72%, negligible effect size), equally safe, less complete, and preferred 58% of the time (72% among early-career raters) — because shorter answers contain less to mark wrong; closing with an npj Digital Medicine perspective proposing peer-reviewed machine-readable interpretive metadata plus a plain-language claims-and-scope label, motivated by ~5,000 AI summaries across ten models over-generalizing even when prompted for accuracy and by a nonsense phrase from an AI summary now repeated in 20-plus peer-reviewed papers. https://www.nature.com/articles/s41591-026-04571-8 2026-08-25-receptor-and-reason Tue, 25 Aug 2026 12:00:00 +0000 1106 Tuesday, August 25, 2026, twelve segments. LEAD: adult human hippocampal neurogenesis in depression is stalled, not emptied. A Nature Medicine multiomic atlas — 495,037 nuclei, unmedicated MDD versus controls, single-nucleus RNA and ATAC plus spatial transcriptomics and proteomics — recovers a neurogenic lineage and a granule cell dematuration path, then finds more quiescent stem cells and fewer neuroblasts in MDD, with protein validation and, critically, no change in the persistently immature population. The earliest module up is interferon signaling, and HTR4, HTR1E, HTR7, 5-HT2A and 5-HT2C are all down in the cells the neurogenic hypothesis says an antidepressant should be reaching. Then: psilocybin lengthening hippocampal sharp-wave ripples and lowering their frequency in all four rats while blocking NREM and filling the recording with high-voltage spindles — the ripple signature of learning without sleep; ketamine cutting occupancy of a rumination-linked default-mode-plus-frontoparietal coactivation state in healthy volunteers, transiently, with lamotrigine only nominally attenuating; social isolation raising BLA-to-mPFC excitability in males and lowering it in females while tracking drinking in both directions; a NIDA-funded phase 3 push for a CB1-pathway candidate in cannabis withdrawal; ecopipam's NDA accepted, a D1/D5 antagonist that would be Tourette's first new mechanism in fifty years; triple-color photometry showing dopamine is a pathway-specific gain term rather than a movement code, moved in opposite directions by methylphenidate and reserpine; the Lothian 1936 cohort tying severe cognitive decline to more oligodendrocytes and thicker myelin on large axons via oligodendrocyte NRF2 loss, reproduced in mice and handing dimethyl fumarate a cell type; GUIA's agent-to-agent network, whose 45-to-93-percent benchmark jump is the specialist's tools arriving rather than better reasoning, though CASZ1 survived real wet-lab validation; K-Bench, where scientific accuracy trails communication by 1.11 points in every one of nine models, overclaiming leads the failure tags, and half of all runs leave no file on disk; and a psychiatry patient-education model that sweeps the automated rubric and then loses the psychiatrists 58 to 42, rated less accurate precisely because it answered more completely. false Daily roundup for August 24, 2026 — eleven segments. LEAD: Tim Hales's group at Dundee closes a 25-year loose end in opioid tolerance. Morphine tolerance needs the mu receptor, beta-arrestin2 and c-Src; the c-Src leg rested on promiscuous inhibitors (dasatinib, PP2). Using the conformationally selective c-Src inhibitor eCF506/NXP900, daily oral dosing before morphine held the antinociceptive ED50 at 2.6 mg/kg over ten days where vehicle animals went from 1.5 to 9.7, and sustained analgesia in mu-heterozygous mice that lost the response by day 4. Mechanistically it halves agonist efficacy for beta-arrestin2 recruitment (DAMGO 100 to 56 percent, morphine 33 to 14) with no change in potency and no change in cAMP inhibition — even after beta-funaltrexamine strips receptor reserve to expose hidden efficacy loss. A cereblon PROTAC degrading c-Src reproduces it, dose-dependently with residual c-Src. Across six mu agonists, c-Src Y416 phosphorylation correlates with beta-arrestin2 efficacy at r=0.90, placing the kinase inside the arrestin arm. The unexplained wrinkle: Src phosphorylation is blocked in under an hour but arrestin recruitment doesn't move until 8 hours (4 for the degrader), with total beta-arrestin2 and surface receptor unchanged. Established tolerance was only transiently reversed. Also: an IGNITE ADOPT-PGx secondary analysis where tramadol added to oxycodone or hydrocodone doubles cumulative morphine milligram equivalents (122 vs 205; 79 vs 206) with indistinguishable pain scores and worse mobility, robust to CYP2D6-normal and single-site restriction; a ten-year claims target trial emulation in 86,290 post-breast-surgery patients finding gabapentin plus opioids comparable on respiratory complications, MACE and arrhythmia but with no reduction in subsequent high-dose opioid use; a conductance-based model showing SK-mediated deep after-hyperpolarization recruits Kv4 and confines the interspike trajectory to a narrow inter-nullcline corridor that damps channel noise, absent in the shallow-AHP accumbens-shell subpopulation, with the same AHP-then-ramp signature documented in dorsal raphe, locus coeruleus and tuberomammillary neurons — making SK blockade a regularity drug rather than a rate drug; iPSC cerebral organoids of Williams syndrome and its reciprocal duplication imaged day 30 to 150, diverging oppositely in rosette morphogenesis, with Dup7 never establishing network synchrony and maraviroc raising spine density in control and Williams but not Dup7; azobenzene liposomes (azosomes) delivered through optofluidic cannulas for light-triggered oxytocin release in awake mice, verified with a genetically encoded sensor and blocked by ornithine vasotocin analog, plus the sobering control that the sensor needed micromolar peptide in vivo against a 20.5 nM in vitro EC50; brain-wide FISH raising proglucagon neuron populations from three to nine, with the posterior hypothalamic/interfascicular group upregulated by 30 h fasting like an AGRP neuron despite making an anorexigenic peptide, opposite to medullary GLP-1 neurons, and Cre driver lines that disagree about which populations they label; a single-cell PBPK tutorial whose most useful output is the negative midazolam result and the resulting criterion for when the added ODEs buy nothing; the AIntibody challenge, the first blinded prospective antibody design benchmark with every submission expressed as full IgG and tested by HT-SPR, KinExA and a five-assay developability panel, where 13 percent of affinity-maturation submissions gained 20-fold and the winner tied the best experimental antibody at 94.7 pM, but eight participants produced no developable binder, a no-ML consensus method placed third, and at cluster-level hit expansion the algorithms beat the abundant-clone baseline only 9.8-13.8 percent of the time against 39 percent for random picking; an ICML GenBio workshop paper where LLM-written multi-metric ranking policies shortlist protein binders at 0.589 recall@10 versus 0.571 for the best single existing confidence score, selling interpretability rather than accuracy; and an ontology of ten therapeutic moves validated by five licensed psychologists, showing models over-inquire at up to three times the human clinician rate, carry forward human-initiated strategies without initiating their own, swing to over-interpretation and abandon skill building when leading a session, and close roughly half the distributional gap when the moves are exposed as tools with no fine-tuning. https://bpspubs.onlinelibrary.wiley.com/doi/10.1111/bph.70647 2026-08-24-receptor-and-reason Mon, 24 Aug 2026 12:00:00 +0000 1043 Monday, August 24, 2026, eleven segments. LEAD: rather than engineer a G-protein-biased mu agonist, delete the kinase that feeds the arrestin arm. A conformationally selective c-Src inhibitor already in oncology development prevents morphine antinociceptive tolerance in mice (ED50 2.6 vs 9.7 mg/kg at day ten), halves DAMGO and morphine efficacy for beta-arrestin2 recruitment without touching potency or cAMP inhibition even with receptor reserve stripped, and is reproduced by a c-Src PROTAC degrader; c-Src Y416 phosphorylation correlates with each agonist's arrestin efficacy at r=0.90, though the 8-hour lag between kinase inhibition and the arrestin effect is unexplained, and established tolerance is only transiently reversed. Then: tramadol added to oxycodone or hydrocodone doubling cumulative morphine milligram equivalents with no analgesic gain and worse mobility; a target trial emulation finding gabapentin-plus-opioids safe but not opioid-sparing; a phase-space account of why slow pacemaking survives single-channel noise, via an SK-deep after-hyperpolarization that recruits Kv4 — generalizable to dorsal raphe, locus coeruleus and tuberomammillary neurons, and making SK blockade a regularity manipulation rather than a rate one; reciprocal Williams and Dup7 organoids diverging in opposite directions from day 30, with Dup7 never synchronizing and maraviroc working in Williams but not Dup7; photoswitchable liposomes releasing oxytocin into CA2 in awake mice, plus a hundred-fold in-vivo-versus-in-vitro gap in the oxytocin sensor; nine brain proglucagon populations instead of three, the posterior hypothalamic one fasting-induced like an AGRP neuron while making an anorexigenic peptide; single-cell PBPK and the criterion for when not to build one; the AIntibody blinded prospective benchmark, where affinity maturation genuinely works, a no-machine-learning consensus method placed third, and cluster-level hit expansion lost to random clone picking; LLM-written ranking policies for binder shortlisting that beat the best single existing metric 0.589 to 0.571; and a validated ontology of ten therapeutic moves showing what language models actually do in a therapy session — over-inquiry, over-interpretation, no skill building — and that exposing the moves as tools halves the gap without fine-tuning. false Daily roundup for August 23, 2026 — eleven segments threaded on dose, meaning how much and for how long before the sign of the effect flips. LEAD: Pascal Fossat's group at Bordeaux settles a fifty-year argument about descending serotonergic pain control. Both chemogenetic activation and chemogenetic inhibition of nucleus raphe magnus serotonin neurons lower paw withdrawal threshold; silencing also raises C-fiber-evoked spiking and windup, so the pathway is tonically analgesic. Holding frequency at 5 Hz and cutting pulse width from 100 ms to 5 ms converts hyperalgesia into analgesia in the same animals, and whole-cell recordings show one homogeneous population rather than two. 5-HT3 on glutamatergic dorsal horn neurons carries the hyperalgesia (abolished by intrathecal granisetron); 5-HT2C and 5-HT2A on Pax2-positive inhibitory interneurons carry the analgesia, with 5-HT2C alone responsible for the tonic component and enriched to comparable levels in human dorsal horn. Also: an exact-residue Ballesteros-Weinstein atlas of 86 human opioid receptor structures where state transition barely moves global contact similarity while rewiring a third more contacts, 85% of them away from the ligand — and where 32 association tests all fail global correction and the authors say so; a fluoxetine PBPK model showing the geriatric exposure difference emerges only after 30-50 days while CYP2D6 phenotype leaves total active moiety nearly unchanged; DrugEvolve, a multi-role LLM system that evolves algorithms across 11 drug-development tasks and 120 test sets for roughly 180 GPU-hours apiece, with no generalist-agent baseline; a cell interaction foundation model run as a generative tissue simulator that catches CXCL9's regulatory-T-cell side effect and nominates CXCL10 plus INHBC knockdown, validated in vitro against anti-PD-1; GEOMeta's finding that flash-tier open models match reasoning models on metadata curation at a tenth the cost, and that public transcriptomics reports sex for 32% of samples and dose for 19% of chemical perturbations; in vivo microdialysis metabolomics where cocaine erases baseline temperament differences by depleting NAAG in every animal; beta-2-containing nicotinic receptors on supramammillary glutamate neurons that respond to nicotine but not cocaine, fentanyl or a peripherally restricted agonist; insular parvalbumin interneurons under-recruited in adolescent mice and bidirectionally controlling impulsivity and alcohol seeking in adults; a chronic-kidney-disease threshold shift running through central amygdala AT1 receptors; and lateral hypothalamic GABAergic projections to dorsal pons and lateral preoptic area producing the same behaviors from both entry points, with the authors' own Fos data raising antidromic activation as the confound. https://www.biorxiv.org/content/10.1101/2025.04.09.647804v2 2026-08-23-receptor-and-reason Sun, 23 Aug 2026 12:00:00 +0000 1200 Sunday, August 23, 2026, eleven segments threaded on dose — how much, for how long, before the sign flips. LEAD: nucleus raphe magnus serotonin neurons are pro-nociceptive both when silenced and when driven hard, and at a fixed 5 Hz the effect flips from analgesia to hyperalgesia as pulse width goes from 5 ms to 100 ms; 5-HT2C carries tonic analgesia, 5-HT3 carries the hyperalgesia, and 5-HT2C is the most enriched spinal serotonin receptor in human dorsal horn too. Then: an opioid receptor contact atlas where activation rewires 35% more contacts while global similarity moves 2.5%, 85% of the rewiring away from the ligand, and no association test survives correction; a fluoxetine PBPK model showing the geriatric accumulation difference is real but slow (1.15-fold at day 10, 2-fold at day 50) and that CYP2D6 phenotype barely moves total active moiety; DrugEvolve evolving algorithms across 11 drug-development tasks at ~180 GPU-hours each, without a generalist-agent control; a generative virtual tissue model that catches CXCL9's Treg side effect and designs around it, beating anti-PD-1 in vitro; GEOMeta showing flash-tier open models match reasoning models at a tenth the cost while public transcriptomics reports dose for 19% of chemical-perturbation samples; cocaine acting as a neurochemical equalizer in high- and low-responder rats by depleting NAAG; a beta-2 nicotinic substrate in supramammillary glutamate neurons that ignores cocaine, fentanyl and a peripherally restricted agonist; insular parvalbumin interneurons under-recruited in adolescence and bidirectionally gating impulsivity and alcohol seeking; chronic kidney disease priming central amygdala AT1 signaling, a target already drugged in these patients; and lateral hypothalamic GABAergic outputs that are not modular, with the authors' own antidromic Fos data offered against their preferred reading. false Daily roundup for August 22, 2026 — eleven segments, and the first two are an argument about whether domain-specific agentic infrastructure has a future. LEAD: Brain Researcher, from the Poldrack lab at Stanford with collaborators at Georgia State, Texas, Singapore and Queensland — a harness that runs inside the researcher's own environment and forces methodological commitments into the open before execution. A commitment card fixes the question, admissible analyses, required checks and claim scope, sealed with a content hash; a review layer afterward assigns every claim one of six states (accepted, qualified, revised, blocked, rejected, deferred) under explicit adjudication rules; a 746,000-node knowledge graph carries provenance. Benchmarks: first-action correct tool selection 23% to 94% across 60 tasks and seven frontier models, all seven improving; verified grounding of cited evidence 4.6% to 22%, which the authors correctly describe as improved but not solved. The real payload is the collaborator studies. A schizophrenia connectivity audit expanded three pre-specified hypotheses into a 480-specification multiverse, and the headline hypothesis that looked supported in half the contrasts was hiding a complete estimator split — 100% favorable under Pearson and Spearman, 0% under partial correlation and mutual information. And the governance lesson: after a server fault dropped execution to a general-purpose coding agent, that agent scored any significant permutation p-value as favorable REGARDLESS OF SIGN, inflating support for a directional hypothesis; the automated review layer missed it and a human reading the specification curve caught it, after which a directionality test and a fallback warning became permanent checks. In the other two episodes the checks did fire — five pre-specified connectivity-behavior associations rejected and converted into a replication plan, and an underpowered cross-cultural meta-analysis blocked from confirmatory status. Then: Pavlo Dral's group at Xiamen surveying all 49 agentic systems in computational chemistry through August 8, with growth from 0.3 systems per month in 2024 to 4.6 in 2026, arguing the explosion is self-terminating because the generalist coding agents that collapsed the build barrier also collapsed the reason to use a specialist — evidenced by the confession that their own group members copy-paste into a chat window rather than use their lab's free, more capable system; PerturbTrace decomposing closed-loop agent behavior into feedback-to-state, state-to-action and action-to-outcome, finding four agents beat the strongest non-agent baseline on at least 15 of 17 tasks while true feedback gave no consistent advantage over random or absent feedback and only 43 of 576 transitions closed the full chain (abstract only — body not yet rendered); evidence-constrained mechanistic synthesis, which classifies what each finding is mathematically entitled to constrain and shifts ensemble frequencies rather than inventing coefficients or probabilities of biological truth, demonstrated on chronic spontaneous urticaria with 114 atomic findings, 18 executable constraints, a frozen 4,096-hypothesis ensemble, 13 findings flagged as contradictory and 29 relations left unresolved rather than declared absent; a Wuppertal benchmark where native molecular-language-model embeddings lose to plain fingerprints across six virtual libraries until the encoders are fine-tuned on the target library; the first electrophysiology of A11 dopamine neurons, the sole source of spinal dopamine, showing sodium-leak-driven pacemaking with no hyperpolarization-activated current, and a mu-opioid agonist that both disinhibits them presynaptically and directly slows their firing via potassium conductance — reconciled by the direct effect desensitizing within minutes while presynaptic mu receptors resist desensitization; a zona incerta to lateral habenula circuit in a single-prolonged-stress-plus-shock PTSD model that rescues anxiety and aversion but not depression, splitting downstream into separate VTA-projecting (aversion) and rostromedial-tegmentum-projecting (anxiety) populations, with the honest control that non-selective habenular output produces aversion and no anxiety at all; three symptom trajectories in 81 medication-naive first-episode psychosis patients, where 31% are delay responders reaching the same endpoint only after week 12 and executive control network connectivity change at week six predicts response while baseline does not; a multimodal microbubble focused-ultrasound safety study showing barrier opening occurs at every dose including the harmless ones, astrocyte activation is not predictive either, targeting is part of the dose, and gross performance recovers in a week while microhemorrhage and edema persist at least three; mitragynine entering an NIH HEAL phase 1 for opioid use disorder weeks after DEA moved to schedule its metabolite, with the wrinkle that a Maastricht group had already run a human phase 1 showing stimulant effects at 5 mg and amnesia plus mild psychopathology at 40; and a pre-IND clearance for a first-in-class teneurin C-terminal associated peptide neuropeptide in generalized anxiety disorder — no efficacy signal, but a genuinely new target class. https://arxiv.org/abs/2608.19902 2026-08-22-receptor-and-reason Sat, 22 Aug 2026 12:00:00 +0000 1120 Saturday, August 22, 2026, eleven segments. The first two items are an argument. LEAD: Brain Researcher, from the Poldrack lab at Stanford — not an autonomous scientist but a harness that runs inside the researcher's own computational environment and forces methodological commitments into the open. The framing is that an analysis is not a claim: an agent can run a pipeline, get a number and declare success, and every failure mode we know from human-executed science comes back dressed in tool calls — selective analysis, premature completion, optimizing a criterion misaligned with the claim. Neuroimaging is the right proving ground because when seventy teams analyzed one dataset, no two used the same workflow. Before execution, a commitment card fixes the question, the admissible analyses, the required checks and the claim scope, sealed with a content hash so later edits are detectable; afterward a review layer writes a claim card assigning one of six states — accepted, qualified, revised, blocked, rejected or deferred — under explicit adjudication rules, over a 746,000-node provenance graph. Benchmarks: first-action correct tool selection 23% without the harness to 94% with it across 60 tasks and seven frontier models, all seven improving; verified grounding 4.6% to 22%, nearly fivefold and still leaving most evidence rows failing, which they say plainly. The payload is the three collaborator studies and specifically the one that went wrong: a 480-specification multiverse on schizophrenia functional connectivity where the headline hypothesis looked supported in half the contrasts but was hiding a complete estimator partition — 100% favorable under Pearson and Spearman, 0% under partial correlation and mutual information, making it estimator-regime-dependent rather than robust. And after a server fault dropped execution to a general-purpose coding agent, that agent scored any specification with a significant permutation p-value as favorable regardless of sign, inflating support for a directional hypothesis to near-universal; the automated review layer did not flag it, a human reading the code and specification curve did, and a directionality test plus a generalist-fallback warning became permanent checks. In the other two episodes the checks fired: a 36-specification multiverse rejected all five pre-specified connectivity-behavior associations and converted the null into a replication plan, and an underpowered cross-cultural meta-analysis with six to eight studies per subgroup was blocked from confirmatory status rather than written up. Then the counterargument, published two days earlier from Pavlo Dral's group at Xiamen: a Perspective surveying all 49 agentic systems for computational chemistry through August 8, with growth from four systems in 2024 to twelve in 2025 to thirty-three in seven months of 2026. Their claim is that the explosion is self-terminating, because the same general-purpose coding agents that collapsed the barrier to building a specialist also collapsed the reason to use one — and the evidence is a confession: they released their own systems free and uptake was slower than expected, including inside their own group, where some members did not realize their lab's agent could fix scripts and instead copy-paste into a chat window and back to the cluster by hand. If the people closest to a free, capable system reach for something else, the binding constraint is not capability. The reconciliation I would offer is that Brain Researcher's value is not tool routing — a generalist will get there — but the commitment card, the claim states, the multiverse and the audit bundle, which are constraints rather than capabilities, and constraints do not arrive by a model getting smarter. Then PerturbTrace, which measures exactly the failure the lead is built to prevent: three-link decomposition of each round-to-round transition, four agents on seventeen screen-derived tasks, every agent beating the strongest non-agent baseline on at least fifteen, and yet no consistent advantage from true feedback over random or absent feedback, with only 43 of 576 transitions — about 7.5% — closing the full chain and 25 of those 43 under random feedback (abstract only; the body has not rendered yet). Then evidence-constrained mechanistic synthesis for the pre-calibration stage of drug development: a finding that X affects Y buys an edge, a directional experiment buys the sign, an ordinal result buys an inequality, a confidence interval buys a bound, and only a calibrated concentration-response experiment buys a magnitude — with evidence shifting the frequency of a relation in a reproducible hypothesis ensemble rather than becoming invented coefficients or probabilities of biological truth, dependency clusters preventing one paper's ten readouts from counting as ten replications, and a replication ceiling stopping one significant study from consuming a relation's whole information budget; demonstrated on chronic spontaneous urticaria with 114 atomic findings from 53 sources, 18 executable constraints, a frozen 4,096-hypothesis ensemble, 13 findings flagged contradictory, 29 of 43 relations left unresolved rather than declared absent, and objective changes that alter the selected control vector without touching the evidence ensemble. Then a Wuppertal benchmark of four pretrained molecular language models across six virtual libraries, where native embeddings vary wildly and plain fingerprints are consistently strong until domain adaptation flips it. Then the day's best wet-side paper, from Stephanie Gantz's group in Neuropsychopharmacology: the first electrophysiological characterization of A11 dopamine neurons, fewer than two hundred cells in dorsal posterior hypothalamus and the only known source of dopamine to the spinal cord, relevant to nociception, itch, restless legs syndrome and migraine. Seventy percent project to lumbosacral cord, 79% pacemake at about 5 Hz with no detectable hyperpolarization-activated current, and the subthreshold drive is a persistent sodium leak of about 60 pA. Prior work proposed that morphine raises spinal dopamine by disinhibition, and that is confirmed — a mu agonist cut inhibitory current frequency and amplitude — but the same agonist also acted directly, activating potassium conductance, erasing the inward drive and slowing pacemaking outright, so the net direct effect is inhibitory. The reconciliation is kinetic: the direct effect desensitized within minutes while presynaptic mu receptors resist acute desensitization, which is satisfying and also an untested hypothesis. Also: 92% of neurons from male mice were spontaneously active versus 61% from females, in a system whose clinical conditions are more prevalent in women. Then a zona incerta to lateral habenula circuit paper in Molecular Psychiatry using single prolonged stress plus shock, where weakened GABAergic input fits the standard habenular-hyperexcitability story and chemogenetic re-activation rescues anxiety, fear and aversion while doing nothing for depression-like measures; downstream the recipient habenula neurons split into largely separate VTA-projecting and rostromedial-tegmentum-projecting populations, with VTA terminals driving aversion and no anxiety and rostromedial terminals driving anxiety and mild aversion — and the control that increases my confidence is that non-selective excitatory habenular output to either target produced aversion but no anxiety at all, making the anxiety effect specific to the zona-incerta-defined subpopulation. All male animals, and the authors are appropriately suspicious that the rostromedial aversion may be spillover through the known projection to VTA. Then an antipsychotic trial-design finding from Adrienne Lahti's group in 81 medication-naive first-episode psychosis patients over sixteen weeks: three trajectories, with half fast responders, 31% delay responders who reach the same endpoint but only stabilize after week 12, and 19% partial responders — meaning a standard six-to-eight-week acute trial scores a third of eventual responders as failures. Baseline executive control network connectivity did not predict response; change from baseline to week six did, at r about 0.34, climbing in delay responders and falling in partial ones. Then a preclinical microbubble focused-ultrasound safety study combining MRI, histology, acoustic emission dosimetry and kinematic motor readouts in rats: barrier opening occurred at every dose including harmless ones, so it is not a safety readout; astrocyte activation was comparable across groups of which only one had microhemorrhage; the same acoustic parameters aimed at cortex alone versus cortex plus striatum produced different microbubble activity and deficits only when the task-relevant subcortical circuit was in the field, so targeting is part of the dose; and recovery time courses came apart, with gross performance resolving in about a week, fine-grained movement changes persisting two weeks in a subset, and microhemorrhage and edema persisting at least three — meaning a single acute imaging timepoint will both miss transient functional harm and overstate persistent structural harm. Then mitragynine, the principal kratom alkaloid — partial mu agonist, competitive kappa and delta antagonist, plus adrenergic and serotonergic binding — entering an NIH HEAL-funded double-blind placebo-controlled single-ascending-dose phase 1 for opioid use disorder in about 32 healthy volunteers at 25 to 100 mg, with four days of inpatient observation, ECG and corrected QT, pupillometry and sedation scales; the odd regulatory backdrop of DEA moving weeks later to schedule the metabolite 7-hydroxymitragynine as Schedule I in synthesized concentrated form only; and the correction that NIH's claim the isolated compound had never been studied in humans was already wrong, a Maastricht phase 1 in fifteen volunteers having found no movement on most outcomes but increased arousal, sustained-attention accuracy and motor inhibition at 5 mg and amnesia ratings plus mild psychopathology at 40 — the classic kratom double-action claim with human data behind it and a real prior for dose selection. Closing on a marker rather than a result: pre-IND feedback with no clinical hold issues for a first-in-class neuropeptide acting on the teneurin C-terminal associated peptide pathway in generalized anxiety disorder, aiming at the central stress response rather than downstream symptom neurotransmission, with an IND targeted before year end and first patient in the first half of 2027 — existing single- and multiple-ascending-dose data generated outside the United States and not under an IND, no efficacy signal, but a genuinely new target class after two decades of recycled serotonergic and GABAergic pharmacology. false Daily roundup for August 21, 2026 — ten segments. LEAD: MASCOT, a multi-agent molecular optimization framework whose interest is architectural discipline plus a real endpoint. Molecular editing, scoring, constraint enforcement and acceptance stay inside an explicit Monte Carlo graph-editing executor; three LLM agents sit above it and control only how the search proceeds — a trade-off agent reallocating objective weights, a strategy agent tuning the fragment-addition probability, a reflection agent distilling delayed decision-outcome records into shared memory — with every output schema-checked and clipped, and malformed calls retaining the previous value. Best ablation in the paper is a shuffled-memory control that keeps the reflection agent's schedule, prompts and lesson count but randomly re-pairs decisions with the wrong outcomes, separating genuine online learning from the benefit of merely adding an agent. Applied to remimazolam with the scaffold fixed, it prioritized RM-1 (microsomal half-life 19.7 to 6.5 min, mouse ED50 38 to 24 mg/kg, therapeutic index 6.6 to 9.9); derivatization then found RM-7 at an INTERMEDIATE half-life of 12 min, because the three fastest-clearing derivatives were useless — ED50 14.8 mg/kg, walking recovery 5.7 to 2.1 min, therapeutic index 24, stronger GABA-A potentiation, preserved flumazenil reversal, and steadier rabbit hemodynamics across six-hour infusions. Then: GPR139 profiled across fifteen G alpha subtypes plus G protein knockout panels, where broad inhibitory/G-q/G-twelve coupling coexists with a stimulatory arm invisible to proximal BRET and an opposing inhibitory cAMP response that only appears once the stimulatory subunit is deleted, and TAK-041 shows reduced maximal responses in arrestin-3 and cAMP specifically; dual retrograde tracing, input-output rabies tracing and paired whole-cell recordings showing the canonical prefrontal-reuniens-hippocampal relay cannot work as drawn, with posterior and lateral hypothalamus supplying the hippocampal-projecting population instead; GluD1 knockout selectively losing GluN2B-containing NMDA responses, LTP and contextual memory at ventral subiculum with GluN2A untouched; a Molecular Psychiatry review making COMT inhibition the cleanest worked example of an inverted-U drug — 60% of frontal dopamine turnover versus 15% striatal, tolcapone helping Val homozygotes and hurting Met homozygotes, extending to delay discounting, gambling and alcohol use disorder — against the constraint that the only brain-penetrant licensed inhibitor is hepatotoxic in exactly the metabolically compromised populations you would target; the first site-specific human demonstration that transcranial ultrasound modulates a white matter tract, with an FA-by-pressure interaction arguing for anatomy-informed dosing and ~30% grey-matter spillover as the honest caveat; an NGF pain dissection where mechanical allodynia and cold hypersensitivity need Schwann cell p75 feeding TRPA1, mitochondrial ROS and NOX1 while acute nociception and heat hyperalgesia need neuronal TrkA and TRPV1; a 12-model benchmark across 176 protein-ligand systems where validity is saturated, usable yield ranges from 1% to 98% of targets, and the strongest docking model passes multi-parameter filters 2% of the time; a scoping review finding hallucination and sycophancy harms extensively theorized and never measured against symptom outcomes; and a kava result reported as the partial failure it is. https://www.biorxiv.org/content/10.64898/2026.08.17.745149v1 2026-08-21-receptor-and-reason Fri, 21 Aug 2026 12:00:00 +0000 1139 Friday, August 21, 2026, ten segments. LEAD: the paper I have been waiting for someone to write — a multi-agent system that runs from molecular search all the way to a rabbit under continuous infusion. MASCOT inverts the usual agentic-chemistry design: instead of asking a language model to propose structures, molecular editing, property scoring, constraint enforcement and acceptance all stay inside an explicit, auditable Monte Carlo graph-editing engine, and three LLM agents sit above it controlling only how the search proceeds. A trade-off agent reallocates weights across competing objectives; a strategy agent adjusts a single number, the fragment-addition probability, tuning exploration against refinement; a reflection agent converts delayed decision-outcome records into compact shared memory. Outputs are schema-checked and bound-clipped, and a malformed call simply retains the previous value — the agents never touch a molecule. Against a no-agent control the full framework hit its target metric about 55 steps earlier and converged higher, and five reduced controllers each captured only part of the gain; the standout is a shuffled-memory condition that preserves the reflection agent's call schedule, prompt structure and lesson count while randomly re-pairing each decision with the wrong outcome, which isolates genuine online learning from the benefit of merely adding another agent to the diagram. Applied to remimazolam with the scaffold held fixed and edits restricted to side chains, optimizing shorter predicted half-life, greater blood-brain barrier penetration and lower predicted toxicity, it prioritized one candidate out of ~2,700 that held up experimentally: microsomal half-life from about 20 minutes to 6.5, mouse ED50 from 38 to 24 mg/kg, higher brain exposure, therapeutic index from 6.6 to about 10. Then the lesson about proxies — side-chain derivatization produced six more compounds and the SHORTEST half-lives were the worst anesthetics, three clearing in under three minutes and useless. The winner sits at an intermediate 12-minute half-life, cuts the effective dose to about 15 mg/kg, shortens time to resumed walking after loss of righting reflex from 5.5 minutes to 2, potentiates GABA-A currents more strongly than the parent, stays fully flumazenil-reversible, reaches a therapeutic index of 24, and in rabbits recovers faster and more stably after infusions out to six hours at comparable EEG depth with smaller cardiovascular swings and less QT perturbation. The nonmonotonic half-life-to-recovery relationship is the takeaway: the optimizer was pointed at "shorter half-life" and the compound it needed balanced exposure window against potency and brain penetration, which no single oracle encodes — and the authors are appropriately clear that the oracles are surrogates and that synthesis, candidate selection and interpretation stayed expert-directed. Then: GPR139, the habenula-enriched orphan receptor behind Takeda's TAK-041, profiled across fifteen G alpha subtypes by BRET plus CRISPR knockout panels — broad coupling to the inhibitory family, G-q and G-twelve, no stimulatory coupling in the proximal assay, and yet cAMP stimulation that vanishes in stimulatory-G-protein knockouts and is rescued only by reintroducing that subunit, with an opposing inhibitory cAMP inhibition emerging once the stimulatory arm is gone, so the net sign depends on which G proteins your cell line expresses; the canonical prefrontal-to-hippocampal reuniens relay failing three independent tests, with only 2% of neurons dual-projecting, medial prefrontal input to the hippocampal-projecting population almost absent, paired recordings from neighbouring cells under identical stimulation confirming it, and posterior and lateral hypothalamus as the actual dominant input; GluD1, the delta-type glutamate receptor implicated in schizophrenia and intellectual disability, acting as a subtype switch that maintains GluN2B-containing NMDA receptors, LTP and contextual memory at the ventral CA1-to-subiculum synapse; a review synthesizing COMT inhibition into the cleanest available inverted-U pharmacology and blunt about tolcapone's hepatotoxicity landing hardest on exactly the populations worth targeting; the first demonstration that repetitive transcranial ultrasound aimed at the human corticospinal tract raises corticomotor excitability and intracortical facilitation site-specifically while motor cortex stimulation instead reduces short-interval inhibition, with fractional anisotropy moderating the pressure-response slope; an NGF pain dissection splitting the modalities by receptor and cell type and offering a non-neuronal target for the persistent mechanical component; a 12-model, 176-target generative chemistry benchmark showing validity is saturated at 99.8-100%, usable yield differs by two orders of magnitude, and the best-docking model has a 2% median multi-parameter pass rate against 44% for the weakest docker — which agrees with the lead that the value sits in the controller and the staging, not any one generator; a scoping review of 119 articles on LLM chatbot mental health harms where only three studies got past the vignette tier and the AI psychosis literature is six-sevenths conceptual; and a closing mouse study where flavokavain-free kava raises open-field center time without sedation and fails to blunt nicotine's anxiogenic effect. false Daily roundup for August 20, 2026 — ten segments, and the thread through most of them is choosing the right dependent variable. LEAD: more than forty-six thousand single units across thirty-five mice, simultaneous multi-region Neuropixels plus scalp EEG, through the acute hours of psilocybin, with ketanserin as the 5-HT2A dissection (Allen Institute with UC Irvine). Mean firing rate barely moves; burst coding is reconfigured bidirectionally — reticular and anterior thalamus up, geniculate relay nuclei down, hippocampal CA1-CA3 bursting suppressed, striatum up — and ketanserin abolishes nearly all of it, which reads as pharmacological co-optation of the reticular gate rather than a change in excitability. The cortical effects do not survive correction, and the EEG high-gamma and alpha suppression is NOT ketanserin-sensitive. Then a sham-controlled crossover proton spectroscopy study in seventy people with alcohol use disorder where medial prefrontal theta burst lowers glutamine and glycine and dorsolateral 10 Hz does nothing, with e-fields ruling out the boring explanation; an astrocyte STAT3 conditional knockout that produces the hyperglutamatergic prefrontal state and LESS binge drinking, inverting the field's default; the Ingram lab showing that inflammatory hyperalgesia is masked rather than resolved by ongoing 2-AG tone driven by glucocorticoid receptors in the vlPAG, with a narrow therapeutic window for cannabinoid analgesics and GRK inhibition as the way to widen it; the Opendak lab rescuing an infant social deficit with oral chenodeoxycholic acid but not cholic acid, an approved drug acting on a peripheral pathway during a developmental window; anoctamin 2 knockouts with reduced tyrosine hydroxylase and dopamine but unchanged transporter; a macaque D1 challenge where SKF 81297 and SCH 23390 move evidence weighting and working memory decay with large effect sizes and no change in accuracy; a ProteinNPT agent plus robotic lab running twenty-five autonomous rounds on glycoside hydrolase specificity, where the gains come mostly from losing glucose activity; a persistent six-agent fleet with a six-layer trust architecture, honest about where retrieval augmentation makes the model worse, nominating haloperidol as an unvalidated pharmacological chaperone candidate; and a reanalysis showing the MoSeq syllable advantage over scalar behavioral features is largely a preprocessing and hyperparameter artifact. https://www.biorxiv.org/content/10.64898/2026.08.14.744865v1 2026-08-20-receptor-and-reason Thu, 20 Aug 2026 12:00:00 +0000 1088 Thursday, August 20, 2026, ten segments. The thread is choosing the right dependent variable. LEAD: the largest single-unit psychedelic dataset yet — more than forty-six thousand units across thirty-five mice, multi-region Neuropixels with simultaneous EEG, pupillometry and locomotion, through the acute hours of psilocybin, with ketanserin used to dissect 5-HT2A dependence. Psilocybin barely touches mean firing rate. What it reconfigures is bursting, bidirectionally and with an anatomical logic that matches receptor distribution: reticular and anterior thalamic burst rate up, first-order geniculate relay nuclei down, hippocampal CA1 through CA3 bursting suppressed with CA3 hardest hit, striatum up — and ketanserin abolishes nearly all of it. Bursts are a coding channel, not an excitability knob, and the pattern reads as pharmacological co-optation of the reticular thalamic gate, decoupled from corticothalamic feedback. Two honest caveats: the cortical effects, including the deep prefrontal one you would most want, do not survive correction; and the EEG high-gamma and resting-alpha suppression is not ketanserin-sensitive, so the mesoscale biomarker is not 5-HT2A-dependent here. Then: a randomized sham-controlled crossover spectroscopy study in seventy people with alcohol use disorder, where medial prefrontal continuous theta burst lowers glutamine and glycine under the coil and dorsolateral 10 Hz does nothing on any metabolite, delivering dosimetry rather than a direction-of-benefit biomarker; an inducible astrocytic STAT3 knockout that loses GLT-1, produces larger prefrontal excitatory currents, and drinks LESS ethanol, which inverts the standard clearance-drives-drinking model; the Ingram lab at OHSU showing that inflammatory hyperalgesia never resolves but is masked by ongoing 2-AG at CB1, driven upstream by corticosterone at glucocorticoid receptors in the periaqueductal gray, with rimonabant unmasking it three weeks out — and, critically, exogenous agonist on top of that tone desensitizing CB1 rather than adding analgesia, which explains the small clinical effects of cannabinoid analgesics and nominates GRK-directed co-therapy; the Opendak lab at Kennedy Krieger rescuing an infant affiliative approach deficit with oral chenodeoxycholic acid but not cholic acid, with the caveats that the rescue is significant only pooled and the responsible sex flips between models; anoctamin 2 knockout mice with impaired inhibitory control, reduced tyrosine hydroxylase expression and phosphorylation, lower dopamine and an unchanged transporter, putting a calcium-activated chloride channel upstream of dopamine synthesis; a two-macaque D1 challenge in which agonist and antagonist move subjective evidence weighting, priming and memory decay with large effect sizes while accuracy stays flat because the animals adjust their sampling — a caution for every pro-cognitive program reading nulls on summary accuracy; a closed-loop system from the Romero lab at Duke in which a ProteinNPT agent and a robotic laboratory ran twenty-five rounds over a month with no human intervention and shifted glycoside hydrolase specificity toward xylose and mannose, mostly by breaking the native glucose preference, with assay dynamic range shaping what the loop could learn; a persistent six-agent fleet running six months on shared memory, its six-layer provenance and trust architecture, the audit finding that every one of five autonomous research systems had systematic evidence-chain failures, and the paper's own admission that memory augmentation made its model worse on target classes the library did not cover; and a closing reanalysis showing the influential MoSeq claim that behavioral syllables beat scalar features by more than fifty percent shrinks to about eleven percent under different but equally standard preprocessing and tuning. false Daily roundup for August 19, 2026 — eleven segments. The first half asks a single question in different forms: not whether a drug works, but which part of it you are getting, and whether the good part separates from the bad. LEAD — activity-dependent tagging (Sema-TRAP) shows semaglutide recruits GLP-1 receptor neurons in the area postrema but mostly NON-receptor neurons in the nucleus tractus solitarius; reactivating the tagged NTS population reproduces satiation, conditioned flavour avoidance and suppression of progressive-ratio responding for Western diet, and area postrema receptor knockdown abolishes all three while sparing satiety and much of the weight loss (UCL, Brierley lab). Then a 115-person amisulpride-versus-placebo fMRI trial where a single 100 mg dose moves the VTA and not the ventral striatum; a preregistered crossover in which escitalopram increases willingness to wait for delayed rewards and L-dopa does not (Kable lab, Penn); EEG hyperscanning in 25 romantic couples on 50 micrograms of LSD, with theta amplitude-envelope coupling that beats cross-couple and temporal-shift surrogates; BHLHE41 as a lithium-response gene that is not a bipolar-risk gene, with a double-knockout mouse in which lithium fails to reduce excitability or dampen LTP (Molecular Psychiatry); pathway-specific excitation-inhibition imbalance confined to ventral hippocampal inputs on apical dendrites after interneuron NMDA receptor deletion; an Ionis VHH-Fc transferrin-receptor platform with a 10-100 nM affinity optimum, DAR1 beating DAR2 in brain, position-239 conjugation beating C-terminal, and MAPT knockdown across NHP cortex, hippocampus and amygdala after subcutaneous dosing; a GPCR allosteric-modulator docking benchmark from the Miao lab where GaMD ensembles beat PDB structures and Boltz-2 ignores its templates; MERIT, which measures how much published trial-outcome performance is compound memorization and then runs the failure predictor backward to recover 83% of known repositionings; The Little Scientist, a single-author agent framework that forces an LLM through hypothesis, numeric prediction and reconciliation and produced an algorithm beating STREME on 132 ENCODE transcription factors plus the current top ProteinGym zero-shot entry, for 229M tokens on one GPU-less VM; and a 149-study meta-analysis in which cannabis is associated with psychosis onset 2.5 years earlier overall, with no effect in the earliest-onset samples and 6.3 years in the latest. https://www.biorxiv.org/content/10.64898/2026.08.10.744052v1 2026-08-19-receptor-and-reason Wed, 19 Aug 2026 12:00:00 +0000 1132 Wednesday, August 19, 2026, eleven segments. The thread through the first half is specificity — not whether a drug works, but which piece of it you are actually getting. LEAD: a UCL preprint that tags the neurons semaglutide actually activates rather than the ones expressing its receptor. In the area postrema they are GLP-1 receptor neurons, as expected; in the nucleus tractus solitarius the large recruited population mostly does NOT express the receptor. Reactivating those tagged NTS neurons alone reproduces satiation, produces conditioned flavour avoidance as strong as the drug itself, and suppresses motivation to work for a high-fat high-sugar diet — and knocking down GLP-1 receptors in the area postrema before tagging abolishes all three, while semaglutide's effect on satiety and much of its weight loss survives. So the reward-suppressing effect people want for substance use disorders may be tethered to the nausea; satiety is the separable arm. Then: a 115-person amisulpride trial where a single low dose moves the VTA and nothing downstream, which I read as a design lesson rather than a dopamine result; a preregistered crossover from the Kable lab where escitalopram makes people wait longer for delayed rewards and L-dopa does nothing, with an age-dependent exploratory signal; EEG hyperscanning in 25 couples given LSD together, where inter-brain theta coupling rises and survives two surrogate controls, and felt connectedness outlasts the plasma curve; BHLHE41, a clock output gene associated with lithium response but not with bipolar risk, yielding a mouse in which lithium reaches the brain and fails to produce its cellular signature; excitation-inhibition imbalance that turns out to be confined to ventral hippocampal inputs on apical dendrites, a companion to yesterday's SETD1A result; an Ionis platform delivering siRNA and antisense across the blood-brain barrier with a binding-affinity sweet spot rather than a maximum, and tau knockdown across primate cortex, hippocampus and amygdala from a subcutaneous injection; a GPCR allosteric docking benchmark with a usable rule for re-ranking ensembles, and a deep-learning model that ignores the conformations you hand it; MERIT, which shows most published trial-outcome prediction is compound memorization and then runs its failure predictor backward to recover known drug repositionings; The Little Scientist, one researcher's agent that replaces evolutionary search with enforced hypothesis-and-reconciliation and beat the MEME Suite default and the ProteinGym leaderboard on a single virtual machine with no GPUs; and a meta-analysis where cannabis moves psychosis onset earlier by 2.5 years on average, with no effect in the youngest-onset samples and six years in the oldest. false Daily roundup for August 18, 2026 — eleven segments, bioRxiv back online after yesterday's outage, with a spine running through receptors and drugs that misbehave relative to the simple model. LEAD — JAK inhibition overcomes first-line drug resistance in epilepsy (neuroinflammation / repurposing; bioRxiv, University of Wisconsin at Madison, full text via r.jina.ai): chronically epileptic mice (repeated low-dose systemic kainic acid, multifocal TLE) were screened against carbamazepine using the clinical 50%-reduction benchmark — 38% responded, 62% did not — and tofacitinib (CP690550, pan-JAK with JAK1 preference, approved for RA) was then added on top in the resistant cohort. Dual therapy cut median seizure frequency 14-fold and time spent seizing 18-fold; 70% of resistant animals responded, and responders showed a 100-fold drop with 5 of 7 free of observed behavioral seizures. PK is anchored rather than assumed (carbamazepine brain concentration tracks seizure threshold, R2 = 0.71; tofacitinib brain-to-plasma 0.12, ~310 nM brain, exceeding 89% of reported JAK1 heterodimer IC50s, with non-BBB-penetrant trospium undetectable as control). Crucially, working and short-term memory were restored to naive levels including in seizure non-responders, with no correlation between seizure reduction and memory rescue — cognition and seizure control are separable targets. pY705-STAT3 was suppressed equally in responders and non-responders, so non-response is not a target-engagement failure. Mouse preprint, small groups, video- not EEG-scored seizures, and JAK-inhibitor immunosuppression is non-trivial in pediatrics. LXR activation and remyelination (MS drug discovery; bioRxiv, Johns Hopkins, full text): IFN-gamma flips OPC lipid metabolism from synthesis to utilization (fatty acid depletion, increased FAO) and induces immune-like, antigen-presenting OPCs that will not differentiate; the same signature appears in human surgical specimens, mouse demyelination models, and human MS lesions. Oleic acid supplementation or FAO inhibition restores differentiation; the LXR agonist GW3965 rebalances lipid handling and overcomes the IFN-gamma differentiation block in mouse and human-derived OPCs; in an adoptive-transfer (Th17 2D2) plus cuprizone model with dosing delayed 3 days to avoid peripheral immunosuppression, GW3965 (10 mg/kg) left the CNS infiltrate unchanged yet increased Olig2+ASPA+ mature oligodendrocytes and fluoromyelin — the authors state no prior approach has overcome cytokine-mediated differentiation blockade. Human genetic support from an NR1H3 loss-of-function variant in rapidly progressive MS; the translational question is CNS engagement without hepatic lipogenesis. Radiprodil (Br J Pharmacol): whole-cell and outside-out patch clamp in neonatal rat substantia nigra dopaminergic neurons shows this clinical-stage GluN2B-selective NAM produces only ~40% maximal inhibition at 30-100 nM and then loses inhibition at higher concentrations — a bell-shaped curve fit by a two-binding-site 'hypercube' model plus a 12% single-channel current reduction. Dose escalation could cost efficacy, and the curve shape is subunit-composition dependent. Glutamatergic antipsychotics (Br J Pharmacol): chronic MK-801 in rats reduces sucrose preference, GluN2A/GluN2B expression and NMDA-evoked release while raising basal extracellular glutamate and D-serine in orbitofrontal cortex; clozapine or D-cycloserine alone restore sucrose preference and evoked release, but combined they worsen anhedonia and overshoot on evoked release, receptor expression and basal levels — all reversed by memantine, implicating extrasynaptic NMDA overactivation. A tachyphylactic 25 mg/kg D-cycloserine dose down-regulates GluN2A/2B (blocked by the glycine-site inhibitor MDL29951), and clozapine suppresses PP2A. The goal is the synaptic-to-extrasynaptic ratio, not maximal NMDA tone. AUT00206 (J Psychopharmacol): the selective Kv3.1/3.2 positive modulator, acting on the channels that let parvalbumin interneurons fast-spike, attenuated sub-chronic PCP deficits in novel object recognition, reversal learning and social interaction at 10 and 30 mg/kg in female Lister Hooded rats — a clinical-stage compound, but acute dosing, one model, one sex. SETD1A and thalamo-prefrontal synchrony (Neuropsychopharmacology, Gordon lab at NIMH, full text): Setd1a+/- mice learn a delayed non-match-to-sample task normally and fail only under blocked constant delays; hippocampal-prefrontal theta synchrony is preserved (unlike 22q11.2 models), while nucleus reuniens-mPFC delay beta synchrony is reduced and the bidirectional beta/gamma modulation that predicts correct trials in WT is disrupted — different schizophrenia risk genes break different circuits. Psychedelic persisting perceptual abnormalities (bioRxiv, Yale; abstract, body not yet rendered): 186 naturalistic users plus a visual conditioned hallucinations task and a computational model; PPAs track younger first use, higher doses, lower visual thresholds, higher conditioned-hallucination rate/confidence and worse discrimination, but among model parameters only reduced decision precision tracked both past and current PPAs and mediated the dose-PPA relationship — a noisy detection-biased visual system, not simply overweighted priors, which cuts against the naive REBUS reading. SV2C ligand discovery (computational pharmacology; bioRxiv, SandboxAQ; abstract, body not yet rendered): SV2A cryo-EM homology model plus MD/GaMD, a CNN scoring function validated at r = 0.72 on a 39-ligand SV2A benchmark, 5.96M Mcule compounds funneled to 94 candidates, 71 profiled by thermal shift plus [3H]-padsevonil SPA, 22 actives (31% hit rate); best lead Ki 3.25 uM at SV2C with >10-fold selectivity over SV2A and ~12-fold over SV2B, common tryptophan-cage binding mode confirmed to 0.76 A binding-site Ca RMSD against an unpublished SV2A-plosaracetam cryo-EM structure. Micromolar binders, not yet functionally characterized, but the first selective probes for a Parkinson's-linked dopaminergic target. PertMind (agentic/LLM in computational biology; arXiv, Zhejiang University and Tencent, PDF read in full): reframes the Tahoe-100M atlas (~100M single-cell profiles) as an RL environment where measured up/down/no-change gene responses are computable rewards, replacing hand-curated reasoning traces; Qwen3-4B Base plus trusted-trajectory SFT and gene-, pathway- and format-level rewards, trained only on forward perturbation-response prediction, transfers zero-shot to reverse perturbation identification (approaching a task-specialized specialist and flatter across hard perturbations), unseen double perturbations, AssayBench phenotypic-screen prioritization (improving GPT-5.4, Gemini 3 Pro and Qwen3.5-397B-A17B) and biological-process naming, with only small MMLU/CMMLU degradation. Limitation the authors state: the reward sees only the terminal endpoint, so nothing verifies the mechanistic narrative in between. Nociceptor CB1 (bioRxiv, UT Dallas; abstract, body not yet rendered): new CB1R floxed-stop-floxed mice crossed to Nav1.8-cre restore CB1 only in peripheral nociceptors; systemic WIN55,212-2 gives robust tail-flick analgesia in WT but none in global knockouts or nociceptor-rescue animals, so nociceptor CB1 is not sufficient for acute analgesia — a real negative for peripherally restricted cannabinoids — while global knockouts develop mechanical and thermal hypersensitivity earlier after nerve injury, so nociceptor CB1 does restrain neuropathic pain. CLOSE — methadone and synthetic cannabinoid receptor agonists (Br J Pharmacol): in National Programme on Substance Use Mortality data, methadone was the most co-detected long-QT-liability medication in SCRA deaths, and median fatal methadone blood concentration was significantly lower when an SCRA was present; in isolated perfused guinea pig hearts methadone alone prolonged QTc, 5F-ADB alone did not, and co-application prolonged it further; in silico and patch clamp show 5F-ADB is a low-affinity hERG inhibitor — a stacking pro-arrhythmic interaction nobody screens for in opioid agonist therapy. Eleven segments per receptor-and-reason PROMPT.md. Source notes: bioRxiv details API pulled day-by-day 2026-08-11 to 2026-08-18 with full pagination (1,427 preprints); fetch_preprint.py fell through to the abstract-only channel on every candidate, so full text was pulled by curling r.jina.ai directly against the .full pages — that recovered complete bodies for the JAK/epilepsy lead and the LXR remyelination paper but confirmed the SV2C, nociceptor-CB1, psychedelic-PPA and LLM-perturbation-atlas preprints are not yet rendered (deferred to abstract-scale tail mentions; the bioRxiv PDF channel 429s through the proxy). Nature NPP and Molecular Psychiatry listings and open-access bodies via r.jina.ai; Elsevier/Wiley/Oxford journals via PubMed; arXiv from the shared daily cache with the PertMind PDF extracted via mutool. Deferred for lack of rendered full text and worth revisiting: the SV2C, nociceptor-CB1 and psychedelic-PPA preprints. Paper links in show notes. https://www.biorxiv.org/content/10.64898/2026.08.06.743311v1 2026-08-18-receptor-and-reason Tue, 18 Aug 2026 12:00:00 +0000 1076 Tuesday, August 18, 2026, eleven segments, with bioRxiv back after yesterday's outage. LEAD — in mice that had already failed carbamazepine, adding tofacitinib, the JAK inhibitor approved for rheumatoid arthritis, cut median seizure frequency fourteen-fold, and among responders a hundred-fold, with most animals free of observed seizures; it also restored working and short-term memory, and it did that even in the animals whose seizures did not respond, decoupling cognitive rescue from seizure control. Liver X receptor activation rebalances oligodendrocyte lipid metabolism and lets remyelination proceed despite inflammation, which the authors say no prior approach has managed. Radiprodil's GluN2B block turns out to be bell-shaped, so more drug means less inhibition. Clozapine plus D-cycloserine worsens negative symptoms in rats because it tips signaling from synaptic to extrasynaptic NMDA receptors, and memantine reverses it. A Kv3.1 and 3.2 modulator, AUT00206, rescues cognition and social behavior in the sub-chronic PCP rat by acting on parvalbumin interneurons. SETD1A haploinsufficiency leaves hippocampal-prefrontal synchrony intact and breaks the thalamic reuniens to prefrontal beta circuit instead, so different schizophrenia risk genes break different circuits. Persisting perceptual abnormalities in psychedelic users track reduced decision precision rather than stronger priors. An AI-enhanced virtual screen of nearly six million compounds delivers the first selective SV2C binders, a Parkinson's-linked dopaminergic target. PertMind turns a hundred-million-cell perturbation atlas into a reinforcement-learning environment and gets biological reasoning that transfers to tasks it never trained on. Nociceptor CB1 is not sufficient for acute analgesia but does restrain neuropathic pain. And a synthetic cannabinoid lowers the fatal methadone threshold through hERG. Paper links in show notes. false Daily roundup for August 17, 2026 — ten segments, a journals day after bioRxiv's details API returned empty on every query (a source outage; Tier 1 preprints unavailable, logged), drawn from Molecular Psychiatry, Lancet Neurology, Biological Psychiatry CNNI, Neuropharmacology, the International Journal of Neuropsychopharmacology, Basic and Clinical Pharmacology and Toxicology, the Journal of Psychopharmacology, Psychiatric Times, and arXiv. LEAD — Adolescent alcohol and astrocyte-synaptic coupling (glia/AUD mechanism; Molecular Psychiatry, full text via nature.com): adolescent intermittent ethanol in male rats (binge-range BECs) drives peripheral astrocytic processes to retract from dorsal-hippocampal synapses and stay retracted into adulthood; astrocytes go hypo-responsive (reduced calcium response despite more glutamate) and animals over-respond in fear conditioning; a Gq DREADD switching dHipp astrocytes back on partially normalizes the fear phenotype and fully restores the adenosine-freezing relationship — astrocyte activity causally sufficient; male-only caveat. NOP receptors and alcohol, longitudinal primate PET (addiction pharmacology; Basic and Clinical Pharmacology and Toxicology): [11C]NOP-1A PET in 11 cynomolgus monkeys, ethanol-naive then after 4-month induction plus 6 months heavy drinking; no baseline sex difference, but baseline NOP binding predicted future intake (orbitofrontal cortex in males, caudate in females), and chronic drinking remodeled NOP availability across reward regions (up in putamen both sexes; opposite-direction moves in insula/amygdala by sex) — a bidirectional, sex-dependent target-engagement biomarker for NOP-directed AUD drugs. Guanabenz in vanishing white matter (integrated-stress-response repurposing / clinical; Lancet Neurology, worked from the published trial report — Lancet body paywalled, structured report only, presented as reported): VWM is a fatal childhood leukodystrophy caused by eIF2B mutations that dysregulate the ISR; guanabenz, an alpha-2 antihypertensive that inhibits the ISR and rescued a mouse model, was given to 33 still-ambulant children in a single-arm phase 1/2 trial titrated over years; vs 66 matched historical controls, HR ~0.33 (95% CI 0.16-0.69, log-rank p=0.0061) for loss of walking-with-support; main safety signal transient hallucinations in ~55% (first 4 months, mostly resolved), plus constipation/hypotension, no deaths; non-randomized historical-control comparison, extension study needed — but a landmark proof of concept for drugging the ISR in human neurodegeneration (relevant to ALS, tauopathies). White-matter microstructure in autism, DTI meta-analysis (neuroimaging synthesis; Biological Psychiatry CNNI): multilevel random-effects meta of 881 effect sizes across 66 studies finds robustly lower fractional anisotropy across nine major tracts (corpus callosum, arcuate/SLF, uncinate, cingulum, IFOF/ILF, thalamic radiation, corticospinal) with moderate effect sizes (g -0.30 to -0.50); consolidates a widespread, reproducible reduced-integrity story and maps age/sex/IQ moderators. Neurabin-1 (PPP1R9A) haploinsufficiency in human iPSC cortical neurons (neurodevelopmental genetics; Molecular Psychiatry, full text): CRISPR heterozygous knockout neurons are morphologically over-built (hyperspinogenesis, increased neurite complexity) yet electrically underpowered (reduced firing, blunted action potentials, defective axo-somatic coupling), with coordinated downregulation of sodium-channel and glutamatergic machinery — a structure-function dissociation that gives clinical geneticists mechanistic evidence to start reclassifying these variants of uncertain significance. Tail of the VTA and maternal behavior (circuit pharmacology; Neuropharmacology): intra-tVTA/RMTg microinjection of glutamate, muscimol, DAMGO, oxytocin, or atosiban in postpartum rats shows this GABAergic brake on dopamine neurons bidirectionally gates maternal behavior (activation suppresses, inhibition enhances), with opioid and oxytocin sensitivity placing two social-bonding peptide systems at a specific anatomical node. Parvalbumin interneurons under stress (mechanistic review; International Journal of Neuropsychopharmacology): frames PV interneurons and their perineuronal nets as the convergence point for redox, metabolic, inflammatory/microglial, and epigenetic stress vulnerabilities, and maps candidate protective interventions (antioxidants, anti-inflammatories, ketamine and derivatives) onto specific mechanisms, with sex differences. Viloxazine ER (Qelbree) real-world Phase 4 in comorbid ADHD (clinical pharmacology; Psychiatric Times): open-label, 104 completers with ADHD plus clinically significant depression/anxiety, nicotine/cannabis permitted; ADHD symptoms down ~45-51% and MADRS/HAM-A down ~50% by week 14; viloxazine is an NRI with 5-HT2C partial agonism plus 5-HT2B and 5-HT7 antagonism (its IR form was a European antidepressant) — but open-label, no comparator, 14 weeks, mostly white/female, boxed suicidal-ideation warning (3 discontinued), and mood benefit can't be separated from ADHD relief. Localize, Then Reason (computational chemistry / vision-language; arXiv): a molecular-property VLM that first localizes the chemically meaningful substructure in a molecular image, then reasons in a compact latent workspace before answering — ~9.6x throughput vs a text-reasoning baseline at matched accuracy; incremental but a clean look-before-you-think result. CLOSE — Defining 'psychedelic' (phenomenology/taxonomy; Journal of Psychopharmacology, Carhart-Harris group): over 200 people rated subjective effects of a classic serotonergic psychedelic, an MDMA-family entactogen, and a ketamine-family dissociative; factor analysis yielded three-to-four independent experiential dimensions and an ML classifier separated the classes from the subjective profile alone — arguing 'psychedelic' is a definable category anchored in a distinguishable subjective state, not just a pharmacological label. Ten segments per receptor-and-reason PROMPT.md. Source notes: journal content read via r.jina.ai proxy of open nature.com pages (adolescent-alcohol lead and Neurabin-1 in full) and via PubMed structured records (NOP/alcohol, guanabenz, white-matter meta-analysis, PV-interneuron review); guanabenz Lancet Neurology body was paywalled so the item is worked from the published structured report and hedged accordingly; Psychiatric Times viloxazine read via r.jina.ai; arXiv from the shared daily cache. Dedup: amisulpride social-reward fMRI, the BMI clustered-MR paper, and the static-analysis-guided agentic Rust-bioinformatics paper were all excluded as already shipped 2026-08-15. bioRxiv Tier 1 unavailable today (source outage, one retry, logged). Paper links in show notes. https://www.nature.com/articles/s41380-026-03817-8 2026-08-17-receptor-and-reason Mon, 17 Aug 2026 12:00:00 +0000 661 Monday, August 17, 2026, ten segments, a journals day after bioRxiv's API went dark (source outage, logged). LEAD — adolescent binge alcohol makes dorsal-hippocampal astrocytes retract their fine processes from synapses and stay retracted into adulthood, leaving the astrocytes hypo-responsive despite extra glutamate and the animals over-reacting in fear conditioning; chemogenetically switching those astrocytes back on rescues the fear phenotype, so astrocyte activity is causally sufficient and a candidate drug target (males only). Longitudinal PET in monkeys shows NOP opioid-receptor tone both predicts future drinking and is reshaped by chronic drinking, in a sex-dependent way. Guanabenz, an old alpha-2 antihypertensive that inhibits the integrated stress response, cut the risk of losing the ability to walk by about two-thirds in children with vanishing white matter, a fatal leukodystrophy caused by eIF2B mutations, versus matched historical controls, with transient hallucinations the main early side effect, a landmark for drugging the integrated stress response though not a randomized comparison. A 66-study diffusion-imaging meta-analysis pins widespread lower white-matter integrity across nine tracts in autism. Neurabin-1 haploinsufficiency builds over-connected but under-firing human cortical neurons, a structure-function split that helps reclassify an uncertain variant. The tail of the VTA gates maternal behavior through opioid and oxytocin inputs. A review frames parvalbumin interneurons as the convergence point of stress vulnerability. Real-world Phase 4 viloxazine ER improves ADHD with comorbid depression and anxiety, though open-label and hard to disentangle from ADHD relief. A molecular-property vision-language model that localizes before it reasons runs about ten times faster. And the Carhart-Harris group argues, from subjective-effect ratings of three drug classes, that 'psychedelic' is a definable category, not just a label. bioRxiv Tier 1 was unavailable today. Paper links in show notes. false Daily roundup for August 16, 2026 — ten segments; a full pharmacology day after yesterday's bioRxiv outage, running heavy on mechanism with a mid-episode arc through psychedelics and biased-receptor signaling, and a recurring sex-differences theme. LEAD — Reversible m6Am methylation of snRNA by FTO controls morphine reward and tolerance without altering analgesia (opioid epitranscriptomics; bioRxiv, full text via r.jina.ai): a Cornell-led group shows morphine reward (conditioned place preference) and antinociceptive tolerance to morphine and fentanyl each require the RNA demethylase FTO, while analgesia is preserved — the long-sought dissociation. Mechanism surprise: mRNA m6A unchanged across 20,139 sites; instead FTO loss raises m6Am (N6,2'-O-dimethyladenosine) on snRNA, and the benefit vanishes in mice that cannot make m6Am. Anatomy split: FTO knockdown in nucleus accumbens shell cuts reward not tolerance; in dorsal root ganglion cuts tolerance not reward. Pitch: FTO inhibitors as opioid adjuvants; caveat: FTO is pleiotropic. DMT vs 5-MeO-DMT preclinical comparison (psychedelic pharmacology; bioRxiv): DMT gives a bell-shaped head-twitch dose-response, 5-MeO-DMT monotonic; 5-HT2A/5-HT1D antagonism or 5-HT1A agonism flattens head-twitch without abolishing marble-burying effect (hallucination dissociable from a therapeutic-like behavior); both raise TrkB phosphorylation and synaptic proteins at 12 days, DMT engages more broadly (DMN, hippocampus) yet uniquely lowers hippocampal BDNF and reprograms frontal glutathione/energy metabolism — rational-design fodder for a scalable, less-hallucinogenic psychedelic. MDMA-enhanced exposure therapy in a learned-helplessness PTSD model (bioRxiv): MDMA plus brief trauma-cue exposure produced rapid, sustained recovery on active avoidance beyond either alone, with dose-dependent anxiety/depression effects — proof-of-principle that the drug can compress therapist-intensive protocols; rodent, temper the title. Oxytocin-receptor Gi signaling and social-fear extinction (bioRxiv): a caudal-lateral-septum oxytocin-receptor neuronal population must be transiently silenced at the moment of social contact for extinction; chemogenetic mistiming impairs it; downstream, Gi-coupled (not Gq-coupled) signaling accelerates extinction and promotes social approach — a biased-agonism target for treatment-resistant social anxiety. Beta-arrestin-2, sex, and Alzheimer's (bioRxiv, APP/PS1): Arrb2 deletion in males reduces amyloid-beta oligomers, boosts autophagy, quiets glia, and broadly rescues cognition (lower Akt/GSK-3-beta phosphorylation, GSK-3-beta/ZBTB16 autophagy); in females amyloid/astrogliosis unchanged, microglia more reactive, only recognition memory improved — sex-stratified design warning for GPCR-targeted AD therapeutics. Polyamine stress response in schizophrenia (framework/review; Progress in Neuro-Psychopharmacology & Biological Psychiatry): proposes a stress-driven polyamine-pool reset via the arginine-nitric-oxide-agmatine-polyamine axis, with agmatinase as a regulatory node and irreversible upstream inhibitors as probes — a non-monoamine reframing with concrete druggable set points; hypothesis, not data. Kynurenine metabolites and psychotropics in bipolar disorder (FACE-BD, n=173; PNPBP): valproate users had lower tryptophan/kynurenine/xanthurenic/quinolinic acid, lithium users higher kynurenine and KYN/TRP ratio, lamotrigine users higher quinolinic acid, surviving covariate adjustment; antipsychotics/antidepressants no signal — current medication is a confounder kynurenine biomarker studies must model. Dimethyl fumarate in the prenatal-valproate autism model (Neuropharmacology): the model's cortical phenotype is itself sex-specific (male neuropeptide-signaling genes Pdyn/Adora2a/Drd2 up; female vascular/inflammation genes down); DMF, an Nrf2 activator approved for MS, rescued recognition/anxiety/social behavior, corrected oscillatory power and cross-frequency coupling, and suppressed microglia in females only — a repurposing lead and a third sex-stratified result. LungChat, a conversational multi-agent multi-omics system (agentic AI in computational biology; bioRxiv, abstract-only, brief mention): hierarchical supervisor decomposes NL questions into parallel tool-grounded tasks; a direction-aware repurposing-and-targeting module distinguishes disease-reversing from disease-reinforcing compounds per cell type; independently re-derived saracatinib (Src-family kinase inhibitor) for IPF matching a real trial and recovered a COPD steroid; ablations show hierarchical orchestration improves grounded abstention and token efficiency — lung, not brain, but the verifiable, abstaining architecture is the transferable idea. CLOSE — FDA approval of florquinitau F 18 (Tauklarify / MK-6240) tau-PET (regulatory; Lantheus): approved for imaging tau neurofibrillary-tangle pathology in adults being evaluated for Alzheimer's, to complement amyloid PET, not validated for non-AD tauopathies; leaned on the Lancet HEAD study plus blinded-read studies — really trial infrastructure to enrich/stratify anti-tau drug trials. Ten segments per receptor-and-reason PROMPT.md. Source notes: bioRxiv full corpus pulled day-by-day 08-09 to 08-16 (details API, ~1,344 preprints, full pagination); FTO lead read in full via r.jina.ai; three Elsevier journal abstracts (polyamine, kynurenine, dimethyl fumarate) via PubMed as ScienceDirect/Wiley were CAPTCHA-walled; Nature/NPP/MP listing pages auth-walled, read via r.jina.ai; arXiv from the shared daily cache (bio-relevant agentic hits were already-shipped or off-topic); Tauklarify facts via press/news search; LungChat full text not yet rendered (abstract-only, brief tail mention). Paper links in show notes. https://www.biorxiv.org/content/10.64898/2026.08.06.743062v1 2026-08-16-receptor-and-reason Sun, 16 Aug 2026 12:00:00 +0000 713 Daily roundup for August 16, 2026, ten segments, a full pharmacology day after yesterday's bioRxiv outage, heavy on mechanism with a mid-episode run through psychedelics and biased-receptor signaling and a recurring sex-differences theme. LEAD — morphine reward and tolerance, but not analgesia, require the RNA demethylase FTO; the mechanism is not messenger-RNA m6A but a related mark, m6Am, on small nuclear RNA, and the effect maps to the nucleus accumbens shell for reward and the dorsal root ganglion for tolerance, making FTO inhibitors a candidate opioid adjuvant. DMT and 5-MeO-DMT are pried apart so that 5-HT2A-driven head-twitch can be blocked while a therapeutic-like behavior and lasting TrkB-linked plasticity persist. MDMA plus a brief trauma-cue exposure compresses PTSD-like recovery in a learned-helplessness model. In the lateral septum, oxytocin-receptor neurons must be silenced at the moment of social contact for fear extinction, and it is Gi-coupled, not Gq-coupled, signaling that accelerates it, a biased-agonism target for social anxiety. Beta-arrestin-2 deletion rescues amyloid and cognition in male but not female APP/PS1 mice, a sex-stratified warning for GPCR-targeted Alzheimer's drugs. A polyamine-stress-response framework offers a non-monoamine target theory for schizophrenia. In bipolar disorder, valproate, lithium, and lamotrigine each leave distinct kynurenine-pathway fingerprints, a confounder for biomarker studies. Dimethyl fumarate, an Nrf2 activator approved for MS, rescues the prenatal-valproate autism model in females only. LungChat, a verifiable multi-agent system, does direction-aware drug repurposing and re-derived saracatinib for pulmonary fibrosis. And the FDA approved florquinitau, brand name Tauklarify, the MK-6240 tau-PET tracer, for Alzheimer's evaluation, useful trial infrastructure for anti-tau drug development. Paper links in show notes. false Daily roundup for August 15, 2026 — nine segments, a journals day after the bioRxiv details API returned empty on every query (historical dates included), so Tier 1 preprints were unavailable; drawn from Neuropsychopharmacology, Molecular Psychiatry, and Biological Psychiatry, with one arXiv computational piece to close. LEAD — Amisulpride and social-reward anticipation in depression (dopamine pharmacology / drug-challenge fMRI; Neuropsychopharmacology, full text via nature.com): a single 100 mg dose of amisulpride, a D2/D3 antagonist that at low doses preferentially blocks presynaptic autoreceptors and thus raises dopamine, tested double-blind/placebo in 115 people (58 mild MDD, 57 HC) on a Social Incentive Delay task using smiling faces (social, not monetary, reward). Surviving correction: amisulpride increased LEFT VTA activation during reward anticipation across both groups; it did NOT move ventral striatum and had no behavioral effect — a midbrain-source vs striatal-target dissociation contrasting with monetary/pramipexole work. Wrinkle: in MDD-amisulpride, higher anhedonia tracked with HIGHER VTA/putamen/insula activation (compensatory/inefficient recruitment; wanting vs liking), the opposite of the naive hypoactivation-restored story. Mildly depressed sample attenuates effects; a mechanistic probe, not a therapeutic signal. Clustered Mendelian randomization of BMI on depression symptoms (causal statistical genetics; Molecular Psychiatry): PheWAS-based clustering of BMI genetic instruments across 407 UK Biobank traits, then cluster-wise MR onto individual PHQ symptoms; appetite change is the ONLY symptom with a robust, homogeneous causal effect across all clusters (~0.19-0.34), bidirectional with BMI; depressed mood/psychomotor/suicidality minimal; fatigue/anhedonia/sleep heterogeneous (pleiotropy) — so BMI-to-depression runs largely through an appetite-metabolic arm (a GLP-1-shaped target), and socioeconomic-enriched clusters inflate broader estimates. Caveat: the appetite item conflates increased and decreased eating; European ancestry. Inflammaging and late-life depression (perspective/framework; Neuropsychopharmacology): reframes LLD as a brain-body systems disorder in which inflammaging plus immunometabolic crosstalk (insulin resistance, mitochondrial dysfunction, oxidative stress) moderate illness trajectory and degrade neurotransmission/bioenergetics; precision-psychiatry turn is to personalize on immunometabolic state rather than monoamines; agenda-setting, light on specific targets. Prenatal maternal metabolomics and child neurodevelopment (metabolomics/neurodevelopment; Molecular Psychiatry, NYU CHES): targeted panel of 188 metabolites in maternal urine across three pregnancy timepoints in 1,070 mother-child pairs, child behavior/autism traits ~age 2.3; robust FDR-surviving signal is a late-pregnancy lysophosphatidylcholine associated with LOWER autism/behavior scores in boys only; softer signals — early kynurenine/tryptophan ratio up with autism traits, taurine up with behavior problems in girls (diet-adjusted). Implicates lipid transport, mitochondrial energy metabolism, immune-inflammatory pathways; strongly sex-specific; associational, single cohort, toddler outcomes. Neonatal microRNA signature for ADHD (developmental biomarker; Biological Psychiatry, MoBa cohort): a 19-miRNA cord-blood signature predicts later ADHD diagnosis and tracks symptom severity, enriched for dopaminergic-signaling and circadian pathways; first demonstration that neonatal miRNAs carry ADHD-relevant signal, but the quoted AUC ~0.96 should be read as an optimistic ceiling pending external validation, not a clinical claim. Early-life stress and episodic-memory development (developmental/computational; Biological Psychiatry, ABCD, ~3,466 youth 9-10): latent change score plus connectome-based predictive modeling; stress-related psychopathology statistically MEDIATES the link from early-life stress to less two-year improvement in episodic memory; visual-cortex connectivity predicts less improvement, cerebellar/subcortical predict more; observational, mediation is a model. Theta-gamma coupling as a treatment moderator (biomarker-guided neuromodulation; Neuropsychopharmacology, PACt-MD secondary analysis): 260 older adults (~72y) with MCI or remitted MDD, cognitive remediation + tDCS vs sham, up to 6-year follow-up; prefrontal theta-gamma coupling on an N-back (median split) moderates outcome — high-coupling patients show slowed long-term global cognitive decline under active treatment (p=0.006), low-coupling get only transient executive gains; TGC read as a cognitive-reserve proxy stratifying who benefits; secondary analysis, median split. Selection bias in precision psychiatric neuroimaging (methods commentary / the reason-side caution; Biological Psychiatry): dense, multi-session precision-imaging designs select for scanner-tolerant, low-attrition participants who differ systematically (e.g., anxiety), and conditioning on participation invites collider/Berkson bias, so individualized connectome predictions do not transport to the clinic; remedy is bias-aware causal design, not bigger scanners — a skepticism to carry into today's other prediction claims. CLOSE — Static analysis-guided agentic translation of bioinformatics code to Rust (agentic AI in computational biology; arXiv, full text): pairs static analysis with LLM coding agents (Claude Code early, then Codex) for conservative, auditable, function-by-function translation of legacy Python/C to memory-safe Rust; on a real single-cell tool, memory dropped ~80x, build time ~10x, key steps over 3x faster, Unix dependencies removed; took ~11 weeks of human-guided work across frontier models, explicitly not a model comparison — a proof that agentic translation can retire the performance tax on scientific Python without a from-scratch rewrite. Nine segments per receptor-and-reason PROMPT.md. Source notes: journal content read via r.jina.ai proxy of open nature.com and Biological Psychiatry article pages (amisulpride lead, BMI-cluster MR, prenatal metabolomics, theta-gamma, and the two Biological Psychiatry items read in full); arXiv from the shared daily cache; bioRxiv Tier 1 UNAVAILABLE today (details API and MCP both returned empty 200s from this host — a source outage, one retry, logged); arxiv:2608.11483 (agentic hit-to-lead) excluded as already shipped 2026-08-14. Paper links in show notes. https://www.nature.com/articles/s41386-026-02521-z 2026-08-15-receptor-and-reason Sat, 15 Aug 2026 12:00:00 +0000 684 Daily roundup for August 15, 2026, nine segments, a journals day after the bioRxiv details API went dark, from Neuropsychopharmacology, Molecular Psychiatry, and Biological Psychiatry with one computational piece to close. LEAD — a single low dose of amisulpride, a D2/D3 antagonist that at low doses raises dopamine by blocking presynaptic autoreceptors, increased left VTA activation during anticipation of social reward (smiling faces) in 115 people, but left the ventral striatum untouched and changed no behavior, a midbrain-source versus striatal-target dissociation; and in depressed patients, more anhedonia tracked with more recruitment, not less, a wanting-versus-liking wrinkle. Clustered Mendelian randomization pins appetite change as the one depressive symptom higher BMI causally drives, a GLP-1-shaped, symptom-specific target, with socioeconomic gene clusters inflating the broader estimates. A perspective reframes late-life depression as an inflammaging, immunometabolic trajectory to personalize on inflammation, mitochondria, and insulin rather than monoamines. Prenatal maternal metabolomics ties a late-pregnancy lysophosphatidylcholine to lower autism traits in boys, with sex-specific kynurenine and taurine signals. A nineteen-microRNA cord-blood signature predicts later ADHD, though its area under the curve near 0.96 should be read as an optimistic ceiling. Stress-related psychopathology mediates early-life stress into slower episodic-memory development in ABCD youth. Prefrontal theta-gamma coupling stratifies who gets durable benefit from cognitive remediation plus tDCS in dementia-risk elders. A commentary warns that selection bias makes precision-neuroimaging connectome predictions non-transportable. And static-analysis-guided agentic AI translates legacy bioinformatics code to Rust, roughly eighty times less memory and over three times faster on a real single-cell tool. bioRxiv Tier 1 was unavailable today (source outage, logged). Paper links in show notes. false Daily roundup for August 14, 2026 — eleven segments, a journal-heavy day out of Neuropsychopharmacology, Molecular Psychiatry, Neuropharmacology, and Progress in Neuro-Psychopharmacology, plus two arXiv computational pieces to close. LEAD — Paternal epigenetic inheritance of methamphetamine taking (addiction epigenetics; Neuropsychopharmacology, full text via nature.com): in rats, male offspring of methamphetamine-self-administering sires show increased meth susceptibility — the directional mirror of the previously reported cocaine-resistance in cocaine-sired sons; the effect is male-only, methamphetamine-specific (not sucrose, cocaine, or morphine), and abolished by delaying mating past a sperm washout. Single-nucleus RNA-seq and ATAC-seq of the nucleus accumbens localize chromatin changes to a D1 medium-spiny-neuron population in the Islands of Calleja, but the transcription-factor logic diverges by drug: cocaine tracks neuronal programs in canonical MSNs whereas methamphetamine implicates astrocytes and oligodendrocytes, with a fibroblast-growth-factor gene (FGF14) downregulated across cell types — i.e., drug-specific and cell-type-specific inheritance. Caveats: multiomics males-only, candidate genes correlational, low transcriptome-epigenome overlap. The neuropsychopharmacology of treatment discontinuation (conceptual perspective; Neuropsychopharmacology, open access): argues discontinuation is an active neuroadaptive state, not pharmacological absence, and lays out four layers — sigmoidal-occupancy pharmacodynamic rebound (the case for hyperbolic tapering), receptor adaptation (D2 upregulation, supersensitivity psychosis misread as relapse), allostatic counterregulation, and gene-regulatory carryover (candidate mechanism for discontinuation-induced refractoriness, best documented for lithium); antidepressant discontinuation-symptom incidence ~15-43% across meta-analyses; agenda-setting, and the withdrawal-vs-recurrence distinction is expert-proposed, not validated. Low-dose lithium, neuroprotection, and brain aging in bipolar disorder (clinical pharmacology; Neuropsychopharmacology hot-topics commentary): the case that serum lithium ~0.2-0.6 mmol/L retains mood-stabilizing efficacy with less toxicity, especially in older adults, argued alongside recent findings of amyloid-sequestered brain-lithium depletion in MCI/AD and an amyloid-resistant formulation rescuing mouse pathology; hypothesis-generating, a low-dose lithium MCI pilot missed efficacy, and the authors warn against lithium orotate for dementia prevention. Seven-tesla MR spectroscopic imaging in early-phase schizophrenia (imaging pharmacodynamics; Neuropsychopharmacology): ultra-high-field spectroscopy of thalamus and caudate during acute psychosis and after eight weeks of antipsychotic finds lateralized, left-right-asymmetric glutamate, GABA, and neuronal-marker abnormalities read as asymmetric astroglial-neuronal imbalance, with non-remitters showing lower baseline left-sided neuronal and GABA markers; small (n=29), response analyses exploratory, laterality needs replication. Chondrolectin-somatostatin interneurons in schizophrenia (cell-type pathology; Molecular Psychiatry): the molecular deficits in this genetically defined SST subtype are present in prefrontal cortex but absent in caudate (Bayesian evidence for no striatal difference), with a diagnosis-by-region interaction — vulnerability is cell-intrinsic features times circuit context; postmortem, 18 pairs, small-to-medium cortical effect sizes. Glial-neuronal crosstalk via GABA (mechanism review; Molecular Psychiatry): consolidates GABA as a bidirectional glia-neuron messenger — glia sense, synthesize, take up, and release it — and flags that GABA turnover is largely mitochondrial, so glial GABA signaling is a bioenergetic control point whose disruption drives mitochondrial dysfunction and glial state transitions; narrative synthesis, no new data. Hippocampal teneurin-4 and oligodendrocytes (depression mechanism; Neuropharmacology): chronic stress lowers hippocampal teneurin-4; knocking it down produces depression-like behavior plus reduced myelin basic protein via an oligodendrocyte differentiation arrest (fewer mature OLs, more precursors, same total lineage), and clemastine, an antihistamine repurposed as a remyelinating agent, rescues both; rodent model, region-specific, causal chain inferred. Nicotinic acetylcholine receptors in mood dysregulation (receptor pharmacology review; Progress in Neuro-Psychopharmacology): centers the alpha-4-beta-2 and alpha-7 subtypes modulating monoamine, glutamate, GABA, plasticity, and neuroimmune signaling with a bidirectional role in mood, pitching subtype-selective ligands, epigenetic regulation, and neuroimaging-guided dosing; candid that the translational record is poor. A method to disentangle risk and resilience genes (statistical genetics; Molecular Psychiatry): an adversarial multi-task neural network separates true genetic resilience from mere low risk by learning to distinguish high-risk-unaffected from affected-at-similar-risk while unlearning low-risk patterns; applied to Alzheimer's it yields resilience scores protective independently of the polygenic risk score and flags five resilience variants; proof-of-concept on simulations and existing datasets, variants need replication. SABLE (agentic AI / computational pharmacology; arXiv, full text): Synthetically-accessible Agentic Bayesian Ligand Exploration uses an LLM orchestrator routing to reaction-templated analog enumeration, ADMET/physicochemical prediction, structure-based affinity scoring, and Bayesian optimization — a computational twin of the analyze/prioritize stages of design-make-test-analyze with per-number provenance; open-source and modular, synthetic accessibility built in via reaction templates (not free generation); decision support, not autonomous chemistry, and limited by scoring-function accuracy. CLOSE — PRISM (mechanistic interpretability / computational neuroscience; arXiv, full text): adapts neuroimaging subtraction analysis to perturbed transformers, maps the seven Philadelphia Naming Test error categories, and runs a structurally matched analysis on 213 chronic post-stroke aphasia patients; both the layer-perturbed vision-language model and the human brains recover the same phonemic-favoring dissociation (a deep-layer cluster and a frontal-perisylvian cortical cluster, both replicating), the semantic direction a non-significant trend in both — a falsifiable, spatially resolved test of functional specialization, with the confirmatory causal intervention left to future work. Eleven segments per receptor-and-reason PROMPT.md. Source notes: journal content via PubMed metadata plus open nature.com and ScienceDirect article pages (Neuropsychopharmacology lead and discontinuation perspective read in full); bioRxiv neuroscience surveyed via the details API and category search (MCP intermittently 503/timeout); arXiv shared daily cache used for Tier 2/3; five bioRxiv picks deferred on 2026-08-13 (FTO/morphine, DMT vs 5-MeO-DMT, JAK/epilepsy, psilocybin cognitive-flexibility null, gut x LRRK2) remain unrendered and have now aged past the 7-day window. Paper links in show notes. https://doi.org/10.1038/s41386-026-02520-0 2026-08-14-receptor-and-reason Fri, 14 Aug 2026 12:00:00 +0000 742 false Daily roundup for August 14, 2026, eleven segments, a journal-heavy day out of Neuropsychopharmacology and Molecular Psychiatry with two computational pieces to close. LEAD — in rats, sons of methamphetamine-taking fathers inherit heightened meth susceptibility, the directional mirror of the cocaine-resistance seen in cocaine-sired sons; the effect is male-only, drug-specific, requires drug action on sperm, and, strikingly, the two drugs seem to write their accumbens marks through different cell types, neuronal for cocaine and glial for methamphetamine. A perspective argues that stopping a drug is an active neuroadaptive state deserving its own pharmacology, across rebound, receptor supersensitivity, allostasis, and gene-regulatory carryover. A hot-topics commentary makes the case for individualized low-dose lithium as a dose-tunable neuroprotective agent. Seven-tesla spectroscopy finds lateralized, left-right-asymmetric glutamate and GABA abnormalities in early-phase schizophrenia that partly track antipsychotic response. Chondrolectin-somatostatin interneuron deficits show up in cortex but not striatum, so vulnerability is cell type times circuit context. A review reframes glial GABA handling as a mitochondrial, bioenergetic control point. Hippocampal teneurin-4 knockdown drives depression-like behavior through an oligodendrocyte-differentiation block that clemastine reverses. A nicotinic-receptor review maps the alpha-4-beta-2 and alpha-7 subtypes as an underexploited mood target. An adversarial neural network separates genetic resilience from mere low risk and finds Alzheimer's resilience variants independent of polygenic risk. SABLE is an open-source agentic hit-to-lead framework with synthetic accessibility built in and provenance for every number. And PRISM imports neuroimaging subtraction statistics to give language-model interpretability a falsifiable, spatially resolved test, validated in parallel against 213 aphasia patients. Paper links in show notes. Daily roundup for August 13, 2026 — nine segments, psychedelic-heavy up top but broadening into opioid-reward circuitry, GLP-1 neuropharmacology, human disease modeling, neuroinflammation targets, and computational pharmacology. LEAD — DT120 / lysergide phase 3 in GAD (psychedelic drug development; full text via Psychiatric Times / Definium topline): a single supervised 100-microgram dose of an orally disintegrating LSD tablet vs placebo in 214 patients cut the Hamilton Anxiety score by ~11.6 vs ~6.2 points at week 12 (placebo-adjusted 5.4, p<0.0001, Cohen's d ~0.8), with separation by week 1, response 43% vs 16%, remission 14% vs 4%, no suicidality signal; second positive phase 3 (after an MDD trial in June) and would be the first new GAD mechanism in ~two decades. Caveats: functional-unblinding risk (a low-dose 50-microgram pivotal arm reads out next month) and an average >6 hours of in-clinic monitoring per dose. TCB-2 multidimensional phenotyping (serotonergic-psychedelic pharmacology; bioRxiv full text): head-fixed and freely moving mice measured for licking, pupil, eye position, blinking, locomotion, and automated head-twitch; the serotonin-2A agonist TCB-2 produced pupil constriction, broad licking suppression with preserved cue-locked timing, reduced locomotion, and dose-dependent head-twitches, but the selective 2A antagonist volinanserin only partially blunted pupil and head-twitch and did not block the rest — a methodological argument that single-assay characterization compresses a multidimensional pharmacology. Psychedelics and astrocytes (mechanism perspective; Molecular Psychiatry review, abstract): a hypothesis piece proposing a neuro-glial ensemble — astrocytic calcium, gliotransmission, metabolic coupling, immune gating — as the minimal functional unit translating serotonin-2A activation into durable circuit plasticity; framework, not new data. Psilocybin circulating microRNAs (psychedelic biomarker; bioRxiv full text): placebo-controlled, N=62, single oral dose, blood at baseline/6h/7d; two microRNAs changed at 6h (mapping to neuroplasticity and inflammatory pathways), none at 7d and none in placebo — transient, small, proof-of-concept for a peripheral molecular readout. Semaglutide, orexin, and hippocampal acetylcholine (GLP-1 neuropharmacology; bioRxiv full text): immediate-early-gene mapping shows acute semaglutide activates lateral-hypothalamic orexin and basal-forebrain cholinergic neurons in both sexes, and microdialysis shows increased acetylcholine efflux in ventral hippocampus in young and aged rats — a mechanistic GLP-1-to-cholinergic-cognition bridge, though the Fos/microdialysis design shows engagement, not direct GLP-1-receptor binding on orexin neurons. Cocaine-abstinence autocrine enkephalin (addiction circuitry; bioRxiv full text): after abstinence from repeated cocaine, D2 medium spiny neurons release enkephalin that acts on their own mu opioid receptors to suppress GABA release onto the ventral pallidum, disinhibiting it and driving cocaine seeking; a conditional Penk knockout in D2 neurons abolishes the GABA suppression and exogenous met-enkephalin recapitulates it — a druggable node for cocaine use disorder, ex vivo slices. Mesocortical assembloids (human disease modeling; bioRxiv full text): iPSC forebrain-midbrain (corticodopaminergic) assembloids that are morphologically identical spontaneously bifurcate into disease-prone and non-prone states, resolved nondestructively by high-density electrophysiology, a Raman-based tau/neurofilament readout, and spatial transcriptomics, and rolled into a composite risk score — reframing inter-organoid variability as biologically meaningful vulnerability; in vitro, a signature not a validated disease. NLRP3 inflammasome (neuroinflammation target; Molecular Psychiatry review, abstract): consolidates NLRP3 biology, evidence for dysregulation in schizophrenia and related disorders, inhibitors tried, and biomarker potential, candid that causal and clinical evidence remain limited. CLOSE — DegradeQuery (computational pharmacology; arXiv full text): PROTAC degradation depends jointly on the degrader and cellular context, but labels exist for a small fraction of molecule-target-E3 records; a counterfactual tuple pretraining objective contrasts recorded tuples against perturbed alternatives to turn label-missing records into pretraining signal, beating supervised baselines on classification and generalization — not agentic, but a transferable trick for the sparse, context-dependent data endemic to computational pharmacology. Nine segments per receptor-and-reason PROMPT.md. Deferred (full text not yet rendered on bioRxiv, posted Aug 11-12, revisit on a later day): FTO/m6Am control of morphine reward and tolerance without altering analgesia (2026.08.06.743062); DMT vs 5-MeO-DMT behavioral/TrkB dissociation (2026.08.06.743248); JAK inhibition overcoming drug-resistant epilepsy (2026.08.06.743311); a psilocybin cognitive-flexibility null result (2026.08.06.743309); gut infection x LRRK2 G2019S driving Parkinsonian pathology (2026.06.08.728789). Source notes: bioRxiv details API fetched day-by-day (multi-day-range pagination broke at offset 390; the API was also unusually slow, ~37s/request); Nature and Elsevier psychiatry journals auth-walled (review abstracts via PubMed); r.jina.ai to bioRxiv full text intermittently HTTP 429 (retried with spacing); arXiv shared daily cache used for Tier 2/3. Paper links in show notes. https://www.psychiatrictimes.com/view/dt120-lysergide-shows-rapid-lasting-relief-for-generalized-anxiety-in-phase-3-trial 2026-08-13-receptor-and-reason Thu, 13 Aug 2026 12:00:00 +0000 645 false Daily roundup for August 13, 2026, nine segments, psychedelic-heavy up top and broadening underneath. LEAD — Definium's DT120, a single-dose orally disintegrating LSD (lysergide) tablet, posts a large, durable phase 3 win in generalized anxiety (Hamilton Anxiety down ~11.6 vs ~6.2 at week 12, Cohen's d ~0.8, rapid and lasting), which would be the first new GAD mechanism in ~two decades; caveats are functional unblinding (a low-dose pivotal arm reads out next month) and a supervised, multi-hour clinic day per dose. Multidimensional mouse phenotyping shows a serotonin-2A tool compound's effects aren't fully blocked by a 2A antagonist, so the standard head-twitch readout undersells the pharmacology. A Molecular Psychiatry perspective argues astrocytes, not just neurons, are the integrative unit of psychedelic action. A single psilocybin dose shifts two circulating microRNAs at six hours but nothing by seven days. Semaglutide activates orexin and basal-forebrain cholinergic neurons and raises hippocampal acetylcholine in young and aged rats, a GLP-1-to-cognition bridge. Cocaine abstinence drives autocrine enkephalin from D2 neurons that suppresses their own GABA output and disinhibits the ventral pallidum to fuel drug seeking. Morphologically identical human assembloids spontaneously split into disease-prone and non-prone states on a nondestructive multimodal platform. A review consolidates the NLRP3 inflammasome as an actionable-but-unproven psychiatric target. And DegradeQuery uses counterfactual pretraining to turn label-missing PROTAC records into signal. Paper links in show notes. Daily roundup for August 12, 2026 — nine segments, an almost entirely bioRxiv day (the arXiv listing API was hard rate-limited from this host and the Nature psychiatry journals were behind their login wall), heavy on synaptic and receptor mechanism plus computational drug design. LEAD — LRRC57/RABIN, a presynaptic Rab brake (synaptic neuropharmacology; full text): a St. Jude group mines the conserved, brain-enriched dark proteome and identifies LRRC57 — named RABIN, for Rab inhibitor — a 27-kDa horseshoe leucine-rich-repeat protein on glutamatergic synaptic vesicles that binds multiple GTP-loaded synaptic Rabs and competitively occupies the effector-binding surface, a noncanonical decoy-effector mechanism beyond the classical GEF/GAP/GDI machinery; forebrain-excitatory deletion increases glutamate release, expands vesicle pools, speeds turnover, and produces hyperexcitability with epileptiform activity and a lowered kainic-acid seizure threshold that is rescued by the SV2A antiseizure drug levetiracetam, while overexpression suppresses release and protects against induced seizures — not an etiologic disease gene but an epistatic modifier with therapeutic upside, explicitly analogized to PCSK9. HOMER1 R297W de novo variant (synaptic scaffolding; abstract): an arginine-to-tryptophan substitution in the coiled-coil tetramerization domain acts dominant-negative, blunting BDNF-directed growth-cone turning and store-operated calcium entry, lowering spine density and ER infiltration into spines, reducing mGluR5 expression and group-I mGluR calcium responses, and declustering mGluR5/IP3R/STIM by super-resolution imaging — a mechanistic bridge from one point mutation to epilepsy/ASD connectivity defects; pairs with RABIN as loss-of-restraint-at-the-synapse. DMT and microglia (psychedelics / neuroinflammation; abstract, body not yet rendered): N,N-dimethyltryptamine directly pushes LPS-activated microglia back toward homeostatic morphology and lowers phagocytosis; proteomics show selective suppression of cytokine/chemokine and oxidative-stress pathways while sparing arachidonic-acid-prostaglandin synthesis; DMT binds microglial sigma-1 receptors at micromolar affinity, morphological reprogramming needs both sigma-1 and serotonergic signaling whereas the phagocytosis drop is largely receptor-independent, and in oxygen-glucose-deprived slices DMT reduces spreading depolarizations and neuronal injury via serotonergic signaling — mechanistic support for DMT's stroke-neuroprotection program, with the sigma-1 arm hinting at neuroprotection separable from the psychedelic experience. Lateral-habenula subtypes (reward/aversion circuits; abstract): multidimensional profiling ties spatial gene expression, intrinsic electrophysiology, and projection targets of transgenically defined LHb subpopulations to behavior, showing two genetically defined subtypes differentially encode value/prediction-error versus directional movement across parallel projections to serotonergic and dopaminergic midbrain — the subtype resolution needed before precise habenular antidepressant manipulation. Arc capsid signaling and sleep (neuromodulation; abstract): in Drosophila, retrotransposon-derived Arc capsid formation links serotonergic and dopaminergic PAM neurons to set sleep depth — loss of fly Arc deepens/consolidates sleep and raises arousal threshold, requires dArc1 in serotonergic (and, less, PAM) neurons, and depends on capsid assembly, identifying a domesticated viral protein as an intercellular signaling channel in the monoaminergic sleep circuit. NDST1 substrate reduction for MPS IIIC (computational pharmacology; abstract): a structure-based pipeline docks ~4.1 million drug-like plus FDA-approved compounds against the NDST1 sulfotransferase domain, then applies pharmacokinetic filtering, microsecond molecular dynamics, and MM/PBSA free-energy calculations, and in parallel diffusion-designs peptide binders, yielding four small-molecule scaffolds and three peptides (lead predicted binding free energy ~-13.4 kcal/mol, large error bar) — an in-silico case that NDST1 substrate reduction is worth pursuing for Sanfilippo syndrome type C, where enzyme replacement barely crosses the blood-brain barrier. Boltz-Perturb (computational structure/drug discovery; abstract): protein-ligand co-folding models already contain correct binding-mode basins (shown by true-coordinate injection), so the failure is sampling; two training-free inference-time conditioning perturbations improve top-20 pose success 2.6-7.8x and beat high-diffusion-temperature sampling with over 75% less compute. De novo binder prioritization (computational protein design; abstract): designing binders for capsaicin, limonene, and quercetin shows that high global structure scores do not guarantee a valid local pocket, and only a subset preserve pocket geometry and ligand contacts after sequence design and re-folding — evaluate the local ligand environment explicitly. CLOSE — Vorapaxar and aripiprazole suppress HBV (drug repurposing / pharmacology; abstract): an FDA-approved-drug screen finds the antipsychotic aripiprazole (via ERK/JNK-dependent HNF4 knockdown) and the PAR-1 antagonist vorapaxar (via PAR-1/SRC/EGFR) both suppress hepatitis B transcription through host STAT3 signaling and hold up in human liver chimeric mice — a liver-and-virus result included as a reminder of CNS drugs' promiscuous host-target pharmacology. Nine segments per receptor-and-reason PROMPT.md. Deferred (not-yet-rendered full text, revisit): a cortico-thalamic multi-head self-attention theory of cognition validated against human intracranial recordings (bioRxiv 2026.08.06.743205, posted Aug 10, body HTML not yet rendered). Sources unavailable this run: arXiv listing API (HTTP 429/rate-limited, web-search fallback returned only out-of-window items); Neuropsychopharmacology and Molecular Psychiatry (idp.nature.com login wall); Psychiatric Times (HTTP 403). Paper links: https://www.biorxiv.org/content/10.64898/2026.08.05.743108v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.741753v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.742931v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.743065v1 ; https://www.biorxiv.org/content/10.64898/2026.08.06.743357v1 ; https://www.biorxiv.org/content/10.64898/2026.08.09.743834v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.742877v1 ; https://www.biorxiv.org/content/10.64898/2026.08.08.743643v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.743121v1 https://www.biorxiv.org/content/10.64898/2026.08.05.743108v1 2026-08-12-receptor-and-reason Wed, 12 Aug 2026 12:00:00 +0000 719 false Daily roundup for August 12, 2026, nine segments, an almost entirely bioRxiv day (arXiv was rate-limited and the Nature psychiatry journals were behind their login wall), heavy on synaptic mechanism and computational drug design. LEAD — a St. Jude group names a conserved brain protein RABIN and shows it is a noncanonical decoy that competitively blocks active synaptic Rab GTPases from their effectors; deleting it drives glutamate release, hyperexcitability, and levetiracetam-rescuable seizures, while overexpressing it protects — a presynaptic-release rheostat framed as an epistatic, PCSK9-like target. A new de novo HOMER1 variant declusters the mGluR5-calcium scaffold and blunts spine and growth-cone signaling. The endogenous psychedelic DMT directly reprograms microglial inflammation through sigma-1 and serotonergic signaling, an actionable mechanism for its stroke-neuroprotection program. Genetically defined lateral-habenula subtypes split value and prediction-error coding from movement coding across parallel serotonergic and dopaminergic channels. Arc retrotransposon-derived capsid signaling between serotonergic and dopaminergic neurons sets sleep depth in flies. A four-million-compound virtual screen with microsecond molecular dynamics and diffusion peptide design nominates NDST1 substrate-reduction candidates for Sanfilippo type C. Boltz-Perturb recovers correct protein-ligand co-folding poses at inference time, three-to-eight-fold better for a quarter of the compute. A de novo binder study shows global structure scores lie about the local pocket. And the antipsychotic aripiprazole and the antiplatelet vorapaxar both suppress hepatitis B transcription through host STAT3 signaling. Paper links in show notes. Daily roundup for August 11, 2026 — nine segments spanning agentic AI for scientific discovery and drug safety, an AI-safety cautionary tale in pharmacogenomics, Alzheimer's receptor mechanism, human fear-extinction pharmacology, a sex-specific addiction circuit, and psychedelics. LEAD — Model Discovery Agent (agentic AI / computational neuroscience; full text): a UBC computer scientist couples an LLM hypothesis-proposer with sequential Monte Carlo, simulation-based inference, and value-of-information experiment design to discover latent mechanistic models from few interventions, operating in the "M-open" regime where the LLM expands the hypothesis space when the current model fails a predictive check. On a new biology benchmark, NeuronBench, built from six simulated Hodgkin-Huxley "mystery neurons" tuned to be indistinguishable under textbook stimulation protocols, the combined system is substantially more data-efficient than an LLM baseline alone; entirely simulated data, not real electrophysiology recordings. Recursion/Genentech agentic target discovery (industry / agentic AI; full text): Genentech has advanced the first target from the companies' neuroscience collaboration into early small-molecule discovery, sourced from a whole-genome CRISPR knockout map built from over a trillion iPSC-derived-neuron data points; Recursion now credits a "Target Discovery Agent" pairing frontier LLM reasoning with its proprietary maps for the find, alongside sibling drug-design and clinical-strategy agents already cutting analysis time and improving trial enrollment — target still undisclosed, early discovery stage only. ATLAS (agentic AI / computational pharmacology; full text): a Hong Kong University of Science and Technology system builds a patient-specific medication-conflict graph via targeted follow-up questions, then runs a risk-first multi-agent policy; beats Gemini 3.1 Pro by 54 points on Strict Success Rate and about 15 points on aggregate safety-reasoning score with zero unsafe recommendations under automated scoring, and outscores it on all five criteria in a 40-case blinded clinician review (modest sample, moderate inter-rater agreement). Genotypic Triggers (AI safety / pharmacogenomics; full text): a backdoor shifts generative antimicrobial-peptide models' output toward elevated predicted immunogenicity specifically for carriers of a targeted HLA allele — a more than sevenfold average increase — while potency, general toxicity, and diversity stay near baseline, so the poisoned model passes conventional safety screens; proof of concept, not an observed in-the-wild attack. Pharmacogenomic variant-effect benchmarking (computational pharmacogenomics; abstract): current predictors, including AlphaMissense and protein-language-model methods, reliably flag CYP2C9 loss-of-function variants but struggle with gain-of-function fast-metabolizer variants; a structure-plus-coevolution method, StructureDCA, performs best. p75 neurotrophin receptor and Tau (Alzheimer's mechanism; full text): Carlos Ibáñez's group engineers p75 mutants that surgically spare or cut only the receptor's RhoA-ROCK arm; mice lacking just that arm improve on every Tau-pathology, gliosis, atrophy, and memory measure tracked, while the receptor's NF-kB and JNK/caspase arms stay intact — first genetic (not purely pharmacological) evidence pinning Tau pathology to that one signaling branch, relevant to a p75-modulator drug that showed biomarker but not cognitive benefit in a phase 2 AD trial. Sodium butyrate and human fear extinction (translational psychiatry; full text): in 180 healthy volunteers, the HDAC inhibitor sodium butyrate reduces one-week return of fear at retrieval only after a sufficiently long extinction protocol, with no effect on shock-triggered reinstatement — consolidation of adequately learned extinction, not blanket relapse protection. Sex-specific binge-drinking circuit (addiction; full text): a Florey Institute (Melbourne) team finds the basolateral-amygdala-to-nucleus-accumbens-core projection preferentially engaged in female mice during binge-like alcohol access, and silencing it cuts alcohol intake only in females with no effect on a sucrose control — modest effect size, one of likely several contributing circuits, but a female-specific pathway a male-only screen would have missed. CLOSE — Psilocybin and dissociable empathy (psychedelics; abstract, self-contained): a Maastricht-led placebo-controlled trial splits empathy into cognitive accuracy, emotional arousal, and empathic concern; psilocybin acutely impairs positive-emotion identification while raising arousal and concern, all effects gone within a week, but the acute rise in empathic concern specifically predicts later positive psychological change, independent of a concurrent oxytocin increase. Nine segments per receptor-and-reason PROMPT.md. Deferred (not-yet-rendered full text): companion psilocybin miRNA biomarker preprint (bioRxiv 2026.08.04.742716, posted Aug 10, same cohort as the empathy study) and a maternal NKG2D-immune-activation/fetal-neurodevelopment paper (Molecular Psychiatry, DOI 10.1038/s41380-026-03809-8) whose body text was not retrievable through the Nature paywall bypass on this run. Paper links: https://arxiv.org/abs/2608.09696 ; https://www.globenewswire.com/news-release/2026/08/05/3339126/0/en/recursion-reports-second-quarter-financial-results-genentech-options-first-neuroscience-target-into-early-discovery-program.html ; https://arxiv.org/abs/2608.09443 ; https://arxiv.org/abs/2608.06779 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742561v1 ; https://www.nature.com/articles/s41380-026-03810-1 ; https://www.nature.com/articles/s41380-026-03802-1 ; https://www.nature.com/articles/s41386-026-02513-z ; https://www.biorxiv.org/content/10.64898/2026.08.04.742704v1 https://arxiv.org/abs/2608.09696 2026-08-11-receptor-and-reason Tue, 11 Aug 2026 12:00:00 +0000 757 false Daily roundup for August 11, 2026, nine segments spanning agentic AI for scientific discovery and drug safety, an AI-safety cautionary tale in pharmacogenomics, Alzheimer's receptor mechanism, human fear-extinction pharmacology, a sex-specific addiction circuit, and psychedelics. LEAD — a UBC researcher's Model Discovery Agent couples an LLM hypothesis-proposer with Bayesian experiment design to discover hidden mechanisms from few interventions, beating an LLM-alone baseline on a new simulated single-neuron electrophysiology benchmark. Recursion credits an agentic "Target Discovery Agent" for the first neuroscience target Genentech has advanced into early discovery from a trillion-datapoint CRISPR knockout map. ATLAS, a graph-and-multi-agent medication-safety system, beats the strongest proprietary LLM by wide margins on safety metrics for older adults on multiple medications. A backdoor attack shows antimicrobial-peptide generators can be poisoned to spike immunogenicity risk for one genetic subgroup while sailing through normal safety screens. Current pharmacogenomic variant predictors still miss fast-metabolizer CYP2C9 variants. Genetically isolating one signaling arm of the p75 neurotrophin receptor reverses Tau pathology and memory deficits in mice. The HDAC inhibitor sodium butyrate strengthens human fear-extinction memory, but only with sufficiently long extinction training. A female-specific amygdala-to-accumbens circuit drives binge-like alcohol drinking in mice. And psilocybin dissociates empathy into pieces, with empathic concern — not accuracy — predicting later well-being. Paper links in show notes. Daily roundup for August 10, 2026 — ten segments across agentic-AI GPCR drug discovery, receptor biophysics, Parkinson's synaptic and clinical dynamics, Alzheimer's resilience genetics, an autism receptor-pharmacology rescue, an addiction model, threat and safety circuits, and an EEG study of altered bodily self. LEAD — Multi-agent GPCR subtype-selective allosteric discovery (agentic AI / computational pharmacology; full text): a five-to-six-agent workflow orchestrated by Claude Sonnet 4.6 and Opus 4.7 encodes the molecular surfaces of 286 non-olfactory class A GPCRs (AlphaFold models) with dMaSIF fingerprints, aligns them by Ballesteros-Weinstein position, and identifies the most divergent lipid-facing regions between subtype pairs; the top divergent regions overlap ~85 to 99 percent of known allosteric sites within 8 angstroms, validating the signal; paired target vs off-target docking of one million lead-like ZINC20 compounds with Uni-Dock, then multi-seed detail-mode redocking, ADMET-AI, MM/GBSA, and retrosynthesis, yielded 352 candidate selective ligands across 104 of 163 receptor pairs, filed into a searchable database. Two takeaways: fast-mode docking overstated selectivity, with the median target-minus-off-target gap collapsing by more than 2 kcal/mol after careful redocking (most quick-screen hits were sampling artifacts); and the language model plans, synthesizes, and interprets but cannot override the numerical eligibility gates, with deterministic docking/energy/property tools, SHA-256 provenance, and human-review checkpoints doing the deciding — a disciplined agentic-science architecture. All in silico; static structures and docking only; no wet-lab validation yet — a hypothesis-and-database generator. Neuromuscular acetylcholine-receptor allostery (receptor biophysics; abstract): kinetic modeling, phi and activation-energy estimates across 60+ residues, and MD argue for parallel major and minor gating pathways, non-identical for liganded vs unliganded receptors, with C-loop capping initiating allosteric communication independent of agonist. Cortex-restricted alpha-synuclein pathology (Parkinson's / synaptic; full text): injecting preformed fibrils into M2 cortex confines aggregates to vGLUT1-positive corticostriatal terminals, sparing SPN somata and dopamine terminals; six weeks in, glutamatergic transmission is impaired (reduced evoked and miniature EPSCs, lower synaptic density) with no dopamine loss or cortical cell death — pathologic alpha-synuclein alone is sufficient for corticostriatal synapse failure, nominating early synaptic dysfunction as a target. Levodopa and whole-brain microstate dynamics (Parkinson's / MEG; abstract): in 13 bradykinetic patients, resting MEG microstate transition patterns are reproducible within but differ between OFF and ON medication, and the extent of departure from the OFF-state baseline tracks motor improvement — a candidate objective efficacy readout (small n). Oligogenic Alzheimer's resilience (AD genetics; full text): single-nucleus and spatial transcriptomics plus WGS of entorhinal cortex in PSEN1 E280A carriers show the strongly protected male RELN-COLBOS carrier (dementia delayed ~25 years) has unique excitatory-neuron and oligodendrocyte populations signaling through a non-canonical Reelin receptor (LRP6) and carries additional rare variants absent in his weakly protected sister — resilience is oligogenic, though the shared Reelin-Dab1 axis remains the tractable thread. Neurotensin-receptor-1 rescue in autism (receptor pharmacology; full text): the brain-penetrant NTSR1 agonist PD149163, given systemically or infused into entorhinal cortex, restores odor generalization in novel backgrounds in Cntnap2 and Shank3B mouse models by reducing neural responses to novel distractor odors, plausibly a general generalization mechanism. Oral fentanyl self-administration (addiction; abstract): a route-appropriate open model in C57 and CD1 mice dissociates escalation-and-relapse (inbred) from high-intake-no-escalation (outbred) phenotypes; port-approach latency, not lever presses, reveals cue learning in outbred mice; severity scores vary continuously. Parabrachial oxytocin-receptor neurons (threat circuits; abstract): lateral parabrachial Oxtr neurons are activated by direct aversive stimuli and by observing a conspecific's distress; chemogenetic inhibition alters social proximity and pain contagion without changing general anxiety, while a selective Oxtr agonist is anxiogenic — a node fusing felt and witnessed threat. Fear versus safety learning (fear circuits; abstract): safety learning produces tone-specific, context-generalizing freezing suppression and a distinct amygdalar signature (elevated PSD-95 and GluA1) versus fear's elevated GluN1, a molecular handle for disorders of impaired safety learning. CLOSE — Out-of-body and unusual bodily experiences (EEG / consciousness; abstract): 36 UBE episodes across REM, light sleep, arousals, and meditative wakefulness show reduced neural irreversibility (flattened cortical hierarchy) with concurrently elevated whole-brain complexity toward waking levels — a dynamical profile resembling psychedelic states. Ten segments per receptor-and-reason PROMPT.md. Deferred (not-yet-rendered full text): a foundational in vivo home-cage platform predicting human clinical outcomes from 24 hours of rodent behavior (bioRxiv 2026.08.03.742611, posted Aug 9, body HTML not yet rendered). Paper links: https://www.biorxiv.org/content/10.64898/2026.08.06.743397v1 ; https://www.biorxiv.org/content/10.64898/2026.08.02.742316v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742532v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742400v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742644v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741644v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741697v1 ; https://www.biorxiv.org/content/10.64898/2026.08.06.743398v1 ; https://www.biorxiv.org/content/10.64898/2026.08.06.743388v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741513v1 https://www.biorxiv.org/content/10.64898/2026.08.06.743397v1 2026-08-10-receptor-and-reason Mon, 10 Aug 2026 12:00:00 +0000 752 false Daily roundup for August 10, 2026, ten segments spanning agentic-AI GPCR drug discovery, receptor biophysics, Parkinson's synaptic and clinical dynamics, Alzheimer's resilience genetics, an autism receptor rescue, an addiction model, threat and safety circuits, and an EEG study of altered bodily self. LEAD — a five-to-six-agent workflow orchestrated by Claude language models mines the divergent GPCR-lipid membrane interface across 286 class A receptors, validating that the most divergent surface regions coincide with known allosteric sites, then paired target versus off-target docking of a million compounds nominates 352 subtype-selective allosteric ligand candidates; fast-mode docking badly overstated selectivity until careful redocking, and the language model orchestrates but cannot override deterministic numerical gates — a disciplined agentic-science template, though entirely in silico. Cortex-restricted alpha-synuclein fibrils impair corticostriatal glutamate transmission with no dopamine loss, isolating presynaptic synucleinopathy as sufficient for synapse failure. Levodopa reconfigures whole-brain MEG microstate dynamics, and departure from the OFF-state baseline tracks motor gains. Deep phenotyping shows Alzheimer's resilience in the RELN-COLBOS carrier is oligogenic rather than one gene. The neurotensin-receptor-1 agonist PD149163 rescues olfactory generalization in two autism models via the entorhinal cortex. An open oral fentanyl model dissociates escalation-prone from relapse-prone phenotypes by genetic background. Parabrachial oxytocin-receptor neurons fuse felt and witnessed distress into defensive behavior. Fear and safety learning leave distinct amygdalar receptor signatures. And out-of-body EEG shows a flattened cortical hierarchy with elevated complexity, a psychedelic-like fingerprint. Paper links in show notes. Daily roundup for August 9, 2026 — nine segments spanning striatal dopamine physiology, glutamate-receptor biology in Alzheimer's and Rett, GPCR structural dynamics, two structure-based design tools, a pharmacogenomics variant-prediction reality check, and two behavioral items on alcohol and food reward. LEAD — Dopamine's physiological actions in vivo (dopamine/striatum; full text): whole-cell membrane-potential recordings from 300+ identified striatal projection neurons in awake mice, with independent optogenetic control of cortical input and dopamine release plus a genetically encoded dopamine sensor; phasic dopamine over seconds-to-minutes produced only modest synaptic effects and NO detectable change in membrane-potential dynamics or intrinsic excitability, challenging the slice-derived acute-excitability model; associative learning strengthened corticostriatal synapses onto both D1- and D2-SPNs, but only D1 potentiation required dopamine (D2 persisted under dopamine-receptor blockade, implicating adenosine A2A) — reframing dopamine's principal role as gating learning-related plasticity, reconciled with slice work via a tonic-dopamine excitability set-point. NMDA-receptor density across the AD continuum (NMDAR/AD; full text): quantitative in vitro autoradiography of hippocampus, entorhinal, and parietal cortex with NMDAR and tau ligands across controls/MCI/AD; steepest loss in CA1 in both sexes, but a diagnosis-by-sex interaction driven by lower NMDAR density in MCI women (not MCI men); NMDAR density correlated positively with MMSE, and entorhinal NMDAR correlated negatively with tau only in AD men (correlational). AMPA-receptor PAM in Rett (AMPAR/neurodevelopment; full text): the clinically advanced ampakine CX1632 in Mecp2-null males and Mecp2-het females shows efficacy gated by developmental stage — brief neonatal treatment gave long-lasting gains in survival, disease progression, motor, and cognition and sustained neuronal/synaptic gene programs plus rescued AMPAR transmission weeks after withdrawal, while later dosing helped symptomatic null males much less; disease stage as a determinant of responsiveness (preclinical). GLP-1R-Gs structural dynamics (GPCR/structural pharmacology; full text): hydrogen-deuterium exchange mass spectrometry on a pre-coupled GLP-1R-Gs complex; non-peptide agonists danuglipron and an experimental compound gave overlapping flexibility fingerprints with drug-specific transmembrane effects, the natural GLP-1 hormone stabilized the backbone more weakly, and its inactive metabolite destabilized locally; both peptides uniquely modulated the G-protein switch-III loop — efficacy encoded in dynamics, supporting dynamics-integrated design of oral non-peptide agonists. CryoLigATE (comp drug discovery/cryo-EM; full text): a hybrid convolutional-transformer network trained on 6,000+ protein-ligand complexes enhances ligand density in cryo-EM maps, tested on ~649 complexes, improving resolvability of poorly resolved binding sites while preserving good density, in seconds on a desktop GPU with no manual prep. De novo ligand-induced conformational switch (protein design; full text): adding a mobile lid domain that closes behind a bound ligand converts a preorganized exatecan (topoisomerase-I-inhibitor) binder into a drug-triggered shape-changer with sub-nanomolar affinity, 100x selectivity, months-long residence, and tunable kinetics, and yields a genetically encoded fluorescent drug biosensor — programmable mechanical response, not just recognition. CYP450 variant-effect prediction breaks down (pharmacogenomics/AI; full text): homology/conservation-based predictors (AlphaMissense, ESM-2) fail on cytochrome-P450 pharmacogenes — AlphaMissense nearly doubles its ambiguous-call rate and its scores are uncorrelated with CYP2C9 deep-mutational-scanning activity; a k-NN model over ESM-2 embeddings ensembled in improved the ambiguous-class correlation roughly tenfold, yet still failed to match clinical annotations, arguing that for multi-substrate enzymes function must be redefined as substrate-conditioned. Bmal1 in the dorsomedial striatum and alcohol (circadian/addiction; abstract): MSN-specific Bmal1 deletion in the dorsomedial (not dorsolateral) striatum reduced alcohol consumption and preference in females only, without changes in anxiety/depressive/motor/sucrose measures, and ovariectomy abolished the effect — subregion-, cell-type-, and ovarian-hormone-specific. CLOSE — Protein restriction and accumbens dopamine (reward/dopamine; abstract): with a genetically encoded dopamine sensor in NAc core, protein restriction increased operant responding (especially for grain) and selectively enlarged the pellet-delivery dopamine response for grain over sucrose during restriction under FR1 (absent on normal diet and under progressive ratio) — internal state biasing dopamine's valuation toward a needed nutrient. Nine segments per receptor-and-reason PROMPT.md. Deferred (not-yet-rendered full text): a five-agent workflow uncovering subtype-selective allosteric sites at the GPCR-lipid interface (bioRxiv 2026.08.06.743397) and alpha-synuclein aggregates impairing corticostriatal glutamatergic transmission (bioRxiv 2026.08.03.742532). Paper links: https://www.biorxiv.org/content/10.64898/2026.08.05.742858v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.743051v1 ; https://www.biorxiv.org/content/10.64898/2026.08.04.742773v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742488v1 ; https://www.biorxiv.org/content/10.64898/2026.08.04.742718v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742366v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741616v1 ; https://www.biorxiv.org/content/10.64898/2026.08.02.742221v1 ; https://www.biorxiv.org/content/10.64898/2026.07.31.741998v1 https://www.biorxiv.org/content/10.64898/2026.08.05.742858v1 2026-08-09-receptor-and-reason Sun, 09 Aug 2026 12:00:00 +0000 708 false Daily roundup for August 9, 2026, nine segments spanning striatal dopamine physiology, glutamate-receptor biology in Alzheimer's and Rett, GPCR structural dynamics, two structure-based design tools, a pharmacogenomics reality check, and two behavioral items on alcohol and food reward. LEAD — in vivo whole-cell recordings in behaving mice show phasic dopamine does not acutely tune striatal excitability; instead it selectively gates learning-driven corticostriatal potentiation, and only in D1-type spiny projection neurons, reframing dopamine's principal physiological job as plasticity gating rather than moment-to-moment excitability control. Quantitative autoradiography finds sex-divergent NMDA-receptor loss across the Alzheimer's continuum, appearing earlier in MCI women. The AMPA-receptor potentiator CX1632 rescues Rett models only within a neonatal developmental window, with benefits outlasting the drug. Hydrogen-deuterium exchange on a pre-coupled GLP-1R-Gs complex maps agonist efficacy to backbone dynamics and a G-protein switch-III loop and distinguishes the non-peptide agonist danuglipron from the natural hormone. CryoLigATE sharpens ligand density in cryo-EM maps for structure-based design. A de novo lidded protein that changes shape on drug binding turns an exatecan binder into a sub-nanomolar, months-residence fluorescent biosensor. Homology-based variant-effect predictors break down on cytochrome-P450 pharmacogenes, and even a tenfold accuracy fix fails clinical annotations because enzyme function is substrate-conditioned. Dorsomedial-striatal Bmal1 deletion cuts alcohol intake in females through an ovarian-hormone-dependent mechanism. And protein restriction selectively biases nucleus-accumbens dopamine toward protein-containing food. Paper links in show notes. Daily roundup for August 8, 2026 — ten segments (eleven preprints) running from a psychedelic through the receptors and circuits shaping mood, motivation, and social pain, into a metabolic angle on Parkinson's, then two computational drug-discovery tools and a light phenomenology close. The psilocybin plasticity map held earlier in the week has now rendered and leads. LEAD — Psilocybin synaptic-plasticity map (psychedelics; full text): a single 1 mg/kg dose in mice, then whole-cell patch-clamp of miniature excitatory postsynaptic currents (mEPSCs) 24 h later across many cortical/limbic regions; psilocybin raised mEPSC frequency (not amplitude) selectively in ventrolateral orbital, prelimbic, anterior insula, and basolateral amygdala, sparing neighboring subregions; effect size tracked local Htr2a/Htr2c/Htr1a expression (matched to a spatial-transcriptomic atlas), and a region-confined CRISPR/SaCas9 Htr2a knockout abolished the frequency increase — a causal, spatially-resolved link between a psychedelic's lasting synaptic effect and where its 5-HT2A target lives (24-h snapshot; new vs unsilenced synapses undetermined). GPR151 habenula target (depression; abstract): the orphan GPCR GPR151 is conserved-topography enriched in the habenula; Gpr151 knockout blunts LPS-induced passive coping despite equal weight loss/immune activation, and adult habenula-restricted re-expression restores inflammatory behavioral vulnerability in males (female rescue insufficient) — a regionally precise candidate target for inflammation-associated depression. Kappa-opioid PET in partner separation (KOR; full text): in monogamous titi monkeys, nucleus-accumbens KOR availability fell during long-term partner separation in both sexes (increased dynorphin occupancy vs downregulation unresolved) alongside elevated cortisol, while plasma oxytocin fell only in males — accumbens KOR and peripheral oxytocin look like partially independent processes (n=16; PET displacement caveat). Mu-opioid receptor and emotion regulation (endogenous opioid; abstract): [11C]carfentanil PET plus fMRI reappraisal shows higher baseline MOR availability predicts stronger frontal/temporal/parietal responses, driven mainly by passive viewing, with some regions attenuating during active reappraisal (correlational, cross-sectional). Dopaminergic signatures of pessimism (dopamine; full text): DA→NAc recordings in a Pavlovian ambiguous-cue task dissociate state pessimism (reduced cue dopamine plus a compensatory, self-correcting positive outcome shift) from trait-like pessimism (reduced cue dopamine plus a negatively shifted outcome balance from larger omission responses); modeling favors asymmetrically weighted prediction-error signaling over asymmetric learning rates. Guanosine metabolism in Parkinson's (bioenergetics; full text; Ryan lab): guanosine-nucleotide metabolism has an outsized effect on synaptic transmission, with nucleoside diphosphate kinases (NDKs) connecting nucleotide pools to release machinery; the pathway runs through GCH1 (a known PD gene) and is supported by human PD genetics (GP2), with guanosine rescuing PARK-mutation synaptic dysfunction and IMPDH inhibition (ribavirin) perturbing transmission — Parkinson's as a synaptic-bioenergetics failure with a supplementable metabolite (preclinical). Microglia and alcohol (addiction/neuroimmune; abstract): chronic intermittent ethanol drove pro-inflammatory microglial activation; escalated self-administration tracked anterior-insula interferon regulatory factor 7 (IRF7), and minocycline blunted both IRF7 and escalation in early abstinence — neuroimmune signaling as a driver of drinking, with a repurposable drug. Histamine-GABA vesicle architecture (neurotransmission; abstract): a glutaraldehyde-NaBH4 epitope method reveals histamine condensed into a dense intraluminal core with GABA toward the rim inside conventional GABAergic vesicles, conserved across central/autonomic/endocrine systems — a structural basis for temporally differentiated dual-transmitter release that dissolves the clear-vs-dense-core dichotomy (needs functional confirmation). Computational drug-discovery tools (comp pharm/AI; abstracts): Nesso-1 (Recursion) is a coarse-grained cofolding binding-affinity model at ~1 s/prediction on one GPU, >10x faster than Boltz-2 while matching/beating its accuracy including out-of-distribution, open-sourced; ClinOracle is a hierarchical graph neural network predicting patient-derived functional activity conditional on target engagement, with an ADME/drug-likeness Priority Score, prospectively validated across five oncology/autoimmune/neuroinflammatory targets out to in vivo efficacy. CLOSE — ALTERD naturalistic reports (psychedelics; abstract): language/topic modeling of ~2,242 app reports finds psychedelic experiences most semantically like dreams yet uniquely aligned with meditation on personal-growth themes, whereas cannabis reads like meditation semantically but lacks the transformative content (self-selected users; suggestive). Ten segments per receptor-and-reason PROMPT.md. Deferred (not-yet-rendered full text): a five-agent workflow uncovering subtype-selective allosteric sites at the GPCR-lipid interface (bioRxiv 2026.08.06.743397) and a parabrachial oxytocin-receptor social-distress circuit (bioRxiv 2026.08.06.743398). Paper links: https://www.biorxiv.org/content/10.64898/2026.07.31.741703v1 ; https://www.biorxiv.org/content/10.64898/2026.08.02.742327v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.739928v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.739761v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741818v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741503v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741570v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.743008v1 ; https://www.biorxiv.org/content/10.64898/2026.08.01.742196v1 ; https://www.biorxiv.org/content/10.64898/2026.08.02.742378v1 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741354v1 https://www.biorxiv.org/content/10.64898/2026.07.31.741703v1 2026-08-08-receptor-and-reason Sat, 08 Aug 2026 12:00:00 +0000 676 false Daily roundup for August 8, 2026, ten segments running from a psychedelic through the receptors and circuits shaping mood, motivation, and social pain, into a metabolic angle on Parkinson's, then two computational drug-discovery tools and a light phenomenology close. LEAD — the largest spatially-resolved map of psilocybin-induced excitatory synaptic plasticity: a single dose raises mEPSC frequency 24 hours later, selectively in orbital, prelimbic, insular, and amygdala circuits, tracking local 5-HT2A/2C/1A expression, with a region-confined 5-HT2A knockout proving necessity. The habenula-enriched orphan GPCR GPR151 is nominated as a regionally precise target for inflammation-associated depression. Titi-monkey PET shows nucleus-accumbens kappa-opioid-receptor availability falls during partner separation in both sexes, while plasma oxytocin drops only in males. Baseline mu-opioid-receptor availability predicts emotion-regulation fMRI responses. Mesolimbic dopamine dissociates self-correcting state pessimism from asymmetric-prediction-error trait pessimism. Guanosine-nucleotide metabolism and NDKs power synaptic transmission and implicate GCH1 in Parkinson's, rescued by guanosine. Pro-inflammatory microglia and insular IRF7 drive escalated alcohol intake, blunted by minocycline. A polarized histamine-core, GABA-rim architecture sits inside GABAergic synaptic vesicles. Nesso-1 and ClinOracle are two computational drug-discovery tools. And a language-model read of naturalistic psychedelic, cannabis, meditation, and dream reports closes the show. Paper links in show notes. Daily roundup for August 7, 2026 — nine items with a serotonin thread up front, then ketamine, GLP-1, Alzheimer's/neurodegeneration drug discovery, a computational mass-spec tool, and a light opioid-receptor closer. Housekeeping: the psilocybin synaptic-plasticity map promised yesterday (bioRxiv 2026.07.31.741703) is still abstract-only and stays held. LEAD — Simulated 5-HT2A activation and brain complexity (computational neuropsychopharmacology): a biophysically grounded whole-brain mean-field model incorporating the brain-wide 5-HT2A receptor distribution represents psychedelic action as a reduction in leak potassium conductance; simulations reproduce an asynchronous-irregular state with higher firing, reduced alpha power, and increased spontaneous Lempel-Ziv complexity, while the perturbational complexity index (PCI) does NOT rise — dissociating two complexity metrics as distinct physiology (model; K+ leak is one of several 2A effects). TCB-2 multidimensional phenotyping (5-HT2A agonist, tool compound): in head-fixed (trace conditioning: licking, pupil, eye, blink) and freely moving mice (locomotion, video head-twitch), TCB-2 constricted pupil, reduced high-dose licking without disrupting cue-locked timing, reduced locomotion, and drove dose-dependent head-twitch; the selective 5-HT2A antagonist volinanserin partially attenuated effects — showing head-twitch, autonomic, and learned-behavior channels are dissociable. SonoKet ultrasonic ketamine (precision psychiatry): ketamine-loaded acoustically activatable liposomes uncaged by focused ultrasound for millimeter-scale delivery; solid-phase microextraction coupled to LC-MS/MS sampled ketamine/metabolites, glutamate, GABA, serotonin, and dopamine in real time in prefrontal cortex, nucleus accumbens, and retrosplenial cortex of awake rats; ultrasound alone was inert but ultrasound plus ketamine amplified the neurochemical response — a closed loop between targeted delivery and live neurochemistry. Semaglutide CNS mechanism (GLP-1 repurposing): c-Fos mapping after acute semaglutide activated orexin/hypocretin and basal-forebrain cholinergic systems, and in vivo microdialysis showed acutely increased acetylcholine efflux in the ventral hippocampus in both sexes and in young and aged rats — a concrete circuit hypothesis for GLP-1's cognitive/arousal effects (acute, correlational). AIK3a305 IGF-1 sensitizer (Alzheimer's): a blood-brain-barrier-penetrant small molecule that sensitizes endogenous IGF-1 signaling rather than delivering ligand; in APP/PS1 mice dosed from 12 months (established disease) for 3 months, Y-maze working memory normalized within a month, cognition stayed preserved at 3 months, anxiety-like behavior was corrected, and brain amyloid-beta fell (single model, preclinical). RTX-117 eIF2B activator (neurodegeneration drug discovery): chronic integrated-stress-response (ISR) activation is a shared feature across tRNA-synthetase and other axonal Charcot-Marie-Tooth type 2 subtypes; the CNS-penetrant, Phase 1 molecule RTX-117 activates eIF2B (cryo-EM binding mode to the eIF2B decamer solved), and dosing after symptom onset in a Gars-mutant CMT2D mouse reduced ISR and improved function/electrophysiology, with GDF-15 and FGF-21 as candidate biomarkers; ISR activation recurs across neurodegeneration models including Alzheimer's (AD link correlational). Clusterin microglial brake (Alzheimer's genetics): the LOAD risk gene clusterin is low at rest and induced by inflammation in microglia; deletion amplifies microglial activation while recombinant clusterin cools it, and in a human mutant amyloid knock-in mouse clusterin loss impaired learning/memory despite reduced amyloid, with single-nucleus sequencing showing disrupted excitatory/inhibitory balance — a second cognition-vs-plaque dissociation this week. AIMe neuro-symbolic mass-spec annotation (computational): a multi-agent framework pairing chemical reasoning with structure-informed learning predicts tandem (MS2) spectra by modeling fragmentation as a sequence of likelihood-weighted actions, linking peaks to explicit fragment formulas/structures and outperforming prior methods; predicted MS2 for >100M PubChem molecules into MS2KOSMOS (>800M spectra), expanding searchable space ~3 orders of magnitude, analogous to sequence-homology search (predicted != measured; prospective validation pending). CLOSE — Mu-opioid receptor and photoperiod: seasonal photoperiod cycling versus constant lighting, assessed by ex vivo radioligand binding, markedly reduced inter-individual variability in mu-opioid-receptor density in brain (cerebellum, striatum) and periphery (adrenal), with no tissue growing noisier — implicating daylength dynamics (not just level) in opioid-system set point, and flagging vivarium lighting as a hidden source of between-animal variance. Nine items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.1101/2025.10.20.683366v2 ; https://www.biorxiv.org/content/10.64898/2026.08.01.742216v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741494v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742612v1 ; https://www.biorxiv.org/content/10.64898/2026.08.01.742197v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.743063v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741791v1 ; https://www.biorxiv.org/content/10.64898/2026.08.05.743095v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742441v1 https://www.biorxiv.org/content/10.1101/2025.10.20.683366v2 2026-08-07-receptor-and-reason Fri, 07 Aug 2026 12:00:00 +0000 739 false Daily roundup for August 7, 2026, nine items with a serotonin thread up front, then ketamine, GLP-1, Alzheimer's/neurodegeneration, a computational mass-spec tool, and a light opioid-receptor closer. The psilocybin plasticity map promised yesterday is still abstract-only and stays held. LEAD — a biophysical mean-field model links 5-HT2A activation, modeled as a single reduction in leak potassium conductance, to the high-complexity psychedelic brain state, and dissociates rising spontaneous Lempel-Ziv complexity from an unchanged perturbational complexity index. A multidimensional dissection of the 5-HT2A agonist TCB-2 (volinanserin-sensitive) shows head-twitch, autonomic, and learned-behavior channels are dissociable. SonoKet delivers focused-ultrasound-uncaged ketamine with real-time solid-phase-microextraction neurochemistry. Semaglutide switches on orexin and cholinergic systems and raises ventral-hippocampal acetylcholine in young and aged rats. A BBB-penetrant IGF-1 sensitizer, AIK3a305, rescues cognition and lowers amyloid in established-disease APP/PS1 mice. The clinical-stage eIF2B activator RTX-117 targets a convergent integrated-stress-response node across neurodegeneration, with a solved cryo-EM binding mode. Clusterin acts as a microglial-inflammation brake where cognition tracks inflammation, not plaque. AIMe, a multi-agent neuro-symbolic system, charts the small-molecule universe from tandem mass spectra. And seasonal photoperiod cycling tightens inter-individual mu-opioid-receptor variability. Paper links in show notes. Daily roundup for August 6, 2026, nine items spanning dopamine circuits, psychiatric and Alzheimer's drug discovery, and computational pharmacology. Two of today's picks are pharmacology preprints held over from yesterday because their full text had not yet rendered; both bodies are now up, so they lead. One item stays on the shelf — a mapping of psilocybin-induced excitatory synaptic plasticity in mice (bioRxiv 2026.07.31.741703), posted August 5 and still abstract-only. LEAD — Ophthalmate as a striatal neuromodulator (pharmacology, Parkinson's): building on a prior finding that dopamine-independent levodopa motor benefit tracks an ~8-fold striatal rise in ophthalmate (ophthalmic acid, a glutathione analog with cysteine replaced by aminobutyrate), tritiated-tracer and whole-cell recordings in mouse striatal slices show ophthalmate is taken up by a saturable, glutathione-competing transporter and released in a calcium-dependent, depolarization-evoked manner; it reciprocally regulates dopamine (ophthalmate boosts evoked dopamine release; D2 activation suppresses ophthalmate release), selectively enhances GABA but not glutamate release, and potentiates AMPA-receptor currents and release probability specifically at excitatory synapses onto direct-pathway medium spiny neurons — meeting the working definition of a neuromodulator (ex vivo; in vivo circuit role still inferred). Dorsomedial dopamine and interval timing (fiber photometry, dLight1.3b): over a timed interval dorsomedial-striatal dopamine ramps DOWN before the reward-time increase, and the downward slope predicts the animal's internal estimate of elapsed time; ramping differs by sex and is reshaped by amphetamine — sharpening what "dopamine encodes time" means at the signal level (correlative). Compartmental wiring of nigral dopamine neurons (spatial optogenetics): substantia nigra pars reticulata inhibition targets somas/proximal dendrites; pedunculopontine excitation targets somas/proximal dendrites while subthalamic-nucleus excitation selectively targets the deep pars-reticulata dendrites — so STN and PPN drive dopamine cells via physically separate compartments that can be integrated independently, relevant to STN as a DBS target. SETD1A schizophrenia rescue (conditional knockout mice): loss of the H3K4 methyltransferase SETD1A produces region-specific transcriptomic/neuronal disruption in dorsal striatum and mediodorsal thalamus (not only PFC), each driving selective behavioral deficits; screening six H3K4 demethylase inhibitors, the selective LSD1 (KDM1A) inhibitor TAK-418 restores H3K4 methylation and gene expression and rescues synaptic and schizophrenia-like behavior — an epigenetic-rebalancing proof of concept with a human-safety-tested tool compound (LSD1 is a global regulator; caveats apply). Sob-AM2 in 5xFAD (Alzheimer's drug discovery): the brain-penetrant thyromimetic Sob-AM2 induces microglial TREM2 and, dosed 12 weeks in 7-month-old 5xFAD mice, improves spatial and associative memory and raises synaptophysin and PSD-95 — with no change in amyloid-beta plaque burden or Iba1, and a hippocampus-specific rise in CD68 — separating cognitive benefit from plaque clearance. Recombinant ApoE2 replacement (APOE4 mice): human-cell-produced, heavily sialylated recombinant ApoE2 raises hexokinase-2, lowers endogenous ApoE4, and protects APOE4 neurons from amyloid-beta oligomer toxicity and oxidative stress; paired with a cadherin-derived peptide (ADTC5) that transiently opens the blood-brain barrier, weekly IV dosing in middle-aged/aged APOE4 mice improves learning and memory and boosts synaptosomal glycolytic/exocytotic activity (early; sex-dependent) — protein-replacement framing of a protective genotype plus a reusable delivery trick. Ptarmigan-1 (computational drug discovery): a contrastive model that co-embeds protein residues and small molecules in a shared latent space from sequence and 2D chemistry alone, with no 3D pose, trained on chemoproteomic and bioactivity data; scores a compound in ~10 ms with residue-level attribution, matches docking/co-folding on orthosteric targets and matches or beats them on covalent, cryptic, and disordered sites (including held-out targets), and screens the whole human proteome against 3.4B compounds in under a day — recasting virtual screening as a nearest-neighbor lookup (prospective wet-lab validation still to come). Boltz2ESI (foundation-model enzymology): built on the Boltz biomolecular foundation model, native co-folding captures active-site plasticity to predict enzyme-substrate interactions without predefined pockets, beating sequence-based and rigid-docking baselines and discriminating tight sub-family specificities, demonstrated by prioritizing candidates in the withanolide biosynthetic pathway — useful for pathway de-orphaning and biocatalyst discovery. CLOSE — BrainBench (arXiv): a benchmark for comprehensive, instruction-conditioned EEG understanding across four subsets (foundational analysis, sleep, neurocognitive, physiological integration), 17 datasets, 100K+ executions, grading LLM outputs on numerical, categorical, set, sequence, semantic, and artifact validation rather than plausible prose. Nine items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.08.03.742629v1 ; https://www.biorxiv.org/content/10.64898/2026.08.01.742215v1 ; https://www.biorxiv.org/content/10.64898/2026.08.02.742307v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742543v1 ; https://www.biorxiv.org/content/10.64898/2026.07.31.741838v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741628v1 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741295v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741672v1 ; https://arxiv.org/abs/2608.04156 https://www.biorxiv.org/content/10.64898/2026.08.03.742629v1 2026-08-06-receptor-and-reason Thu, 06 Aug 2026 12:00:00 +0000 716 false Daily roundup for August 6, 2026, nine items from dopamine circuits to a benchmark for language models reading EEG. Two picks are pharmacology preprints held over from yesterday now that their full text has rendered; one item (a psilocybin synaptic-plasticity map) stays held, still abstract-only. LEAD — ophthalmate, a glutathione-analog tripeptide, behaves as a bona fide striatal neuromodulator, with saturable uptake, calcium-dependent release, reciprocal dopamine cross-talk, and direct-pathway-selective synaptic potentiation — a candidate new axis in basal ganglia signaling with Parkinson's relevance. Dorsomedial striatal dopamine ramps DOWN over a timed interval, its slope encoding elapsed-time estimates. Subthalamic vs pedunculopontine excitation is wired onto physically separate dendrites of nigral dopamine neurons. An LSD1/KDM1A inhibitor (TAK-418) rebalances H3K4 methylation and rescues a SETD1A schizophrenia model, which shows dorsal-striatal and mediodorsal-thalamic dysfunction. The TREM2-inducing thyromimetic Sob-AM2 improves cognition in 5xFAD mice without touching amyloid plaque. Recombinant sialylated ApoE2 plus a BBB-opening peptide borrows APOE2's protection for aging APOE4 brains. Ptarmigan-1 recasts virtual screening as a structure-free, pose-free nearest-neighbor query — 3.4B compounds against the whole proteome in a day, strongest at cryptic and covalent sites. Boltz2ESI co-folds enzyme-substrate pairs for pathway de-orphaning. CLOSE — BrainBench benchmarks LLMs on comprehensive EEG understanding. Paper links in show notes. Daily roundup for August 5, 2026, ten items spanning disease-driven excitability, Alzheimer's mechanisms, and a substantial reckoning in computational drug discovery. Note on the mix: the freshest pharmacology preprints (an LSD1/KDM1A-inhibitor rescue of Setd1a schizophrenia phenotypes, a focused-ultrasound ketamine liposome system, a striatal ophthalmate signaling story) were posted in the last ~24h and their full text has not rendered, so they are held for a coming episode rather than summarized off an abstract. LEAD — Traveling beta waves in the parkinsonian STN (Parkinson's): multi-contact DBS-electrode LFP recordings, fit to a spherical traveling-wave model with anti-overfitting controls, show parkinsonian beta (13–35 Hz) organizes into mesoscopic traveling waves in 11 of 20 hemispheres (~16% of recording time), sourced in the dorsal motor STN, with propagation speed positively correlated with residual motor impairment — a spatial feature invisible to single-contact amplitude sensing and relevant to adaptive, sensing-guided DBS. Nav1.3 and early AD hyperexcitability (5xFAD mice): the developmentally-restricted sodium-channel subtype Nav1.3 is aberrantly re-expressed at DG-CA3 mossy fiber terminals; CA3 population activity is elevated at three months, before oligomeric amyloid-beta accumulates (which rises by seven months), and lentiviral shRNA knockdown of Nav1.3 in CA3 normalizes network activity — a subtype-selective, in-principle-druggable driver of early circuit dysfunction. ABCA7 brain proteome (AD risk gene): label-free proteomics across wild-type, Abca7 knockout, APP, and double-mutant mice at four ages converges on insulin-receptor/PI3K-AKT signaling, MAPK, immune/complement, vesicle trafficking, and the ubiquitin-proteasome system, with a membrane-protein bias — reframing ABCA7 loss as sitting at the insulin-signaling/neuroinflammation intersection (associative; whole-brain proteomics blurs cell types). Lifespan architecture of the human EEG: resting-state EEG from 1,763 healthy people aged 5–85 maps spectral, complexity, and morphology features onto strongly non-linear, spatially-heterogeneous trajectories with recurring regional transition sequences — a normative model for flagging age-atypical EEG in neurological/psychiatric disease. Computational drug-discovery reckoning, three papers: Mavchen-1 (pose prediction) shows receptor conformational-ensemble poses beat a matched static AlphaFold2 baseline on 21/29 targets (mean RMSD 3.4 vs 5.6 Å), concentrated in induced-fit and water-mediated classes — but under blind, ground-truth-free pose selection its autonomous scorer only matches/modestly-trails AlphaFold's top poses, naming scoring as the rate-limiter (company preprint, proprietary scorer). "How Bias Shapes the Leaderboard" identifies pocket bias (coordinate-frame leakage from static pocket extraction) and structural data leakage that inflate ML scoring-function benchmarks; under progressively stricter, out-of-distribution evaluation, ML models drop substantially and classical AutoDock Vina outperforms them in 5 of 7 screening tasks — demand OOD numbers. A Multi-Conformation Continuum Electrostatics study of nine kinase domains with eighteen FDA-approved inhibitors finds protein net charge stable but inhibitor charge dynamic upon binding, and tautomer binding-free-energy differences of 3–6 kcal/mol where a solution-minority tautomer dominates the bound state — protonation/tautomer ensembles matter for affinity prediction. An alphabet of allostery: a transferable dictionary of 131M+ local contact "words" carrying Gaussian-network-model transfer-entropy distributions mined from the non-redundant PDB; annotated allosteric sites behave as transfer-entropy sinks (weak but highly significant, ROC-AUC ~0.54), with source/sink channels predicting apo-to-holo rewiring, plus a proposed switch vocabulary for allosteric design. PHI-Reason (LLM in comp bio): keeps heterogeneous biological evidence as modular, named, perturbable records and lets a general-purpose LLM reason over them without task-specific training; species-level top-1 accuracies in the 50s–60s% on phage-host and eukaryotic virus-host prediction, with systematic evidence knockouts and a Jacobian readout quantifying when rationales depart from supplied evidence. CLOSE — Activation-guided neuron intervention (computational psychiatry): in an 8B open LLM, scaling up feed-forward neurons that fire more for Alzheimer's transcripts produces graded impairments in story recall, fluency, working memory, and discourse and reduces lexical surprisal, idea density, and syntactic complexity — paralleling human AD speech; attenuation preserves or slightly improves performance. A controllable, dissectable model of a linguistic phenotype, with the obvious caveat that inducing a phenotype is not modeling the disease. Ten items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.27.740919v1 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741298v1 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741021v1 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741219v1 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740840v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.740774v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.741060v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741458v1 ; https://www.biorxiv.org/content/10.64898/2026.06.10.727770v2 ; https://arxiv.org/abs/2608.03067 https://www.biorxiv.org/content/10.64898/2026.07.27.740919v1 2026-08-05-receptor-and-reason Wed, 05 Aug 2026 12:00:00 +0000 860 false Daily roundup for August 5, 2026, ten items from parkinsonian oscillations to a language model taught to speak like Alzheimer's. (Note: the freshest pharmacology preprints — an LSD1-inhibitor schizophrenia rescue, a focused-ultrasound ketamine system, a striatal ophthalmate story — posted in the last day and had not rendered full text, so they are held for a coming episode.) LEAD — parkinsonian STN beta organizes into mesoscopic traveling waves out of the dorsal motor territory, with propagation speed tracking residual motor impairment, a feature invisible to single-contact sensing and relevant to adaptive DBS. The developmentally-restricted sodium channel Nav1.3 is re-expressed at DG-CA3 mossy fibers and drives CA3 hyperexcitability before amyloid accumulates in 5xFAD mice, rescued by knockdown. An ABCA7 whole-brain proteome points at brain insulin/PI3K-AKT signaling and neuroinflammation. A non-linear, spatially-resolved lifespan atlas of the human EEG from 1,763 people. Three hard looks at ML docking: conformational-ensemble poses beat static AlphaFold2 but blind pose selection stays unsolved; pocket bias and data leakage inflate ML scoring leaderboards, with classical AutoDock Vina winning out-of-distribution; and ensemble protonation/tautomer states shift kinase-inhibitor binding energetics by 3–6 kcal/mol. A transferable transfer-entropy "alphabet of allostery" mined from the PDB. PHI-Reason, an LLM that keeps biological evidence uncompressed and testable for host prediction. CLOSE — activation-guided neuron edits induce Alzheimer's-like language phenotypes in an 8B LLM. Paper links in show notes. Daily roundup for August 4, 2026, ten items spanning channel pharmacology, addiction circuits, computational drug discovery, and a Parkinson's biomarker model. LEAD — GIRK2 channelopathy and drug repurposing (pharmacology): a rare KCNJ6 Gly154Cys variant found in a patient with mild Keppen-Lubinsky features widens the GIRK2 selectivity filter, causing loss of potassium selectivity, aberrant sodium permeation, and loss of inward rectification — a severe gain-of-function, mechanistically a cousin of the weaver mutation one residue over — and an in-silico plus electrophysiological screen of FDA-approved compounds identifies nefazodone (an antidepressant) and eletriptan (a triptan, behaving as a voltage-dependent intracellular pore blocker) as potent blockers of wild-type and mutant channel, offering repurposing starting points. Two papers on punishment-resistant reward-seeking, the diagnostic core of addiction: the first shows chronic estradiol raises punishment resistance in female mice not by boosting dopamine peaks on rewarded actions but by exaggerating dopamine dips on unrewarded actions (optogenetically mimicking the dips accelerates resistance in both sexes), fitting a policy-based view of dopamine; the second shows adolescent chronic unpredictable stress makes a central-amygdala-to-substantia-nigra-pars-lateralis population persistently hyperexcitable, disinhibiting dopamine in the tail of the striatum and newly recruiting it into the behavior — normalizing either node prevents the phenotype. Striatal acetylcholine and reversal learning: dorsomedial-striatal phasic acetylcholine is needed for reversal, best modeled as a signal to create a new latent state when experience is poorly explained; a model-predicted rescue — making reward omission explicit with an auditory cue — reverses the deficit from acetylcholine knockdown. Neurotensin brake on exploration: predator-odor-responsive lateral-septum neurons are GABAergic and neurotensin-expressing; chronic activation drives avoidance without a predator, silencing abolishes predator-enhanced avoidance. ADGRD1/GPR133 (adhesion GPCR): knockout mice show reduced nest-building and exploration, a trend toward impaired LTP, heightened kainate excitotoxicity, and transcriptomic shifts toward reduced synaptic stabilization. Rett astrocytes (MECP2 knockout): reduced nuclear SREBP2, downregulated cholesterol biosynthesis/transport, intracellular cholesterol/desmosterol accumulation, low ABCA1, and defective ApoE lipidation despite preserved ApoE secretion — astrocytes ship ApoE without properly loading cholesterol, starving neuronal synapse maturation. DODock/DOScore (computational drug discovery): a physics-plus-machine-learning docking-and-scoring hybrid built to generalize out of distribution; a blind PCSK9 prediction matched the later crystal pose to 1.2 angstroms, and prospective screens against CD73, IRAK4, Factor XI, and IL-17 yielded chemically novel active inhibitors with up to ~100-fold better hit rate than a recent ML screen (industry group, patents filed — independent replication warranted). AT1R allostery (interpretable ML): a Bayesian-network model over residue interaction energies from MD maps ligand-to-G-protein communication pathways, validated by alanine scanning plus in-silico deep mutational scanning, revealing a cryptic intracellular pocket; structure-based screening against it yields a fragment-like negative allosteric modulator, Q2, that dampens angiotensin-driven Gq signaling. CLOSE — Vocal biomarkers in a songbird: human acoustic analysis of zebra-finch song shows non-linear aging trajectories (middle-aged birds differing from young and old) and, with alpha-synuclein overexpression in song nucleus Area X, larger changes in pitch variability, frequency modulation, and intensity than controls — a tractable model to dissociate aging from Parkinson's-like pathology by voice. Ten items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.28.741201v1 ; https://www.biorxiv.org/content/10.64898/2026.08.01.742224v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741478v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741321v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741073v1 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741204v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741426v1 ; https://www.biorxiv.org/content/10.64898/2026.08.03.742480v1 ; https://www.biorxiv.org/content/10.64898/2026.08.01.742107v1 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740830v1 https://www.biorxiv.org/content/10.64898/2026.07.28.741201v1 2026-08-04-receptor-and-reason Tue, 04 Aug 2026 12:00:00 +0000 667 false Daily roundup for August 4, 2026, ten items from channel pharmacology to a Parkinson's biomarker model. LEAD — a rare KCNJ6 Gly154Cys variant widens the GIRK2 selectivity filter into a sodium-permeant, non-rectifying gain-of-function channelopathy, and an FDA-drug screen surfaces nefazodone and eletriptan as candidate blockers, taking a rare variant from mechanism to repurposing. Two independent papers converge on punishment-resistant reward-seeking: estradiol raises it in female mice by exaggerating dopamine dips on unrewarded actions, and adolescent stress recruits a latent central-amygdala-to-nigra-to-tail-of-striatum dopamine circuit. Striatal acetylcholine signals when to build a new latent state during reversal learning, with a model-predicted behavioral rescue. A neurotensin population in the lateral septum brakes exploration under chronic threat. The orphan adhesion GPCR GPR133 earns a hippocampal-vulnerability phenotype. Rett astrocytes ship ApoE without loading cholesterol, starving neuronal synapse maturation. A physics-plus-ML docking hybrid, DODock/DOScore, posts prospective virtual-screening wins (with the usual industry caveats). An interpretable Bayesian-network allostery map of AT1R yields a fragment negative allosteric modulator. CLOSE — zebra-finch song separates normal aging from alpha-synuclein pathology. Paper links in show notes. Daily roundup for August 3, 2026, nine items spanning structural pharmacology, Parkinson's, plasticity, neuroimmune therapeutics, and agentic AI for neuroscience and drug discovery. LEAD — Graded activation of the dopamine D1 receptor (structural pharmacology): cyclic-AMP assays reclassify two canonical "neutral antagonist" tool compounds — LE300 is an inverse agonist suppressing constitutive D1R activity, SCH-23390 is a weak partial agonist (~26% of dopamine's Emax), and several antipsychotics filed as D1 antagonists are inverse agonists too; cryo-EM using these ligands captures a missing inactive state (LE300) and an intermediate active state (SCH-23390) that bridge known active conformations, and with MD and mutagenesis shows efficacy is encoded as progressive engagement of a conserved tryptophan toggle switch; a single chemical modification converts SCH-23390 into a high-efficacy partial agonist, giving a design framework relevant to D1 partial agonists like tavapadon. Cortical remodeling in parkinsonism (MitoPark mice): as striatal dopamine falls, layer-5 pyramidal-tract motor-cortex neurons lose alpha-5 GABA-A-mediated inhibition from somatostatin interneurons and gain excessive dendritic NMDAR activation plus spine loss; L-DOPA prevents (early) or rescues (late) these adaptations, casting the cortex as an active, dopamine-sensitive participant and the alpha-5 GABA-A receptor as a handle. AMPA phosphorylation and LTP maturation (rat hippocampus, weaning to adulthood): total GluA1/GluA2 rise but basal GluA1 phosphorylation at Ser831/Ser845 falls, priming those sites for activity-dependent phosphorylation as theta-burst LTP in CA1 progressively strengthens; NMDA-subunit and CaMKII phosphorylation increase. Minocycline against panic (rats): hypoxia (7% O2) drives panic-like jumping and immunoresponsive microglial remodeling in the periaqueductal grey; five days of minocycline reduces jumping as well as acute alprazolam and prevents the dorsomedial-PAG microglial changes — a non-benzodiazepine repurposing angle. SeekBrain (arXiv): an autonomous multi-agent neuroscience-discovery framework grounded in analysis "recipes" mined from code-paper pairs, beating Claude Code and Codex by ~12% and ~18% on the BrainArena benchmark and doing real work on zebrafish and mouse decision-making datasets. CyberNeuro (arXiv): a privacy-preserving agentic workbench (Planner/Validator/Dispatcher/Reporter over a secure MCP bridge) driving a local model, WandaMind, lifting held-out NeuroBench accuracy 40%→69% while using a fraction of cloud token budgets and keeping clinical data in-house. MGMG (bioRxiv): morphology-guided molecule generation conditions on cellular imaging profiles plus text to design bioactive molecules with no target information, extending to genetic perturbations; validated by in-silico docking. ADMET quantum-inspired preprocessing (bioRxiv): descriptor correlations encoded as a many-body Hamiltonian whose simulated expectation values feed a CatBoost model, reaching ~0.75 log MAE on aqueous solubility — within existing error bars, classically simulated, marginal. CLOSE — Dynamic connectivity in smartphone addiction (resting-state fMRI): attractor-state and quasi-periodic-pattern analyses show reduced dorsal-attention-network participation and context-dependent DMN-attention differences in high scorers, most during internal-external state transitions. Nine items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.28.741251v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.741031v1 ; https://www.biorxiv.org/content/10.64898/2026.07.30.741885v1 ; https://www.biorxiv.org/content/10.64898/2026.07.29.741424v1 ; https://arxiv.org/abs/2607.29347 ; https://arxiv.org/abs/2607.28841 ; https://www.biorxiv.org/content/10.1101/2025.07.11.664424v2 ; https://www.biorxiv.org/content/10.64898/2026.06.30.735582v2 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740792v1 https://www.biorxiv.org/content/10.64898/2026.07.28.741251v1 2026-08-03-receptor-and-reason Mon, 03 Aug 2026 12:00:00 +0000 611 false Daily roundup for August 3, 2026, nine items from structural pharmacology to agentic AI. LEAD — two canonical "neutral antagonist" D1 tool compounds turn out to have intrinsic efficacy (LE300 an inverse agonist, SCH-23390 a weak partial agonist), and cryo-EM of the inactive and intermediate states they trap maps how ligand efficacy is encoded on a conserved tryptophan toggle switch, with a single tweak converting SCH-23390 into a high-efficacy partial agonist. The parkinsonian motor cortex remodels through lost alpha-5 GABA-A inhibition and excess NMDAR drive, rescuable with L-DOPA. Falling basal AMPA GluA1 phosphorylation paces the maturation of adolescent hippocampal LTP. Minocycline curbs hypoxia-triggered panic in rats as well as alprazolam by blocking microglial remodeling in the periaqueductal grey. SeekBrain, a recipe-grounded multi-agent system, beats Claude Code and Codex on a neuroscience-analysis benchmark; CyberNeuro runs a privacy-preserving local-LLM agent workbench for cohort neuroimaging. MGMG generates molecules from cell-morphology profiles with no target information. A quantum-inspired feature trick marginally nudges ADMET solubility prediction. CLOSE — dynamic connectivity reveals attention-network shifts in high smartphone-addiction scorers. Paper links in show notes. Daily roundup for August 2, 2026, ten items running computational systems neuroscience into plasticity, sensory gain, neuroimmune therapeutics, a hydrocephalus drug target, and a sleep-and-depression study. LEAD — Climbing fibers as a loss-function gradient (Shadmehr lab, marmosets): reward-valued saccade targets show climbing fibers multiplicatively scale the spatial error vector by reward context (folding in reward prediction error), and spike-triggered suppression measuring each Purkinje cell's "potent vector" reveals the climbing-fiber signal is, on average, proportional to the dot product of the reward-weighted error onto that cell's potent vector — i.e. the teacher hands each cell the gradient of a loss with respect to its own output, the exact quantity gradient descent needs. Coupled neuron–synapse working memory (Abbott lab, theory): recurrent nets with ongoing Hebbian plasticity fluctuating on the neural timescale keep oscillating after an oscillatory input is removed — persistent activity with no explicit storage/retrieval and no prior knowledge of the input — traced analytically to plasticity creating outlier eigenvalues, and exploited to build a dynamic analog of a Hopfield network. Distance to criticality as cortical generator (resting-state fMRI): one parameter, each local network's distance to criticality, fits both local power spectra and functional connectivity across all 360 cortical areas, with a conserved cross-subject rank order and a universal power-spectrum shape-collapse — a single organizing principle for cortical dynamical diversity, though fit to slow BOLD. Astrocytes instructively set neuronal translation (Klann lab): astrocyte-conditioned medium raises neuronal translation (further with BDNF) but neurotoxic reactive-astrocyte medium suppresses it; neuronal mTORC1 tracks output; astrocyte-secreted APOE and its lipid cargo is a negative regulator, the effect requires neuronal endocytosis and drives glutamatergic synaptic remodeling — an APOE-linked route from astrocyte state to the protein synthesis memory depends on, rescuable at several pathway points. Aggrecan, not perineuronal nets, closes the critical period: deleting aggrecan in inhibitory neurons removes perineuronal nets yet the visual critical period still closes; deleting it in excitatory or all neurons sustains adult plasticity — reassigning the brake to diffuse neuropil aggrecan and implying the classic chondroitinase reopening worked through the matrix, not the nets. Insulin gates smell at the first synapse: satiety-driven insulin inhibits the Kv1.3 current in periglomerular cells, raising their activity and presynaptic inhibition of olfactory-receptor-neuron terminals, dampening input before mitral cells — metabolic-state sensory gain control at a druggable channel. Circadian reconfiguration of the visual code (dLGN, awake mice): spontaneous activity peaks in late subjective day; tuning stays stable but information-transfer efficiency depends on scene statistics, favoring efficient coding at night under high contrast, confirmed by optogenetic retinogeniculate probing. gp130 neuroprotective loop: activating microglial gp130 (the shared IL-6-family receptor) triggers LIF secretion that drives neuronal IL-6 back onto microglial gp130, a protective circuit improving TBI, stroke and spinal-cord-injury outcomes and drivable with designer cytokines — inverting the pro-inflammatory reputation of IL-6/gp130. TRPV4–SPAK axis for hydrocephalus: the kinase SPAK controls choroid-plexus CSF secretion via NKCC1 (the bumetanide target) and the Na/K-ATPase; the SPAK inhibitor ZT-1a systemically reduces post-hemorrhagic fluid buildup at 24h, with TRPV4-driven hypersecretion requiring SPAK — a rare pharmacological alternative to shunting, albeit an acute rodent readout. CLOSE — Sleep and self-evaluation in depression: targeted reactivation of positive traits during NREM sleep boosts positive self-evaluation, evokes trait-specific sigma-band (~11–16 Hz) representations that align with waking self-evaluation, and both the benefit and the neural alignment weaken as depressive symptoms rise — a self-evaluative process spanning sleep and wake that is specifically disrupted in depression. Ten items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.27.741034v1 ; https://www.biorxiv.org/content/10.1101/2025.08.20.671131v1 ; https://www.biorxiv.org/content/10.64898/2026.05.21.726898v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.741020v1 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740801v1 ; https://www.biorxiv.org/content/10.64898/2026.07.22.740090v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740496v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740442v1 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740775v1 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740831v1 https://www.biorxiv.org/content/10.64898/2026.07.27.741034v1 2026-08-02-receptor-and-reason Sun, 02 Aug 2026 12:00:00 +0000 638 false Daily roundup for August 2, 2026, ten items from computational systems neuroscience to psychiatry. LEAD — in reward-valued marmoset saccades, cerebellar climbing fibers hand each Purkinje cell the gradient of a reward-weighted loss with respect to that cell's own motor output, making credit assignment look like machine-learning gradient descent. Coupling neuron and synapse dynamics yields a storage-free "persistent oscillation" form of working memory. A single distance-to-criticality gradient reproduces cortical power spectra and connectivity across 360 areas. Astrocyte-secreted APOE clamps down neuronal mTORC1-driven translation in the reactive, Alzheimer's-linked state. Aggrecan diffusely in the neuropil, not perineuronal nets, is the real brake closing the visual critical period. Insulin inhibits Kv1.3 in periglomerular cells to turn down smell after a meal. Circadian phase reconfigures the visual-thalamus code without changing tuning. A gp130 loop makes microglia neuroprotective across TBI, stroke and spinal-cord injury. The SPAK inhibitor ZT-1a curbs post-hemorrhagic hydrocephalus through the TRPV4–NKCC1 axis. CLOSE — sleep reactivation of positive traits boosts self-evaluation but fails as depressive symptoms rise. Paper links in show notes. Daily roundup for August 1, 2026, ten items running from reward circuits through receptor pharmacology to computational drug discovery and neurodegeneration. LEAD — Cerebellar encoding of dopamine reward (NINDS intramural): head-fixed mice pushed for reward delivered as optogenetic dopamine-neuron stimulation, removing consummatory movement; two-photon imaging showed granule cells predictively ramping to the expected reward time (rescaling between 1- and 2-second delays) and climbing fibers spiking just after delivery like an instructive teaching signal, with dopamine encoding matching or exceeding water; chronic granule-cell inhibition disrupted self-stimulation learning and climbing-fiber stimulation alone drove operant learning, arguing the cerebellum is deeply wired into brain-wide reward prediction. Accumbens dopamine (rats, GRAB-DA): NAc-core dopamine during lever-pressing for a peer, palatable food, or shock avoidance showed cue-locked phasic bumps and pre-press ramps scaled by what else was available (food dopamine roughly doubled when food was the only reinforcer), so accumbens dopamine is not an absolute value signal. Chronic fentanyl abstinence (female mice): an oral drinking-in-the-dark model with naloxone-precipitated withdrawal produced a dynamic, week-to-week behavioral course — transient mid-exposure thermal hyperalgesia, shifting affective and exploratory measures across a month of abstinence — a female-specific longitudinal map of relapse-relevant windows. GPR183 biased agonists: optimization from an antagonist scaffold yielded full, completely Gi-biased agonists at low-nanomolar potency; the lead TUG-2604 matches the natural oxysterol yet does not itself drive human dendritic-cell migration and blocks oxysterol-induced migration, dissociating G-protein signaling from chemotaxis. Pyroglutamylation and GPCR selectivity: a solvent-exposed N-terminal pyroglutamate never contacts the receptor but reshapes the peptide's conformational ensemble, flipping activation of paired sea-slug PRXamide receptors in opposite directions and biasing human neuromedin U subtype preference, with pocket tightness setting the direction — distal allosteric selectivity encoded in the ligand. Rhodopsin deep mutational scanning: a three-phenotype cell assay (basal activity, ligand-dependent signaling, receptor abundance) scored roughly 2,000 substitutions (about 6,000 measurements), revealing asymmetric biophysical pleiotropy that concentrates constraint at the retinal pocket and G-protein interface. AiPP (AI protein profiling): a sequence-based multitask model on the ESMC protein language model predicts covalently ligandable cysteines plus reversible-binding and disordered-recognition annotations, using a LatentLift clustering step to reconcile conflicting experimental labels, yielding a proteome-wide, structure-free cysteine ligandability atlas for target ID and covalent drug discovery. EC-Reason-Bench: a training-free diagnostic decomposes LLM enzyme-classification failure into output structure, external knowledge, reasoning structure and robustness, finding that near-zero four-level EC-number accuracy is a knowledge-grounding problem recoverable at inference time without weight updates — the bottleneck is retrieval, not reasoning. LINE-1 and neuronal senescence in Alzheimer's: in donor-aged induced neurons, LINE-1 retrotransposon activity drives senescence and the SASP, and nucleoside reverse-transcriptase inhibitors or antisense oligonucleotides lower p16, quiet interferon programs and reduce reactive astrogliosis; long-read single-cell sequencing localizes a LINE-1-high neuron subset enriched in AD and RNA velocity suggests LINE-1 activation precedes senescence — a repurposing hypothesis for brain-penetrant approved drugs. CLOSE — AQP4 stop-codon read-through: the perivascular water channel is the extended isoform AQP4X made by UGA read-through, and the purine analog 2,6-diaminopurine boosts read-through and raises AQP4X in cells and by about 10% after one intracranial injection in vivo, with no effect in a full read-through mouse confirming mechanism — a proof of concept for pharmacologically tuning a disease-relevant isoform. Ten items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.02.02.703324v4 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740421v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.741002v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.740726v1 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741146v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.740619v1 ; https://www.biorxiv.org/content/10.1101/2025.09.07.670677v3 ; https://arxiv.org/abs/2607.26397 ; https://www.biorxiv.org/content/10.64898/2026.07.27.740588v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.741080v1 https://www.biorxiv.org/content/10.64898/2026.02.02.703324v4 2026-08-01-receptor-and-reason Sat, 01 Aug 2026 12:00:00 +0000 749 false Daily roundup for August 1, 2026, ten items from reward circuits to computational drug discovery. LEAD — mice pushing for dopamine delivered straight into the brain reveal cerebellar granule cells that predictively time the reward and climbing fibers that instructively teach it, and causal manipulations argue the cerebellum belongs in brain-wide reward-prediction networks, not just motor control. Accumbens dopamine scales reward by what else is on offer, so it is not an absolute value signal. A female-specific longitudinal map shows fentanyl abstinence behavior shifts week to week. TUG-2604 is a fully Gi-biased GPR183 agonist that blocks dendritic-cell migration, dissociating signaling from chemotaxis. N-terminal pyroglutamylation steers GPCR subtype selectivity allosterically without touching the receptor. Deep mutational scanning of rhodopsin exposes biophysical pleiotropy at the retinal pocket and G-protein interface. AiPP, an ESMC-based sequence model, maps covalent cysteine ligandability across the whole proteome. EC-Reason-Bench shows LLMs fail four-level enzyme classification for lack of knowledge, not reasoning. LINE-1 retrotransposons drive neuronal senescence in Alzheimer's and reverse-transcriptase inhibitors push back. CLOSE — 2,6-diaminopurine tunes perivascular AQP4 by promoting stop-codon read-through. Paper links in show notes. Daily roundup for July 31, 2026, nine items with a recurring throughline: not just does a system run, but can it show its work. LEAD — Evaluating agentic bioinformatics (FEV framework): surveying 109 agentic or agent-adjacent systems and 28 benchmarks across genomics, single-cell/spatial omics, protein science, drug discovery and computational pathology, the authors argue LLM-agent credibility should rest on the inspectable workflow trajectory (Function, Evidence, Validation), not final-answer correctness; the field's planning and tool execution have outrun replayability, provenance, external validation and prospective testing. Chst9 accumbens MSNs and opioid reward (Day lab, UAB): a rare, conserved, high-mu-opioid-receptor indirect-pathway MSN subtype marked by Chst9 is silenced by opioids, and selective deletion of the mu-opioid receptor gene from this population abolishes fentanyl conditioned place preference, resolving why opioid reward never mapped onto the canonical D1/D2 split (caveat: preprint, target framework not yet drugged). BRD3290 (Broad): a GluA3-preferring AMPA-receptor positive allosteric modulator built to match the GRIA3 schizophrenia-risk genetics engaged the established AMPA-PAM target-engagement biomarker in wild-type mice but did not improve novel object recognition, whereas an older nonselective AMPA-PAM did — a candid negative on subtype-selective AMPA modulation. Metabotropic glutamate control of trans-thalamic communication (Xiao-Jing Wang, NYU): a spiking model gives seconds-slow group-I mGluRs a role — thalamocortical mGluRs boost dendritic plateau potentials and feedforward detection, corticothalamic mGluRs integrate inputs and shift higher-order thalamic firing from burst to tonic (a novelty wake-up call), supported by pulvinar recordings. PEP-3 LRRK2-PP1 peptide (Inserm): a bifunctional, blood-brain-barrier-crossing peptide disrupts the LRRK2-PP1 interface rather than the kinase active site; in alpha-synuclein-overexpressing mice, repeated dosing was well tolerated and reduced substantia-nigra dopaminergic death and pathological phospho-alpha-synuclein (caveat: small in vivo package, peptide delivery). Boldine and SARM1 (Penn): the boldo aporphine alkaloid boldine directly inhibits SARM1 NADase activity (IC50 about 7.5 micromolar), with AlphaFold-3-style modeling plus ML docking predicting dual catalytic-plus-regulatory-site binding, and preserves severed sciatic-nerve axon segments to 7 days (caveat: micromolar potency, explant model). Fungal autophagy modulators for tauopathies (Univ. Maimonides): an in-silico network-pharmacology/molecular-dynamics screen nominates PPAR-gamma, GSK-3-beta and casein kinase 2 as autophagy-tau targets and Lion's Mane/ergot metabolites as candidate binders, with unusually disciplined applicability-domain and proteomic caveats (only casein kinase 2 altered, in AD CA3). MADRS LLM pipeline (Clario/Thermo Fisher): a pipeline transcribes structured depression-trial interviews, maps them to the ten MADRS items, estimates severity and flags problematic ratings, reaching about 0.87 correlation with expert raters — a quality-control layer for the measurement that decides placebo separation. CLOSE — Psilocybin behavioral factors (Nautiyal, Dartmouth): across nearly 700 mice, the acute head-twitch/locomotor response is robust while post-acute antidepressant-like effects are variable and structured by sex, stress protocol and 5-HT1B signaling (sturdiest signal in sucrose preference), reframing the reproducibility problem as identifiable variance. Nine items per receptor-and-reason PROMPT.md. Paper links: https://arxiv.org/abs/2607.27556 ; https://www.biorxiv.org/content/10.64898/2026.07.28.741269v1 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740780v1 ; https://www.biorxiv.org/content/10.64898/2026.07.25.740691v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.740889v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.741059v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740367v1 ; https://arxiv.org/abs/2607.28190 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740586v1 https://arxiv.org/abs/2607.27556 2026-07-31-receptor-and-reason Fri, 31 Jul 2026 12:00:00 +0000 723 false Daily roundup for July 31, 2026, nine items on one question — not just does it run, but can it show its work. LEAD — a Function-Evidence-Validation review of 109 agentic bioinformatics systems argues these LLM agents should be judged on inspectable workflow traces, not final answers, because planning has outrun provenance and validation. A conserved Chst9-marked accumbens MSN subtype resolves the opioid-reward paradox, and deleting its mu-opioid receptor abolishes fentanyl place preference. A GluA3-preferring AMPA-receptor PAM built to match schizophrenia genetics engaged its biomarker but failed to improve cognition — a candid negative. Slow group-I mGluRs give the higher-order thalamus a burst-to-tonic dial on cortical communication. A blood-brain-barrier-crossing peptide disrupting the LRRK2-PP1 interface cuts dopaminergic death and phospho-alpha-synuclein in mice. Boldine is a dual-site SARM1 NADase inhibitor that preserves severed axons. A network-pharmacology screen nominates PPAR-gamma, GSK-3-beta and casein kinase 2 as fungal-metabolite autophagy targets for tauopathies. An LLM pipeline scores depression-trial MADRS interviews at about 0.87 with experts. CLOSE — a 693-mouse dataset shows psilocybin's post-acute effects are structured by sex, stress protocol and 5-HT1B, not random noise. Paper links in show notes. Daily roundup for July 30, 2026, ten items; a glia-and-neuroinflammation arc that turns through drug repurposing, the neuromodulation of stress and addiction, plasticity and memory, and the computational edge of psychiatry. Two receptor-structure papers (dopamine D1 graded activation; a GluA3-preferring AMPA-receptor PAM for schizophrenia) were deferred because their full text had not yet rendered. LEAD — TRPML1 in Alzheimer's astrocytes: using an astrocyte-restricted, lysosome-targeted calcium sensor imaged in the living brain, loss of the lysosomal calcium channel TRPML1 is shown to be a pre-plaque initiating lesion; in 3-month amyloid knock-in mice lysosomal calcium transients are nearly abolished (amplitude ~3-fold down, frequency ~5-fold down) tracking reduced astrocytic TRPML1, and astrocyte-only TRPML1 restoration from 2 months holds 8-month astrocytes at wild-type calcium and cuts plaque area in the transduced region from ~3.9 percent to ~0.2 percent (caveat: overexpression that raises calcium even in wild-type astrocytes, and local-only amyloid effect). SASP-driven senescence spread: conditioned-media transfer across five human brain cell types builds a directional matrix — astrocytes and microglia are spreaders (self, each other, endothelial cells), oligodendrocytes and neurons send but cannot receive — and a cytokine panel plus receptor inference plus target-specific drugs show astrocyte-to-microglia spread is blocked by a DPP-4 inhibitor (sitagliptin target) or a CXCR7 blocker (caveat: immortalized cancer-derived cell lines, single contested senescence marker). Approved-drug NLRP3 screen: a reporter separating priming from assembly flags the antidepressants fluoxetine and duloxetine as inflammasome blockers — fluoxetine acts at priming by reducing endotoxin membrane binding — with sharply sex-dependent survival in a lethal endotoxin model (fluoxetine rescues females, mefloquine males), reframing SSRIs as innate-immune modulators (caveat: prophylactic pretreatment, pure endotoxin, no mechanism for the sex split). Circadian amplitude screen: 5,631 repurposing compounds through a Bmal1-luciferase reporter yield 13 confirmed amplitude enhancers, 8 hitting clock-naive targets (JAK3, PDE-1, ryanodine receptor via dantrolene, among others), confirmed by structurally distinct inhibitors and shown to be cell-autonomous in the SCN (caveat: some dose-independent target attributions, in vitro/ex vivo only). Early-life-stress resilience: per-animal multivariate phenotyping plus unsupervised clustering (null on group averages) shows the identical caudal-dorsal locus coeruleus noradrenergic signature maps onto anxiety in males but resilience in females, with cluster proportions mirroring the human 2:1 female depression risk (caveat: single correlational c-Fos snapshot, no causal manipulation). Alcohol compulsivity: unpredictable intermittent ethanol access, not continuous access, produces durable aversion-resistant (quinine-adulterated) drinking after vapor dependence, argued to be cue-driven rather than pharmacokinetic, with a lick-by-lick within-session blood-alcohol estimate (caveat: sexes pooled, underpowered). Rett memory: 7-day tracking of the same CA1 place cells shows a failure to stabilize, not form, place fields — excess, mis-targeted behavioral-timescale synaptic plasticity, with a minimal model reproducing the phenotype by raising non-specific spontaneous activity (caveat: causal dendritic-inhibition driver inferred, not manipulated). Generative replay: during sleep ripples, primary visual cortex co-activates a never-experienced inferred cue pair for the reward set, with hippocampal content predicting cortical content ~80 ms earlier and cortex decoupling across days — systems consolidation as the hippocampus teaching cortex a generative model (caveat: small-n, correlational). Depression-TMS targeting: connectivity-based parcellation of the left DLPFC in ~391 HCP brains shows only one anterior-central subcluster maximizes subgenual anti-correlation (a neighboring cluster is positively coupled), yielding a probabilistic target map (caveat: young healthy data, no TMS, no outcomes). CLOSE — EEG foundation models for clinical decoding fail targeted negative controls (high accuracy on shuffled labels/corrupted inputs) and degrade across datasets, revealing reliance on dataset identity rather than neurophysiology. Ten items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.23.740052v1 ; https://www.biorxiv.org/content/10.64898/2026.02.10.705129v1 ; https://www.biorxiv.org/content/10.64898/2026.03.05.709979v1 ; https://www.biorxiv.org/content/10.64898/2026.07.23.740283v1 ; https://www.biorxiv.org/content/10.1101/2025.10.11.681820v1 ; https://www.biorxiv.org/content/10.64898/2026.03.31.715677v1 ; https://www.biorxiv.org/content/10.64898/2026.01.29.702595v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740539v1 ; https://www.biorxiv.org/content/10.1101/2025.11.19.689337v1 ; https://arxiv.org/abs/2607.24519 https://www.biorxiv.org/content/10.64898/2026.07.23.740052v1 2026-07-30-receptor-and-reason Thu, 30 Jul 2026 12:00:00 +0000 739 false Daily roundup for July 30, 2026, ten items; a glia-and-neuroinflammation arc through drug repurposing, the neuromodulation of stress and addiction, plasticity and memory, and computational psychiatry (two receptor-structure papers deferred for unrendered full text). LEAD — the lysosomal calcium channel TRPML1 is a pre-plaque initiating lesion in Alzheimer's astrocytes; astrocyte-only TRPML1 restoration normalizes lysosomal calcium and cuts local plaque area from about 3.9 to 0.2 percent. A directional map of SASP-driven senescence spread across human brain cell types makes astrocytes and microglia the spreaders and nominates DPP-4 and CXCR7 as cell-type-specific blocks. An approved-drug screen flags fluoxetine and duloxetine as sex-dependent NLRP3 inflammasome blockers, reframing SSRIs as innate-immune modulators. A phenotypic screen finds novel small-molecule enhancers of circadian amplitude acting cell-autonomously in the SCN. Early-life-stress resilience is an active, sex-specific locus coeruleus state — the same caudal-dorsal signature means anxiety in males but resilience in females. Unpredictable, not merely intermittent, alcohol access drives aversion-resistant compulsive drinking. Rett memory failure is unconstrained, mis-targeted behavioral-timescale plasticity, reproduced by a single spontaneous-activity knob. Generative replay in visual cortex during sleep ripples, taught by the hippocampus, builds a never-experienced inference. Connectivity parcellation refines the depression-TMS prefrontal target beyond the subgenual-anticorrelation heuristic. CLOSE — EEG foundation models fail dataset-identity and negative-control stress tests for clinical decoding. Paper links in show notes. Daily roundup for July 29, 2026, ten items; a mechanism-to-tools arc through Alzheimer's proteostasis, Parkinson's physiology, the neuromodulators of learning and stress, computational psychiatry, and agentic analysis. LEAD — APOE4 in isogenic human iPSC-derived neurons: layering transcriptomics, ribosome profiling (translatomics), and proteomics, the transcriptome barely moves while APOE4 disrupts ribosome occupancy, uncouples protein abundance from transcript level, and globally extends protein half-lives; the proteasome and lysosome both lose ground (APOE binds the neuronal proteasome more), long-lived proteins accumulate during maturation, and the cells slide toward a senescence-like state — a neuron-intrinsic proteostasis failure upstream of amyloid, and a complementary resilience target to anti-amyloid antibodies (which carry ARIA risk in APOE4 carriers). Parkinson's physiomarkers: OFF-medication subthalamic LFPs plus quantitative MRI in 33 patients show distinct pathologies map onto dissociable electrophysiological components — nigral susceptibility (iron) with low-beta bursts, nigral free water with the aperiodic offset/slope, and pedunculopontine (brainstem cholinergic) with high-frequency aperiodic activity strengthening as nigral pathology worsens — so the subthalamic signal is an integrated multi-system fingerprint, not a unitary dopamine readout, a caution for closed-loop DBS tuning. Levodopa-induced dyskinesia: FosTRAP ensemble tagging in the 6-OHDA mouse identifies opposing afferents — secondary motor cortex drives dyskinesia, parafascicular thalamus brakes it; chronic levodopa biases cortical input toward NMDA-receptor excitation onto ensemble neurons while thalamic input recruits polysynaptic inhibition, and ensemble-restricted GluN1 knockdown reduces peak dyskinesia, framing LID as a targetable cortical-vs-thalamostriatal imbalance. Dopamine-modulated STDP: the standard reward-learning model has dopamine merely scale STDP (so it learns at any dose); the authors instead make rising dopamine nonlinearly bias plasticity toward potentiation with receptor saturation, and in networks of Izhikevich neurons robust conditioning emerges only within a realistic dopamine range, self-organizing a feedforward backbone in recurrent circuitry that bursts to reward-predicting cues. Locus coeruleus and cognitive effort: reversible inhibitory-DREADD silencing of noradrenergic LC neurons in rhesus macaques (deschloroclozapine actuator; one control animal) left an easy visible-reward hole-board task intact but impaired the working-memory (hidden-reward) version — a rare causal, reversible primate loss-of-function implicating LC in cognitive, not just physical, effort. Stress-specific noradrenaline: a genetically encoded fluorescent noradrenaline sensor in the hypothalamic PVN of freely moving mice shows stressor-specific release — pulsatile seconds-scale release to acute nociception versus prolonged release tracking restraint — so the noradrenergic input encodes the kind and time-course of a stressor, not a generic on-signal. Schizophrenia thought disorder: resting-state MEG in 25 patients finds no beta-power group difference, but transient beta-burst rate/duration/synchrony show dissociable anterior-posterior patterns, with shorter parietal bursts and increased prefrontal synchrony in the default mode network predicting computationally derived speech-disorganization — a reminder that time-averaged spectra can hide burst-level clinical signal. CogEEGAgent: an LLM agent for cognitive-EEG analysis grounded in MNE-Python that separates semantic authority (intent, proposing registered analyses) from deterministic scientific authority (typed contracts, held-out confirmation data, pre-commitment before confirmation is opened, evidence-bound release), engineering out agentic researcher-degrees-of-freedom; it out-routes a matched deterministic router and its held-out confirmation curbs false positives from adaptive search — a fail-closed template for autonomous scientific agents. CDRPipe: a connectivity-based drug-repurposing pipeline harmonizing the microarray Connectivity Map (~2,000 experiments) with pseudo-bulk profiles from the Tahoe single-cell perturbation atlas (tens of thousands); single-cell-derived signatures recover far more annotated therapeutics, and the two resources overlap by only a few percent — complementary, so single-database repurposing quietly narrows the search space. CLOSE — chronic diphenhydramine (first-generation antihistamine, anticholinergic OTC sleep aid) in 5XFAD and wild-type mice by telemetry: it did not rescue AD-model sleep deficits and instead fragmented NREM2 and pushed drowsiness into the active phase in both genotypes — a cautionary data point given anticholinergic use is itself linked to dementia risk. Ten items per receptor-and-reason PROMPT.md; the psilocybin behavioral-factors preprint was deferred (body not yet rendered). Paper links: https://www.biorxiv.org/content/10.64898/2026.07.15.738801v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739149v1 ; https://www.biorxiv.org/content/10.64898/2026.07.22.740223v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739276v1 ; https://www.biorxiv.org/content/10.64898/2026.07.23.740243v1 ; https://www.biorxiv.org/content/10.64898/2026.07.25.740743v1 ; https://www.biorxiv.org/content/10.64898/2026.07.27.740959v1 ; https://arxiv.org/abs/2607.25045 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739227v1 ; https://www.biorxiv.org/content/10.64898/2026.07.22.739929v1 https://www.biorxiv.org/content/10.64898/2026.07.15.738801v1 2026-07-29-receptor-and-reason Wed, 29 Jul 2026 12:00:00 +0000 609 false Daily roundup for July 29, 2026, ten items; a mechanism-to-tools arc through Alzheimer's proteostasis, Parkinson's physiology, the neuromodulators of learning and stress, computational psychiatry, and agentic analysis. LEAD — in isogenic human iPSC-derived neurons, APOE4 barely changes the transcriptome but arrests proteome renewal: it disrupts ribosome occupancy, uncouples protein levels from transcripts, globally extends protein half-lives, and impairs the proteasome and lysosome, so long-lived proteins accumulate and neurons drift toward senescence — a neuron-intrinsic Alzheimer's mechanism upstream of amyloid. Distinct nigral and brainstem pathologies map onto separable subthalamic LFP components in Parkinson's, so the adaptive-DBS physiomarker is not a unitary dopamine readout. A parafascicular thalamostriatal brake opposes motor-cortex recruitment of striatal ensembles in levodopa-induced dyskinesia, and ensemble-restricted GluN1 knockdown cuts peak dyskinesia. A dopamine-modulated STDP rule that flips depression to potentiation and saturates only supports learning within a realistic dopamine range in spiking networks. Reversible DREADD silencing of the locus coeruleus in macaques impairs an effortful working-memory task but not an easy one — causal evidence for the LC in cognitive effort. Fiber-photometry shows stressor-specific noradrenaline release in the hypothalamic PVN. Beta-burst dynamics, not spectral power, track thought disorder in schizophrenia MEG. CogEEGAgent separates semantic from scientific authority and holds out confirmation data to curb agentic confirmation bias in EEG analysis. CDRPipe shows single-cell perturbation signatures recover more repurposing candidates than microarray Connectivity Map and barely overlap. CLOSE — chronic diphenhydramine fragments NREM sleep and worsens sleep in 5XFAD and wild-type mice, questioning OTC anticholinergic sleep aids. Paper links in show notes. Daily roundup for July 28, 2026, nine items; a glia-and-metabolism arc that turns through novel psychiatric targets, computational drug discovery, and neurodegeneration. LEAD — miR-223, a glia-enriched microRNA that runs abnormal in Alzheimer's patient plasma/CSF: in the App knock-in amyloidosis mouse, both a ventricular miR-223 mimic and long-term AAV overexpression rescued cognition, cut amyloid-beta plaque load, and restored synaptic markers; mechanism runs through microglia (clustering around plaques, upregulating AXL/TREM2/CD11c), abolished by pharmacological microglial depletion, and in human iPSC-derived microglia miR-223 directly targets endo-lysosomal genes including the AD risk gene SPPL2A — a proof of concept for RNAi therapeutics in the AD brain (caveats: mouse amyloid clearance rarely tracks human cognition; a miRNA is a broad microglial reprogrammer). PNAG upstream driver: poly-N-acetylglucosamine, a conserved microbial surface polysaccharide absent from mammalian cells, colocalizes with amyloid plaques in human AD hippocampus/cortex/substantia nigra (minimal in controls); purified PNAG and PNAG-bearing microbial vesicles trigger TLR2-dependent inflammasome signaling and drive amyloid-beta and phospho-tau in human neuronal/3D models; an anti-PNAG vaccine in APP/PS1 mice improved cognition, cut glial activation and amyloid, shifted amyloid processing toward less aggregation-prone species, and spared the gut microbiome — a specific, conserved, druggable antigen with an existing anti-PNAG antibody program in infection (caveat: colocalization is not causation; APP/PS1 vaccine wins are a crowded graveyard). TG2 in injury: whole-body transglutaminase-2 knockout mice show markedly less white-matter and default-mode-network pathology 28 days after repetitive weight-drop mild TBI (diffusion and resting-state MRI); astrocyte multi-omics show wild-type astrocytes impose a post-injury metabolic restriction that TG2 loss lifts, de-repressing glutamate-recycling and lipid-metabolism networks — TG2 (druggable, CNS-penetrant inhibitors exist) as a transcriptional gatekeeper of astrocyte metabolic versatility (caveat: constitutive, non-cell-type-specific, not a timed post-injury intervention). VRAC as antidepressant target: in zebrafish, disrupting the volume-regulated anion channel (pharmacology or lrrc8aa knockdown) produced anxiety/depression-like phenotypes and affective-gene shifts, while activation with zinc pyrithione reversed them and, under chronic unpredictable stress in adults, matched imipramine and normalized elevated monoamine oxidase — a non-monoaminergic, bidirectional loss/gain case (caveat: zebrafish, promiscuous tool compound, body-wide channel so CNS selectivity is the hard problem). Prefrontal CB1: deleting cannabinoid receptor 1 from mPFC neurons, and specifically from GABAergic interneurons, reshapes emotion sex- and cell-type-dependently — pan-neuronal deletion anxiolytic in females, GABAergic-specific anxiolytic in males (with increased interneuron calcium activity), and fear conditioning splitting the opposite way — a reminder that CB1-drug emotional effects can invert by sex and target cell. ProDock: a rank-resolved, multi-engine virtual-screening workflow combining global DiffDock and local GNINA docking with interpretable pose descriptors and Optuna-tuned re-ranking thresholds; on DUD-Z (43 targets) CNN-based precision-recall enrichment rose from about 0.20 to 0.29 with more native-like poses, but gains were non-uniform, the interaction-similarity metric is co-crystal-anchored (partly circular, may not transfer to novel chemotypes), and the held-out split is within-target not across proteins — a sensible engineering consolidation, not a leap. scGENet: turns single-cell foundation-model (scGPT) embeddings, fine-tuned on brain transcriptomes, into interpretable context-specific gene networks that best match curated neuronal pathways, Parkinson's risk loci, and patient signatures; applied to iPSC-derived PD midbrain organoids it recovers a conserved neurogenic program disrupted across genetic and idiopathic PD (reduced dopaminergic neurons; an SNCA/VGLUT2 subtype) — interpretability benchmarked against known biology. Wild-type huntingtin: knocking down the Drosophila huntingtin homolog in the eight lateral-ventral circadian pacemaker neurons (adult-restricted) weakens free-running rhythms, increases sleep, disrupts dense-core-vesicle axonal trafficking and its time-of-day rhythm, and lowers firing rate — a caution that non-allele-selective HTT-lowering therapies would sacrifice wild-type huntingtin's cell-autonomous roles in trafficking, excitability, and circadian output. CLOSE — genipin, a gardenia-derived iridoid, restores total sleep and reconsolidates fragmented nighttime sleep in an alpha-synuclein Drosophila PD model, showing a non-motor phenotype responds to the same intervention that rescues motor deficits. Nine items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.20.738977v1 ; https://www.biorxiv.org/content/10.64898/2026.07.20.739171v1 ; https://www.biorxiv.org/content/10.64898/2026.07.16.738974v1 ; https://www.biorxiv.org/content/10.64898/2026.07.22.740018v1 ; https://www.biorxiv.org/content/10.64898/2026.07.23.740323v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740457v1 ; https://www.biorxiv.org/content/10.64898/2026.07.26.740813v1 ; https://www.biorxiv.org/content/10.64898/2026.07.22.740122v1 ; https://www.biorxiv.org/content/10.64898/2026.07.16.738995v1 https://www.biorxiv.org/content/10.64898/2026.07.20.738977v1 2026-07-28-receptor-and-reason Tue, 28 Jul 2026 12:00:00 +0000 664 false Daily roundup for July 28, 2026, nine items; a glia-and-metabolism arc through novel psychiatric targets, computational drug discovery, and neurodegeneration. LEAD — miR-223, a glia-enriched microRNA, delivered as a mimic or via AAV, rescues cognition and clears amyloid-beta in App knock-in mice through microglial phagocytosis, targeting the AD risk gene SPPL2A. Persistent microbial PNAG colocalizes with plaques, drives TLR2 inflammasome signaling and amyloid/tau, and an anti-PNAG vaccine rescues APP/PS1 mice. Transglutaminase-2 deletion restores post-injury astrocyte metabolic versatility and attenuates repetitive mild-TBI white-matter pathology. The volume-regulated anion channel (VRAC) is nominated as a non-monoaminergic antidepressant target in zebrafish, with zinc pyrithione comparable to imipramine. Prefrontal GABAergic CB1 deletion reshapes anxiety and fear sex-dependently. ProDock adds multi-engine (DiffDock plus GNINA) rank-resolved re-ranking with Optuna for virtual screening, with honest benchmark caveats. scGENet turns single-cell foundation-model embeddings into interpretable gene networks and finds a conserved neurogenic program across Parkinson's. Wild-type huntingtin knockdown in circadian pacemaker neurons disrupts dense-core-vesicle trafficking and sleep — a caution for HTT-lowering therapies. CLOSE — genipin rescues alpha-synuclein sleep deficits in a Drosophila PD model. Paper links in show notes. Daily roundup for July 27, 2026, nine items; a drugs-and-mechanisms Monday across neuropsychopharmacology, receptor biophysics, computational drug design, and neuroimaging methods. Note: the bioRxiv details API returned HTTP 500 for the July 22 and 23 collection days (and partial pages elsewhere) during today's run, so the bioRxiv corpus is a partial scan of the 7-day window; items dated 07-22/07-23 were still recovered as revisions in later collection days. LEAD — elunetirom (Autobahn Therapeutics): a brain-targeted prodrug cleaved by fatty acid amide hydrolase to release LL-340001, a thyroid hormone receptor beta agonist; in primary cortical and hippocampal neurons the active compound grew and branched neurites, increased synapse number within 24 h (persisting to 72 h), protected synapses against amyloid-beta, and raised nuclear PGC-1-alpha, NRF2, functional mitochondria, and ATP; the plasticity effect was blocked by a TrkB antagonist (ANA-12) but NOT by a 5-HT2A antagonist (M100907), placing it on a BDNF-TrkB program distinct from the 5-HT2A route of psychedelics/psychoplastogens; translational authors include Stephen Stahl and Roger McIntyre; caveats — entirely primary-culture pharmacology (no behavior/in vivo) and an industry preprint. Duloxetine and RNA-protein interactions: the SNRI accumulates intracellularly and broadly reduces RNA-binding capacity (~80% of RNA-binding proteins in E. coli IAI1 and human Caco-2 cells); mechanistically it competes with aspartate, the natural substrate of the pyrimidine-biosynthesis enzyme PyrB, weakening PyrB binding to 3'-UTR stem-loops in its RNA partners — an off-target mode of action at the level of RNA-protein recognition, with implications for efficacy variability, toxicity, and microbiome effects. G-protein conformational efficacy: fluorine-19 quantitative NMR of G-alpha-s with the adenosine A2A receptor builds a three-state conformational model and shows ligand-bound receptors both differentially populate the substates AND impose distinct conformational-cycling kinetics — efficacy is an equilibrium-and-transition-rate signature, not a single active-state number; a reframing for design of biased/partial agonists (design for kinetics, not just a pose). Warhead-preserving generative design: a single-author preprint adds inference-time geometric shape guidance to a pretrained structure-based diffusion model; one-shot guidance increased target-volume coverage but produced disconnected fragments, while reformulating as smaller staged scaffold expansions (with target-directed selection, adaptive growth increments, beam search, and soft-scaffold inpainting) yielded controllable molecular growth toward a target ligand shape while preserving a fixed warhead — performance depends on ligand-pair geometric compatibility. Cross-disease neurodegeneration repurposing: pooled transcriptomes of Alzheimer's, Parkinson's, and Huntington's share 274 dysregulated genes and three upregulated pathways (two NF-kappa-B, one RUNX3-BCL2L11 tied to the neurodegeneration-vs-cancer inverse link); five nominated targets (NFKBIA, NFKB1, RELA, TRIM4, SMAD4) map to repurposable, commonly safe FDA drugs (statins, antihypertensives, antidiabetics, analgesics, diuretics); in-silico only, and shared end-stage DEGs can reflect consequence as much as cause. Stress-resilience circuit: machine-learning deep behavioral phenotyping plus longitudinal manganese-enhanced MRI and cFOS mapping in chronic social defeat show active adaptation has a latent pre-stress kinetic signature; resilient responders form a tightly integrated module (periaqueductal gray, VTA, basolateral amygdala, dorsal raphe) while non-responders fragment, with dorsal-raphe-to-basolateral-amygdala connectivity lost in non-responders; chemogenetic silencing of BLA-projecting dorsal raphe neurons during a social challenge shifted adaptive animals to the non-responsive phenotype — a load-bearing serotonergic pathway for coping. Ketamine EEG decoding (methods caution): re-analysis of an open 62-channel dataset (10 volunteers, awake vs subanaesthetic ketamine) under leave-one-subject-out CV; band power decoded the state at balanced accuracy ~0.71 while weighted phase-lag index connectivity was at chance (~0.47), and concatenation did not help; decomposing the drug effect into shared vs subject-specific components, connectivity was almost entirely subject-specific (shared fraction ~0.05) and thus non-transferable, whereas the spectral effect was ~0.53 shared; a five-channel lateral montage matched the full array — a group-level connectivity difference can be real yet useless for prediction. Schizophrenia rs-fMRI parcellation (methods caution): across two cohorts (n=159, n=255), individualized surface-based parcellation detected more pronounced quasi-periodic-pattern dynamics, stronger default-mode/dorsal-attention opposition, and larger, more reproducible patient-control differences than atlas-based pipelines, but PANSS symptom associations did not replicate across cohorts — preprocessing improves case-control sensitivity while stable brain-symptom relationships remain elusive. CLOSE — FAE-20 (ferulic acid eicosyl ester, a Rhodiola rosea constituent): subchronic treatment improved attentional set-shifting (cognitive flexibility) in both sexes of young adult mice, did not affect Y-maze working memory or Barnes-maze spatial learning generally, but rescued a subgroup of aged poor-learners; no change in hippocampal neurogenesis or spine density, with sex-specific modulation of the ascending arousal system (increased lateral-hypothalamic orexin neuron activity in males, reduced laterodorsal-tegmental cholinergic activity in females) — a task-demand- and baseline-dependent enhancer acting via arousal rather than the hippocampus. Nine items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.21.739927v1 ; https://www.biorxiv.org/content/10.64898/2026.07.23.740326v1 ; https://www.biorxiv.org/content/10.64898/2026.07.23.740425v1 ; https://www.biorxiv.org/content/10.64898/2026.07.21.739744v1 ; https://www.biorxiv.org/content/10.64898/2026.07.22.739294v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740522v1 ; https://www.biorxiv.org/content/10.64898/2026.07.21.739494v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740570v1 ; https://www.biorxiv.org/content/10.64898/2026.07.20.739577v1 https://www.biorxiv.org/content/10.64898/2026.07.21.739927v1 2026-07-27-receptor-and-reason Mon, 27 Jul 2026 12:00:00 +0000 719 false Daily roundup for July 27, 2026, nine items; a drugs-and-mechanisms Monday. LEAD — elunetirom, a brain-targeted (FAAH-cleaved) thyroid hormone receptor beta prodrug from Autobahn, grows neurites and synapses and boosts mitochondrial biogenesis in neurons via a TrkB-dependent, 5-HT2A-independent program — a differentiated antidepressant mechanism (in vitro only). The SNRI duloxetine broadly disrupts RNA-protein interactions off-target (~80% of RNA-binding proteins in bacteria and human cells), competing with aspartate at PyrB. Fluorine-19 qNMR of G-alpha-s with the adenosine A2A receptor shows ligands set efficacy via both conformational populations and cycling kinetics across a three-state landscape. Inference-time shape guidance on a pretrained structure-based diffusion model gives controllable, warhead-preserving molecular growth via staged scaffold expansion. A cross-disease transcriptomic analysis of Alzheimer's/Parkinson's/Huntington's nominates a shared NF-kappa-B axis and repurposes common safe FDA drugs. Deep phenotyping plus manganese-enhanced MRI/cFOS pin a dorsal-raphe-to-basolateral-amygdala projection as load-bearing for adaptive stress coping. Spectral power decodes the subanaesthetic ketamine state (~0.71) while phase connectivity is at chance (~0.47) because its drug effect is subject-specific. Individualized surface parcellation improves schizophrenia rs-fMRI sensitivity, but symptom associations fail to replicate. CLOSE — FAE-20, a Rhodiola rosea constituent, selectively enhances cognitive flexibility and rescues aged poor-learners by tuning ascending arousal, not the hippocampus. Paper links in show notes. Daily roundup for July 26, 2026, ten items; a psychedelics-and-computation Sunday spanning neuropsychopharmacology, computational drug discovery, and neurotherapeutics. LEAD — cardiac 5-HT2B safety of repeated low-dose psilocybin: because chronic psilocin exposure engages the cardiac 5-HT2B receptor (the fen-phen/ergot valvulopathy mechanism), the authors build an exposure-response model scoring functional Gq efficacy times exposure duration above endogenous serotonin tone, calibrated with no misclassifications across seven clinical anchors; every human-valvulopathic exposure scores high and a candidate 3 mg/day psilocybin regimen scores ~2 orders of magnitude below the weakest valvulopathic exposure, driven by psilocin being a low-efficacy partial 5-HT2B agonist (~52% of serotonin max vs 96% for norfenfluramine) with a ~2.5 h half-life producing pulsatile rather than sustained receptor drive; 12 days of continuous psilocin in rats above the projected human peak produced no valve lesions on blinded histopathology; caveats — 12 days cannot exclude slow multi-month fibrosis, and psilocin is a fuller agonist at the 5-HT2B beta-arrestin arm whose fibrogenicity is unknown; practical upshot is prospective echocardiography in trials. DOI and cortical network structure: widefield calcium imaging in awake mice shows the 5-HT2A agonist DOI acutely decreases cortical modularity by increasing anterolateral-posteromedial coupling and raising thalamo-cortical and long-range cortico-cortical influence onto the anterolateral domain, a mouse correlate of human psychedelic network desegregation. Tx2Mol (computational drug discovery): a transcriptome-guided generative framework that designs molecules from gene-expression signatures while preserving phenotypic guidance, beating 9 baselines across 10 cancer-relevant perturbation benchmarks (about +24% average, over +50% on HDAC1 similarity to known ligands) and generalizing to single-cell and patient-derived signatures — phenotype-first design when no clean target exists; caveats are preprint status, a rediscovery-biased metric, and cancer-only validation. Chemical chain-of-thought as a hallucination-prone molecular scratchpad: chain-of-thought improves LLM chemistry scores but the intermediate steps are frequently chemically invalid even when final answers are correct — visible reasoning is not verified reasoning, and molecular deployments need external chemical validators. Alpha-synuclein targeted protein degradation: peptide amphiphiles self-assemble into high-aspect-ratio supramolecular nanofibers presenting epitopes that bind alpha-synuclein and recruit chaperone-mediated autophagy, internalizing and selectively lowering alpha-synuclein in vitro — a modular multivalent alternative to bifunctional small-molecule degraders for aggregation-prone CNS proteins (early, in vitro). Parkinsonian freezing and DBS (computational, negative result): a dynamic-neural-field basal-ganglia model doing action selection and specification cleanly separates healthy from dopamine-depleted freeze-prone regimes, but modeling DBS as scaled reduction of afferent drive to the subthalamic nucleus and internal globus pallidus did not consistently restore action selection, signaling that the standard account of DBS is too simple. Medial pulvinar and schizophrenia (marmoset): neonatal (not adult) bilateral medial pulvinar lesions alter adolescent prefrontal diffusion trajectories and produce adult working-memory deficits, with reduced thalamocortical input to layer-3 parvalbumin interneurons, weaker gamma, less parvalbumin, and immature fast-spiking physiology — relocating the origin of prefrontal inhibitory dysfunction to developmental thalamic input. NDUFV2 pseudogene and schizophrenia: NDUFV2P1 is upregulated in patient brain and peripheral cells, inversely tracks NDUFV2 and mitochondrial respiration (not via microRNA sponging per in-silico analysis); overexpression impairs mitochondrial membrane potential, dynamics, and oxygen consumption in healthy lymphoblast lines, knockdown rescues patient-derived lines, and overexpression in rat cortical neurons impairs synapse formation and firing — a candidate lever on schizophrenia bioenergetics. HDAC3 and memory persistence: selective HDAC3 inhibition immediately or 6 h after training enhances novel-object-recognition memory (biphasic), NF-kappa-B inhibition at the same points impairs it, and HDAC3 inhibition increases nuclear NF-kappa-B in hippocampal CA1 — first in-vivo link between HDAC3 and NF-kappa-B translocation during consolidation, with timing part of the mechanism. CLOSE — the "dark magic mushroom" Galerina indica co-produces psilocybin and amatoxins (first single species with both), acquiring psilocybin biosynthesis by horizontal gene transfer twice independently after amatoxin biosynthesis, possibly coinciding with reduced amatoxin potency. Ten items per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.19.739440v1 ; https://www.biorxiv.org/content/10.64898/2026.07.20.739656v1 ; https://www.biorxiv.org/content/10.64898/2026.07.21.739736v2 ; https://arxiv.org/abs/2607.20935 ; https://www.biorxiv.org/content/10.64898/2026.07.20.739556v1 ; https://www.biorxiv.org/content/10.64898/2026.07.21.738458v1 ; https://www.biorxiv.org/content/10.64898/2026.07.23.740162v1 ; https://www.biorxiv.org/content/10.64898/2026.07.23.740347v1 ; https://www.biorxiv.org/content/10.64898/2026.07.16.738883v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739257v1 https://www.biorxiv.org/content/10.64898/2026.07.19.739440v1 2026-07-26-receptor-and-reason Sun, 26 Jul 2026 12:00:00 +0000 766 false Daily roundup for July 26, 2026, ten items; a psychedelics-and-computation Sunday. LEAD — an exposure-response model of cardiac 5-HT2B safety argues repeated low-dose (3 mg/day) psilocybin sits about two orders of magnitude below the weakest valvulopathic exposure, because psilocin is a low-efficacy partial 5-HT2B agonist (~52% of serotonin) with a short (~2.5 h) half-life giving pulsatile rather than sustained Gq drive; 12-day continuous rat dosing above the human peak showed no valve lesions. The psychedelic DOI (5-HT2A agonist) decreases cortical network modularity in awake-mouse imaging by raising thalamo-cortical and cortico-cortical influence onto anterolateral cortex. Tx2Mol generates de novo molecules from gene-expression signatures while preserving phenotype, beating 9 baselines across 10 cancer benchmarks. A preprint shows chemical chain-of-thought is a hallucination-prone molecular scratchpad — chemically invalid intermediate steps even when answers are right. Supramolecular peptide-amphiphile nanofibers degrade alpha-synuclein via chaperone-mediated autophagy in vitro. A dynamic-neural-field basal-ganglia model finds the standard "DBS = reduced STN/GPi drive" formulation does not restore action selection (instructive negative result). Neonatal medial pulvinar lesions in marmosets produce adult working-memory deficits and immature prefrontal parvalbumin interneurons, relocating schizophrenia-relevant dysfunction to developmental thalamic input. An NDUFV2 pseudogene upregulated in schizophrenia suppresses Complex I and respiration, with knockdown rescuing patient-derived cells. HDAC3 inhibition enhances memory persistence biphasically via increased nuclear NF-kappa-B in CA1. CLOSE — the "dark magic mushroom" Galerina indica co-produces psilocybin and amatoxins via horizontal gene transfer. Paper links in show notes. Daily roundup for July 25, 2026, ten items; a dopamine-and-reward-heavy Saturday spanning neuropsychopharmacology, addiction, aversion circuits, receptor biophysics, pain, and computational drug discovery. LEAD — miR-10a-5p and dopamine-agonist impulse-control disorders (rat): dopamine D2/D3 agonists like pramipexole trigger pathological gambling/buying/hypersexuality in a vulnerable subset (~18% of Parkinson's patients on agonists); rats stratified by baseline impulsivity on a delay-discounting task plus subchronic pramipexole and microRNA sequencing of dorsal striatum and nucleus accumbens show pramipexole increases impulsive choice only in low/mid-impulsive animals (ceiling effect in high-impulsive), and microRNA-10a-5p tracks the effect (constitutively high in trait-impulsive animals, drug-upregulated in the responders); viral overexpression confirms downregulation of PI3K-Akt-mTOR and BDNF pathways and, restricted to accumbens (not dorsal striatum), is sufficient to make rats impulsive — an accumbens-localized molecular lever for iatrogenic ICDs and a microRNA-directed therapeutic angle. Two modes of striatal dopamine release: a diffuse mode driving bulk extracellular accumulation vs a spatially restricted point-to-point mode; eliminating diffuse release drops striatal excitability and locomotion, while point-to-point transmission alone maintains spine density and supports motor/associative learning — release geometry multiplexes movement vs learning. VTA astrocyte Gq (chemogenetic): facilitates cocaine place preference in drug-naive rats but suppresses it in cocaine-experienced rats and reduces self-administration — a history-dependent astrocytic node in cocaine use disorder. Met-enkephalin in fear learning: a met-enkephalin sensor shows local amygdala release during auditory fear conditioning that shifts from outcome to predictive cue; the amygdalo-striatal transition zone is the source, spreading by volume transmission to lateral amygdala; enkephalin knockdown enhances fear learning; enkephalin suppresses dopamine via mu-opioid receptors — a diffuse opioid brake on fear memory. Mesohabenular glutamate and uncontrollable stress: VTA glutamatergic axons in lateral habenula report aversive stimuli independent of stress history, but silencing them during each inescapable-stress trial prevents the social and non-social consequences (learned helplessness) — a circuit gate for stress susceptibility. CNIH3 and slow AMPA receptors: abundant slow AMPA responses in CA1 pyramidal cells (mosaic, dendrite-level) are conferred by the auxiliary subunit cornichon-3 (CNIH3), expressed in a ventral-high/dorsal-sparse gradient; shRNA knockdown ablates slow responses in ventral CA1 and dorsal overexpression introduces them — a single auxiliary protein sets coincidence-detection vs temporal integration ("detonator" synapses). Mediodorsal thalamus pain subdivisions: medial-central vs lateral MD dissociate the affective-motivational (avoidance, via anterior cingulate) from sensory-discriminative components of pain, targeting distinct cortical interneuron populations in cingulate/prelimbic cortex — separable thalamocortical channels for hurting vs caring. Target Preference Maps (computational drug discovery): learns atom-type-specific spatial preference maps from local protein microenvironments, excluding whole-ligand topology to avoid memorization and capture transferable interaction preferences; recovers bridging waters and metal-dependent environments; validated retrospectively and prospectively guiding medicinal-chemistry optimization at a challenging protein-protein interface; feeds downstream docking/generative pipelines. CHIMIYA-1 (ADMET foundation model): tops the Therapeutics Data Commons ADMET benchmark (first on 4 endpoints, top decile on 20/22) under a conservative five-seed, structural-overlap-audited protocol — with the pointed subtext that several top public comparators fail that scrutiny (leaderboard leakage / single-seed inflation). CLOSE — edge-controllability and rTMS: in 25 treatment-resistant depression patients, baseline edge-controllability of middle-frontal-gyrus-centered structural connections predicts HAMD improvement after 5 weeks of prefrontal rTMS — exploratory, single-investigator, but the kind of individualized circuit predictor personalized neuromodulation needs. Ten items; neuropsychopharmacology + addiction + aversion/stress circuits + receptor biophysics + pain + computational pharmacology per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.20.739499v1 ; https://www.biorxiv.org/content/10.64898/2026.07.24.740623v1 ; https://www.biorxiv.org/content/10.64898/2026.07.20.739241v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739254v2 ; https://www.biorxiv.org/content/10.64898/2026.07.21.739922v1 ; https://www.biorxiv.org/content/10.64898/2026.07.22.739851v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739109v2 ; https://www.biorxiv.org/content/10.1101/2025.08.01.668090v11 ; https://www.biorxiv.org/content/10.64898/2026.07.20.739289v1 ; https://www.biorxiv.org/content/10.64898/2026.07.11.737986v2 https://www.biorxiv.org/content/10.64898/2026.07.20.739499v1 2026-07-25-receptor-and-reason Sat, 25 Jul 2026 12:00:00 +0000 703 false Daily roundup for July 25, 2026, ten items; dopamine-and-reward-heavy Saturday. LEAD — microRNA-10a-5p as the accumbens-localized trait-vulnerability substrate for dopamine-agonist (pramipexole) impulse-control disorders: pramipexole raises impulsive choice only in low/mid-impulsive rats (ceiling in high-impulsive), miR-10a-5p tracks it, and accumbens-restricted overexpression recapitulates the effect via PI3K-Akt-mTOR/BDNF suppression. Two spatial modes of striatal dopamine release dissociate locomotion (diffuse/bulk) from learning and spine maintenance (point-to-point). VTA astrocyte Gq signaling gates cocaine place preference history-dependently and cuts self-administration. Met-enkephalin volume transmission from the amygdalo-striatal transition zone brakes fear learning by suppressing lateral-amygdala dopamine via mu-opioid receptors. VTA-glutamate to lateral habenula gates the consequences of uncontrollable stress (learned helplessness) while reporting current aversion independent of history. Cornichon-3 (CNIH3) is the auxiliary-subunit signature converting fast AMPA synapses into slow summating "detonators" along a ventral-dorsal CA1 gradient. Mediodorsal thalamus subdivisions dissociate affective (avoidance, via cingulate) from sensory pain. Target Preference Maps learns transferable atom-level drug-receptor interaction preferences without whole-ligand memorization and guides real medicinal-chemistry optimization. CHIMIYA-1 tops the TDC ADMET benchmark under a five-seed, overlap-audited protocol that exposes probable public-leaderboard leakage. CLOSE — edge-controllability of the structural connectome predicts rTMS response in depression (n=25, exploratory). Paper links in show notes. Daily roundup for July 24, 2026, ten items; a psychedelics-and-reward-heavy week across neuropsychopharmacology, addiction, pain pharmacology, and computational drug design. LEAD — Kwan lab (psilocybin dendritic calcium imaging): two-photon imaging of apical tufts in layer-5 pyramidal-tract neurons of mouse medial frontal cortex shows psilocybin transiently raises spontaneous dendritic calcium events tracking brain pharmacokinetics, selectively during quiet wakefulness, abolished by cell-type-specific 5-HT2A deletion; and the normal calcium-predicts-spine-formation relationship breaks after psilocybin (no spinogenesis observed in this head-fixed prep — a flagged limitation) — a cellular correlate for set-and-setting and a dissociation of acute from long-term plasticity relevant to non-hallucinogenic psychoplastogens. Moghaddam lab (psilocybin, probabilistic choice + mPFC single units + RL modeling): uncertainty-dependent behavior (better choice at low uncertainty, more exploration at high), bidirectional learning-rate shift (up from rewarded, down from unrewarded actions), matched by enhanced mPFC coding of rewarded and diminished coding of unrewarded outcomes — an asymmetric optimism update as a bridge to antidepressant mechanism. Orexin→supramammillary (SuM) pathway: OX2R-predominant, terminals activated by footshock/threat cues and suppressed during consumption, yet optogenetic drive is reinforcing and evokes accumbens dopamine — converts threat salience into approach; relevant to orexin antagonists in addiction/relapse. dHPC→NAc vs BLA→NAc chemogenetic inhibition during WIN 55,212-2 self-administration: silencing either input increased addiction-like phenotype and persistence; dHPC input also raised motivation/impulsivity/reward sensitivity/extinction resistance while BLA input selectively raised cue-induced seeking — these glutamatergic projections act as a brake on the transition to compulsive use. Nicotine Go/No-Go self-administration: nicotinic (mecamylamine-sensitive) signaling carries impulsive action while D1 signaling carries motivation (D1 agonist A77636 cut intake without touching No-Go; D1 antagonist SCH23390 confounded by general output reduction) — motivation and impulsivity may need separate targeting in cessation pharmacotherapy. Endocannabinoid DRG electrophysiology: CBD and THC inhibit nociceptor firing (~5 µM IC50), terpenes (linalool, β-pinene, myrcene) too, and CBD+THC / CBD+terpene combinations are synergistic — first rigorous electrophysiological support for the entourage effect; caveats include modest potency and CBD hERG liability. Human DRG Nav1.7 (AM-2099) vs Nav1.8 (suzetrigine): Nav1.7 block raises AP threshold and refractory period but barely affects repetitive firing, while Nav1.8 block dramatically suppresses repetitive high-frequency firing — a clean biophysical explanation for a decade of Nav1.7 clinical failure vs suzetrigine's success. CAFE: a training-free, inference-time protocol that places a competitive orthosteric blocker (e.g. ADP for kinases) to divert fragments into allosteric and cryptic pockets in Boltz-2 co-folding, with binding free energies matching/exceeding crystal poses and cryptic pockets missed by conventional tools; generalized to RAS-MAPK and fragment screening. GEM-GPT: fuses an scRNA-seq foundation model with a molecular GPT for cell-type-resolved generative systems pharmacology, beats baselines, generalizes to unseen cell types, and nominates novel chemotypes plus FDA-approved repurposing candidates in an opioid-use-disorder case study (entirely computational — illustrative, not validated). CLOSE — task-based value generalization (N=163 online, transdiagnostic): breadth of learned-value spread tracks self-reported positive overgeneralization and subclinical bipolar symptoms, captured by an RL model where self-efficacy modulates the influence of anticipated future value. First episode back after the download-driven pause. Ten items; neuropsychopharmacology + psychedelics + addiction + pain pharmacology + AI-driven drug design + computational psychiatry per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.16.738983v1 ; https://www.biorxiv.org/content/10.64898/2026.07.16.733837v1 ; https://www.biorxiv.org/content/10.64898/2026.07.18.739326v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739252v1 ; https://www.biorxiv.org/content/10.64898/2026.07.16.739006v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739255v1 ; https://www.biorxiv.org/content/10.64898/2026.03.26.714428v2 ; https://www.biorxiv.org/content/10.64898/2026.07.19.739466v1 ; https://www.biorxiv.org/content/10.64898/2026.07.17.739269v1 ; https://www.biorxiv.org/content/10.64898/2026.07.12.737635v1 https://www.biorxiv.org/content/10.64898/2026.07.16.738983v1 2026-07-24-receptor-and-reason Fri, 24 Jul 2026 12:00:00 +0000 950 Daily roundup for July 24, 2026, ten items; psychedelics-and-reward-heavy week. LEAD — Kwan lab psilocybin two-photon dendritic calcium imaging in layer-5 pyramidal-tract neurons of mouse medial frontal cortex: transient rise in spontaneous apical-tuft calcium events tracking brain PK, selectively during quiet wakefulness, abolished by cell-type-specific 5-HT2A deletion; the normal calcium→spine-formation coupling breaks after psilocybin (no spinogenesis in this head-fixed prep, flagged) — cellular correlate for set-and-setting and an acute-vs-long-term plasticity dissociation for psychoplastogen programs. Moghaddam lab psilocybin probabilistic choice + mPFC single units + RL: uncertainty-dependent choice, bidirectional learning-rate shift (up from reward, down from non-reward), matched mPFC coding changes — asymmetric optimism update as antidepressant-mechanism bridge. Orexin→supramammillary (OX2R-predominant): threat/salience-activated, consumption-suppressed, but optogenetically reinforcing with accumbens dopamine — threat salience converted to approach; orexin-antagonist relevance for relapse. dHPC→NAc vs BLA→NAc chemogenetic inhibition during WIN 55,212-2 self-administration: both raise addiction phenotype + persistence; dHPC also raises motivation/impulsivity/reward sensitivity/extinction resistance, BLA selectively raises cue-induced seeking — glutamatergic brake on compulsive transition. Nicotine Go/No-Go: nicotinic (mecamylamine) carries impulsive action, D1 carries motivation (agonist A77636 cut intake sparing No-Go; antagonist SCH23390 output-confounded). Endocannabinoid DRG: CBD/THC (~5 µM) + terpenes inhibit nociceptor firing, CBD+THC/terpene synergy — entourage-effect electrophysiology; CBD hERG caveat. Human DRG Nav1.7 (AM-2099) vs Nav1.8 (suzetrigine): Nav1.7 raises threshold/refractory but spares repetitive firing; Nav1.8 suppresses repetitive firing — biophysical why Nav1.7 blockers failed and suzetrigine works. CAFE: training-free orthosteric-blocker trick (ADP for kinases) redirects Boltz-2 co-folding to allosteric/cryptic sites, free energies matching crystal poses; generalized to RAS-MAPK + fragment screening. GEM-GPT: scRNA-seq foundation model + molecular GPT for cell-type-resolved systems-pharmacology molecule generation, OUD case study (computational, illustrative). CLOSE — task-based value overgeneralization tracks subclinical bipolar symptoms, RL model with self-efficacy-modulated future value. First episode back after the pause. Paper links in show notes. false Your daily briefing has been paused because no downloads have been detected in about two weeks. As soon as you download an episode, generation will automatically resume. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-17-goodbye Fri, 17 Jul 2026 09:00:08 +0000 7 Your daily briefing has been paused because no downloads have been detected in about two weeks. As soon as you download an episode, generation will automatically resume. Daily roundup for July 16, 2026: bioRxiv Knowles lab (Cambridge, UK) + Sahtoe lab (Utrecht) — deep-learning-based de novo protein design generates compact single-chain binders engaging the intrinsically disordered NAC domain of α-synuclein via stabilization of its extended β-strand aggregation-competent conformation, 3/21 designs engage target in vitro + in live cells, single partial-diffusion refinement matures strongest binder to Kd ≈1.94 nM with no detectable cross-reactivity to tau/amylin/Aβ, kinetic analysis with monomer/oligomer/fibril affinities pinpoints per-binder microscopic step of inhibition + most potent binder selectively captures on-pathway oligomers (species most closely linked to toxicity) + suppresses fibril formation at substoichiometric ratios without engaging bulk monomer — substoichiometric potency on an IDP target is the paradigm-shift feature vs Prothena/Roche/Biogen antibody programs; bioRxiv anti-α-synuclein single-domain-antibody screen — 5 sdAbs against phospho-Ser129 α-syn in mouse cultures + Drosophila synucleinopathy readouts, sdAb 2H1 (previously unreported) binds aggregation-prone region, lowers phospho-Ser129 α-syn, prevents DA-neuron loss, alleviates mitochondrial dysfunction, improves motor + prolongs fly survival + TurboID biotinylation localizes engagement to synapse/vesicle trafficking machinery; bioRxiv α-syn preformed-fibril seeding differentially affects SNc DA subpopulations — vulnerable neurons show altered tonic firing + T-type Ca current changes while resilient neurons show increased excitability, hands the field a specific ion-channel handle in a Cav3-modulator space with ethosuximide precedent + emerging T-type-selective molecules; bioRxiv AAV-Parkin gene therapy for Parkin-deficient EOPD — pUbSer65 validated as mitophagy-initiation biomarker (reduced in Parkin-KO SH-SY5Y under mitochondrial stress + restored by AAV-Parkin + restored in patient-derived fibroblasts), rat SN delivery with AAV1 EF1α promoter shows dose-dependent Parkin expression, screen of promoters in proprietary Spark100 capsid identifies EF1α + Synapsin1 as most effective for SN DA neurons at well-tolerated doses; bioRxiv preclinical AAV9-E2-dCas9-VP64 (RT101) CRISPR-activation of endogenous SCN1A in PV interneurons for Dravet syndrome — saturating gRNA screen across human SCN1A promoter, dose-dependent SCN1A upregulation in human Dravet iPSC GABAergic neurons, ICV administration in Dravet mice produces dose-dependent survival improvement + reduced hyperthermia-induced seizure susceptibility + increased Nav1.1 with maintained PV-interneuron selectivity + minimal off-target, MRI-guided ICV in juvenile cynomolgus macaques is well-tolerated with broad cortical biodistribution + strong peripheral detargeting — being advanced toward clinical evaluation; bioRxiv P301S tauopathy on Galectin-3 KO background — Gal-3 loss attenuates hyperphosphorylated + pathological tau across cortex/hippocampus/piriform-entorhinal, normalizes tau-kinase signaling + mitochondrial + trafficking pathways, dampens microglial activation, preserves white matter integrity + reduces axonal degeneration, proteomic/phosphoproteomic profiles of Tau-Gal3KO cluster with WT rather than tau-only + mechanism is enhanced microglial myelin phagocytosis + lysosomal degradation + human iPSC neurons show extracellular Gal-3 exacerbates tau seeding reversed by Gal-3 inhibitors — short repurposing path from phase-2 Gal-3 programs in pulmonary fibrosis/NASH into FTD/AD (caveat: tau-only P301S, needs co-pathology validation); bioRxiv human iPSC forebrain organoid with doxycycline-inducible SOX10/OLIG2/NKX6-2 generates mature myelinating oligodendrocytes at scale + APP pathogenic mutations produce Aβ plaques + phospho-tau + reduced MBP + disrupted myelin ultrastructure + spatial transcriptomics reveals plaque-density-dependent oligodendrocyte reprogramming coordinately inducing immune activation + calcium signaling + lipid remodeling + proteasome subunit remodeling + program conserved in human AD postmortem = AD myelin loss is oligodendrocyte proteostasis failure not demyelination signal, reframes white-matter drug strategy toward proteostatic support rather than OPC differentiation drivers; bioRxiv follow-up to 2025 Dendritic Syntaphilin Intrusion (DSI) MS neurodegeneration mechanism — primary hippocampal cultures under inflammatory cytokines sensitize DSI + synaptic GluN2A NMDAR blockade increases DSI + extrasynaptic GluN2B NMDAR blockade reduces DSI + GluN2A/GluN2B balance oppositely regulates SNPH mislocalization — classical synaptic-vs-extrasynaptic NMDA dichotomy on specific neurodegeneration substrate, gives ifenprodil/radiprodil-class GluN2B antagonists a neuroprotective target in progressive MS; bioRxiv v5 macaque V1 layer-4C nicotine local delivery + cortical-depth recording during perceptual task — widespread heterogeneous gain across cortical layers (enhancement + suppression) not explained by stimulus properties but well-predicted by adaptation of normalization model of divisive gain control + circuit does not compensate + gain changes evident during behavior — implications for varenicline/encenicline-class cognitive-enhancement programs (assumed simple attention-boost, actually uncontrolled layer-heterogeneous gain) + alpha-7 nicotinic biology in schizophrenia; bioRxiv 7T fMRI cognitive restructuring paradigm (repeat vs Socratic-challenge negative statements) in 73 clinical participants (depressive/anxiety disorders) vs 70 controls + dynamic causal modeling — clinical + control participants modify negative beliefs equally well behaviorally but depressive symptoms correlate with reduced dlPFC + vlPFC over-inhibition of vmPFC while anxiety symptoms correlate with greater dlPFC + increased dlPFC→amygdala excitation + vlPFC over-inhibition of amygdala = dissociable circuit-level dysfunction for same CBT operation, argues for symptom-profile-guided pharmacotherapy augmentation (amygdala-targeting for anxiety-dominant vs prefrontal modulators for depression-dominant); arXiv Automatic ODE Discovery for Biological Systems using LLM-Powered Agentic System — multi-agent LLM framework with specialized agents for hypothesis generation + numerical ODE solver testing + statistical validation + iterative refinement, tested across synthetic + real biological ODE systems (population dynamics, biochemical networks, gene regulation, metabolic pathways), recovers ground-truth parameters on synthetic data + comparable-or-better vs expert-driven manual modeling with substantially less human time = agentic scaffolding for CNS QSP model structure discovery, bottleneck was never parameter fitting but structure hypothesis generation. Eleven items; neuropsychopharm + Parkinson's + Alzheimer's + MS + Dravet gene therapy + tauopathy + cognitive-restructuring circuits + agentic AI in comp bio. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.13.738012v1 ; https://www.biorxiv.org/content/10.64898/2026.07.07.735601v1 ; https://www.biorxiv.org/content/10.64898/2026.07.10.737840v1 ; https://www.biorxiv.org/content/10.64898/2026.07.09.737487v1 ; https://www.biorxiv.org/content/10.64898/2026.07.12.737793v1 ; https://www.biorxiv.org/content/10.64898/2026.07.07.736964v1 ; https://www.biorxiv.org/content/10.64898/2026.07.08.737318v1 ; https://www.biorxiv.org/content/10.64898/2026.07.08.737141v1 ; https://www.biorxiv.org/content/10.64898/2026.02.22.707315v5 ; https://www.biorxiv.org/content/10.64898/2026.07.12.738091v1 ; https://arxiv.org/abs/2607.13608 https://www.biorxiv.org/content/10.64898/2026.07.13.738012v1 2026-07-16-receptor-and-reason Thu, 16 Jul 2026 12:00:00 +0000 1072 Daily roundup for July 16, 2026, eleven items. LEAD — Knowles (Cambridge) + Sahtoe (Utrecht) de novo protein design generates single-chain binders engaging the intrinsically disordered NAC aggregation-prone domain of α-synuclein via stabilization of its extended β-strand conformation; strongest binder matures to ~2 nM affinity with no cross-reactivity to tau/amylin/Aβ, selectively captures on-pathway oligomers, and suppresses fibril formation at substoichiometric ratios without engaging bulk monomer — substoichiometric potency on an intrinsically disordered target is the paradigm shift vs Prothena/Roche/Biogen α-syn antibody programs. Anti-α-syn sdAb screen identifies previously unreported nanobody 2H1 that binds aggregation-prone region, lowers phospho-Ser129 α-syn, prevents DA neuron loss + mitochondrial dysfunction + motor deficit + shortened survival in Drosophila synucleinopathy + TurboID localizes engagement to synapse/vesicle trafficking. α-syn PFF seeding differentially affects SNc DA subpopulations — vulnerable disrupt T-type Ca + tonic firing, resilient increase excitability — hands the field a Cav3-modulator handle with ethosuximide precedent. AAV-Parkin gene therapy for Parkin-deficient EOPD validated with pUbSer65 mitophagy biomarker + rat SN AAV1 EF1α delivery + Spark100 capsid promoter screen identifying EF1α/Synapsin1 for SN DA neurons. AAV9-E2-dCas9-VP64 (RT101) CRISPR activation of endogenous SCN1A in PV interneurons for Dravet syndrome — saturating gRNA screen + dose-dependent SCN1A upregulation in human iPSC Dravet neurons + ICV in mice improves survival/seizure susceptibility + juvenile cynomolgus MRI-guided ICV shows broad cortical biodistribution + peripheral detargeting, being advanced toward clinical evaluation. P301S tauopathy on Galectin-3 KO robustly attenuates tau pathology + restores mitochondrial + trafficking pathways + enhances microglial myelin phagocytosis + Gal-3 inhibitors reverse tau seeding in iPSC neurons — short repurposing lane from phase-2 pulmonary-fibrosis/NASH Gal-3 programs to FTD/AD. Inducible SOX10/OLIG2/NKX6-2 human forebrain organoid + APP mutations reveals plaque-density-dependent oligodendrocyte proteostatic reprogramming that decouples MBP transcript from MBP protein, conserved in AD postmortem — AD myelin loss is proteostasis failure not demyelination signal. Synaptic GluN2A vs extrasynaptic GluN2B NMDAR balance oppositely regulates Dendritic Syntaphilin Intrusion in inflammatory MS — GluN2B-selective antagonists as neuroprotective lever for progressive MS. Macaque V1 layer-4C nicotine + cortical-depth recording during perceptual task — widespread heterogeneous gain across layers predicted by divisive normalization model with no circuit compensation, implications for varenicline/encenicline-class cognitive-enhancement + alpha-7 nicotinic in schizophrenia. 7T fMRI cognitive restructuring in depression/anxiety + DCM — behavioral output preserved but depression maps to reduced dlPFC + vlPFC over-inhibition of vmPFC while anxiety maps to increased dlPFC + increased dlPFC→amygdala excitation + vlPFC over-inhibition of amygdala = dissociable circuits for same CBT operation, argues for symptom-profile-guided pharmacotherapy augmentation. arXiv Automatic ODE Discovery for Biological Systems using LLM-Powered Agentic System — multi-agent hypothesis-generation + numerical ODE testing + statistical validation + iterative refinement, recovers ground-truth parameters + comparable-to-better vs expert manual modeling — agentic scaffolding for CNS QSP model structure discovery. Paper links in show notes. false Daily roundup for July 15, 2026: bioRxiv Bryan Roth lab (UNC Chapel Hill) revised preprint solves cryo-EM structures of eleven distinct endogenous mGluR2 receptor assemblies isolated from mouse brain via CRISPR mCherry-tagged knock-in mice and rapid immunoaffinity purification — mGluR2/3 heterodimers detected only in active-state complexes (not a static species but an activation-associated ensemble), endogenous ternary complexes with mGluR2 homodimers and mGluR2/3 heterodimers coupled to a single G-alpha-oA heterotrimer differ significantly from recombinant reconstructions and provide a real blueprint for state-selective mGluR2 modulators (relevant to the pomaglumetad/mGluR2 PAM schizophrenia clinical-failure lane); bioRxiv Sato-Bigbee lab (Virginia Commonwealth) find CSF nociceptin dramatically elevated in progressive-MS patients but not RRMS or controls, reproduce phenomenon in EAE in older mice (age equivalent to human progressive-MS transition risk), and rescue with NOP-receptor antagonist — regression of clinical scoring, higher OLG:OPC ratios, increased myelination, reduced reactive astrocytes; human astrocytes upregulate nociceptin under proinflammatory cytokines — NOP antagonism as testable pharmacological lever for progressive MS with existing clinical tool molecules; bioRxiv Vincent Laurent lab (UNSW Sydney, with Stephen Maren) — optogenetic silencing of diagonal-band-to-infralimbic cholinergic projection during fear extinction (not conditioning) enhances retrieval + prevents renewal in ABA/AAB paradigms, anatomically specific (HDB→PL projection inert), mechanism via feedforward inhibition of IL pyramidal cells through superficial-layer interneurons, and nicotinic (not muscarinic) IL blockade reproduces the effect — brief nicotinic antagonism during exposure therapy as surgical intervention against contextual relapse; bioRxiv Pierre-Eric Lutz lab (INCI CNRS Strasbourg) — sciatic nerve cuff neuropathic pain-induced depression in mice profiled via genome-wide DNA methylation + three histone marks + RNA-seq in ACC — both sexes show extensive epigenomic remodeling at largely distinct genomic loci but converging on overlapping synapse-related gene modules with similar regulatory features (TF binding, chromatin annotations) — sex-specific-but-functionally-convergent architecture argues shared-downstream synapse-modulating drugs work in both sexes while upstream epigenetic modifiers (HDAC/DNMT inhibitors) need sex-specific targeting; bioRxiv Rafiq Huda lab (Rutgers) fiber photometry in dorsolateral striatum during locomotion reveals two-timescale DA-ACh-astrocyte-Ca relationship (fast DA/ACh anticorrelation; slower positive covariation with astrocyte Ca), muscarinic/D1/D2 blockade + closed-loop optogenetic CIN silencing both reduce astrocyte Ca, 6-OHDA PD attenuates ACh release + astrocyte Ca — striatal astrocytes as underexplored PD therapeutic target with muscarinic access point that may be doing much of its work through astrocytes not neurons; bioRxiv v3 Aaron Wong lab (University of Delaware) 37 PD patients ON/OFF meds + 39 controls dissociate reward sensitivity vs effort perception vs effort-reward slope — strong Bayes-factor-supported null on reward sensitivity + effort-reward slope, positive finding on sensorimotor effort perception (lower match forces + higher subjective ratings), fatigue ruled out — critically dopamine medication did NOT shift effort perception, so L-DOPA is the wrong lever for effort-perception-driven bradykinesia (look upstream: sensorimotor cortex, cerebellum, non-DA cortico-BG loops); bioRxiv Caldwell lab (University of Alabama) DA-α-synuclein interaction is the molecular hinge uncoupling C. elegans A53T α-syn DAergic neurodegeneration from organismal aging — DA overexpression exacerbates degeneration, cat-2 knockout rescues neurons, DA-interaction-motif mutation renders neurons DA-insensitive; same interaction drives compartmentalized dual response (localized DAergic degeneration via oxidative stress + organism-wide TFEB/HLH-30 proteostatic remodeling extending lifespan) — two-vector pharmacology prescription (autophagy boost without DA-oxidative stress); bioRxiv Walter Reed Army Institute of Research — 26 healthy adults, 2h restricted sleep + 1h SO-tDCS at 0.75 Hz + 46h sleep deprivation, PVT vigilance protected in STIM group + paired-word learning improved at T120 vs pre-sleep in STIM (SHAM deteriorated) — brief slow-oscillation stimulation as pharmacology-adjacent intervention for sleep-restricted cognition, relevant to shift work/military/residency contexts (small sample + tDCS replication caveats acknowledged); bioRxiv TCS Research (Hyderabad) hybrid QSP + reinforcement learning + Monte Carlo tree search framework for personalized ipilimumab immunotherapy dosing — 95.2% remission (matching highest 10 mg/kg fixed dose) achieved with 72 mg/kg median total dose (comparable to lowest 3 mg/kg fixed regimen), drug-activity rate emerges as dominant learned determinant of long-term outcome — mechanistically-grounded RL-plus-MCTS on properly-fit CNS QSP model could rewrite psychiatric dose-finding at fraction of current per-patient exposure risk; bioRxiv Queen's University Belfast neuro-symbolic Octopus framework combines localized privacy-preserving multi-agent LLM swarm with regularised ML predictive environments chained from CRISPR dependency (CCLE) through XGBoost SHAP attribution to PDX tumor trajectory to human survival — unsupervised sweep autonomously prioritises IGF2 as 5-FU resistance vulnerability, BH-FDR q=0.0292 log-rank p=0.0007, PDX tumor-volume shrinkage LMM p=0.0373 — architectural recipe of agentic hypothesis generation constrained by orthogonal statistical validation across biological scales generalises to CNS drug repurposing; bioRxiv v2 Liang Wang (Huazhong) claims general-purpose LLMs possess emergent capability for biological structural discovery — GPT-2 (124M) fine-tuned only on English-paraphrasing achieves 84% zero-shot protein homology detection, Qwen-3 approaches 100% + 75% precision on remote homology in <25% sequence-identity zone, chain-of-thought traces reveal "mental folding" and attention heads collapse language + biology representations into shared difference-detection manifold — direction matters for compute allocation in AI-for-biology arms race (specialized architecture vs domain adaptation with general models). Eleven items; neuropsychopharm + Parkinson's + fear extinction + sex-specific epigenetics + computational pharmacology + agentic-AI-in-biomedicine. Paper links: https://www.biorxiv.org/content/10.64898/2026.04.01.715822v2 ; https://www.biorxiv.org/content/10.64898/2026.07.02.736158v1 ; https://www.biorxiv.org/content/10.64898/2026.07.06.736678v1 ; https://www.biorxiv.org/content/10.64898/2026.07.04.736250v1 ; https://www.biorxiv.org/content/10.64898/2026.07.08.737296v1 ; https://www.biorxiv.org/content/10.64898/2026.03.26.714286v3 ; https://www.biorxiv.org/content/10.64898/2026.07.04.736516v1 ; https://www.biorxiv.org/content/10.64898/2026.07.03.736438v1 ; https://www.biorxiv.org/content/10.64898/2026.06.09.730783v1 ; https://www.biorxiv.org/content/10.64898/2026.07.05.736565v1 ; https://www.biorxiv.org/content/10.64898/2026.01.03.697478v2 https://www.biorxiv.org/content/10.64898/2026.04.01.715822v2 2026-07-15-receptor-and-reason Wed, 15 Jul 2026 12:00:00 +0000 1030 Daily roundup for July 15, 2026, eleven items. LEAD — Bryan Roth lab (UNC) revised preprint solves cryo-EM of eleven endogenous brain-derived mGluR2 assemblies from CRISPR-tagged mouse brain: mGluR2/3 heterodimers only in active-state complexes, endogenous ternary complexes with G-alpha-oA differ from recombinant structures — real blueprint for state-selective mGluR2 modulators after the pomaglumetad-era clinical-failure lane. Sato-Bigbee (VCU) — CSF nociceptin elevated in progressive-MS but not RRMS, NOP receptor antagonist rescues aged EAE (higher OLG:OPC, more myelin, less reactive astrocyte) — testable pharmacological lever with existing tool molecules. Vincent Laurent lab (UNSW; Maren coauthor) — cholinergic diagonal-band-to-infralimbic silencing during extinction enhances retrieval + prevents fear renewal; nicotinic (not muscarinic) IL blockade reproduces effect — brief nicotinic antagonism during exposure therapy against contextual relapse. Pierre-Eric Lutz (INCI CNRS) — sex-specific epigenetic responses at distinct genomic loci but converge on synapse-related pathways in ACC — shared-downstream drugs work both sexes, upstream epigenetic modifiers need sex-specific targeting. Rafiq Huda (Rutgers) — striatal astrocyte Ca driven by ACh and DA on distinct timescales, collapses in 6-OHDA PD — astrocytes as underexplored PD target via muscarinic access point. Aaron Wong (Delaware, v3) — 37 PD ON/OFF vs 39 controls — strong Bayes-factor null on reward sensitivity and effort-reward slope; positive on sensorimotor effort perception, resilient to L-DOPA — dopamine replacement is the wrong lever for effort-perception bradykinesia. Caldwell (Alabama) — DA-α-synuclein interaction is the molecular hinge uncoupling C. elegans DAergic degeneration from lifespan; DA-interaction-motif mutation makes neurons DA-insensitive; same interaction drives TFEB-dependent proteostatic lifespan extension — two-vector pharmacology (autophagy boost without DA-oxidative stress). Walter Reed — SO-tDCS during 2h restricted sleep protects PVT vigilance + improves T120 word recall across 46h sleep deprivation. TCS Research — QSP+RL+MCTS ipilimumab dosing matches highest-fixed-dose efficacy at lowest-fixed-dose cumulative exposure, drug-activity rate emerges as dominant learned determinant — direct CNS translation. Queen's Belfast Octopus — multi-agent LLM swarm + statistically bounded ML predictive envs prioritise IGF2 as CRC 5-FU resistance vulnerability, BH-FDR + PDX + human-survival cross-validated — architectural recipe generalises to CNS drug repurposing. Huazhong Liang Wang (v2) — GPT-2 fine-tuned only on English paraphrasing hits 84% zero-shot protein homology detection, Qwen-3 ~100% + 75% remote-homology precision, "mental folding" attention heads collapse language + biology representations into shared manifold. Paper links in show notes. false Daily roundup for July 14, 2026: bioRxiv Elowitz lab (Caltech) shows dopamine D1/D2 signal integration lives at the adenylyl cyclase, not at the receptor — multiplexed single-cell assay + mathematical modeling + brain-transcriptome analysis reveal that AC isoforms determine whether opposing D1/D2 inputs cancel, suppress, or amplify, and neurons that co-express opposing DA receptors preferentially express AC isoforms predicted to sustain signaling under multi-receptor activation; bioRxiv KIST (Korea) GeoMx spatial transcriptomics of dorsal striatum after chronic restraint stress — both D1 and D2 populations respond, but D2 downregulated genes cleanly resolve into glutamatergic-synapse / postsynaptic-organization / dendritic-spine categories with AMPA-trafficking + EPHB signaling attenuated (D1 gene sets do not show a comparable coherent pattern); bioRxiv v4 (posted 2026-07-13) optogenetically-identified VTA DA neurons in mice in operant sensory discrimination — during sub-criterion performance neurons predicted correct choice pre-outcome AND fired again on reward as if unexpected but were inhibited on punishment as if reward had been expected, consistent with actor-critic architecture where multiple Bayesian belief representations must be reconciled; bioRxiv Kravitz lab (WashU) — dorsomedial striatal NMDARs required for learning from previously rewarded actions + blocking striatal NMDARs abolished in-vivo Ca2+ dynamics but not APs (abstract-only); bioRxiv Maastricht (Mason/Ramaekers) — pooled placebo-controlled psilocybin 15 mg + 2C-B 20 mg + LSD 50 μg (N=72), sex-adjusted analysis shows female participants report significantly greater subjective intensity, reduced vigilance, and impaired control across all three drugs, with matched peak plasma concentrations (Cmax/AUC) — pharmacodynamic sex difference, not PK; bioRxiv Iowa group (Carver Charitable Trust-funded) — first patch-clamp characterization of hypothalamic A11 DA neurons (only descending DA input to spinal cord), most are ~5 Hz spontaneous pacemakers driven by subthreshold net inward current, and MOR activation slows pacemaking via K+ activation + reduced GABAergic input; bioRxiv Pittsburgh (Ellenberger et al) — scalable voluntary two-bottle drinking-in-the-dark oral opioid intake model in mice, oxycodone 0.1-1 mg/mL clean intake but minimal withdrawal; fentanyl 10-100 μg/mL produces dose- and session-duration-dependent naloxone-precipitated withdrawal + female mice show greater intake and stronger withdrawal at high escalating doses; bioRxiv APOE-focused Alzheimer's study — homozygous APOE3/APOE4 mice at 6 mo vs 24 mo across full cerebrovascular battery, age × APOE4 interaction drives brain-volume loss, neuroinflammation, and exaggerated cerebral-artery vasoconstriction to endothelin-1 (with altered ET-1 receptor + endothelin-converting enzyme expression) — mechanistic hinge suggests ET-1-antagonist repurposing (bosentan/macitentan/ambrisentan class) for APOE4-stratified LOAD; bioRxiv Cedersund group (Sweden) M4 drug discovery — first integration of multi-timescale + multi-level + mechanistic + multi-species preclinical data into a single fitted ODE framework, exemplified with GLP-1R agonist exenatide using 16 new + 6 prior human cell studies + 6 rat studies, successfully predicts 30-week human treatment outcome (χ² p>0.05) and shows which experimental system contributes which piece — template that generalizes to CNS pharm with iPSC-derived neurons + BBB-integrated MPS + rodent behavioral pharmacology; arXiv TU Munich/Charité benchmark of 32 frontier LLMs on hematologic-oncology clinical reasoning under uncertainty — models must proactively request data across 3 rounds, best model 68% accuracy, information utilization is strongest predictor of accuracy (R=0.69) but collapses from 57% early to 26% in final round when molecular + cytogenetic data are critical for treatment selection, reasoning-trace rubric 91% above threshold but decorrelated from accuracy, failure modes = search satisficing + anchoring + premature closure (novice-clinician biases); bioRxiv EcoXAI containerized multi-agent knowledge-graph biomedical discovery — proof-of-concept on Alzheimer's drug repurposing evaluated 103 candidates, 79 exceeded randomized baseline, top surviving candidate = CCR5 antagonist maraviroc with independent CCR5-AD literature support. Eleven items today; targets neuropsychopharm + computational pharmacology + agentic-AI-in-biomedicine per receptor-and-reason PROMPT.md. Paper links: https://www.biorxiv.org/content/10.64898/2026.07.10.737756v1 ; https://www.biorxiv.org/content/10.64898/2026.07.12.737112v1 ; https://www.biorxiv.org/content/10.1101/2024.05.06.592735v4 ; https://www.biorxiv.org/content/10.64898/2026.07.08.737179v1 ; https://www.biorxiv.org/content/10.64898/2026.07.08.737263v1 ; https://www.biorxiv.org/content/10.64898/2026.07.06.736848v1 ; https://www.biorxiv.org/content/10.64898/2026.07.07.736860v1 ; https://www.biorxiv.org/content/10.1101/2025.11.03.686224v1 ; https://arxiv.org/abs/2607.10275 ; https://www.biorxiv.org/content/10.64898/2026.07.08.737358v1 https://www.biorxiv.org/content/10.64898/2026.07.10.737756v1 2026-07-14-receptor-and-reason Tue, 14 Jul 2026 12:00:00 +0000 913 Daily roundup for July 14, 2026, ten items. LEAD — Elowitz lab (Caltech) shows dopamine D1/D2 signal integration lives at adenylyl cyclase, not receptor: AC-isoform expression determines whether opposing D1/D2 inputs cancel/suppress/amplify, and neurons that co-express opposing DA receptors preferentially express ACs predicted to sustain signaling; KIST spatial transcriptomics of chronically stressed dorsal striatum shows D2-specific attenuation of AMPA trafficking + EPHB signaling (D1 pattern less coherent); v4 preprint shows optogenetically-identified VTA DA neurons act as a critic signal on choice quality distinct from classical reward-prediction error; Kravitz lab abstract — striatal NMDARs abolish in-vivo Ca2+ dynamics without abolishing spiking + are required for updating action policy from prior reward; Ramaekers psilocybin/2C-B/LSD pool (N=72) shows female participants have larger subjective intensity + impaired control across all three drugs with matched Cmax/AUC — pharmacodynamic, not PK; Iowa first patch-clamp of hypothalamic A11 DA neurons (only descending DA to spinal cord), MOR slows ~5 Hz pacemaking via K+ + reduced GABA input; Pittsburgh voluntary two-bottle DID oral opioid model shows fentanyl 10-100 μg/mL produces dose- and duration-dependent naloxone withdrawal + female mice greater intake/withdrawal at high escalating doses (oxycodone minimal); Homozygous APOE3/E4 mice show age × APOE4 interaction drives brain-volume loss + neuroinflammation + exaggerated cerebral-artery vasoconstriction to ET-1 — mechanistic hinge suggests ET-1-antagonist repurposing for APOE4-stratified LOAD; Cedersund M4 drug discovery integrates cell/MPS/animal/human data into single fitted mechanistic model (exemplified with GLP-1R exenatide, 30-week human prediction) — template generalizes to CNS pharm; TU Munich/Charité 32-frontier-LLM hematologic-oncology benchmark, best model 68%, information-utilization is strongest predictor of accuracy but collapses from 57% to 26% by final round, novice-clinician cognitive-bias failure modes (satisficing/anchoring/premature closure); EcoXAI multi-agent knowledge-graph biomedical discovery — 103 AD-repurposing candidates evaluated, 79 exceeded baseline, top pick = CCR5 antagonist maraviroc with independent literature support. Eleven items; neuropsychopharm + computational pharmacology + agentic-AI-in-biomedicine. Paper links in show notes. false 2026-07-13-receptor-and-reason Mon, 13 Jul 2026 12:00:00 +0000 https://www.biorxiv.org/content/10.64898/2026.07.05.736652v1 Daily roundup for July 13, 2026: bioRxiv Sohal lab psilocybin selectively rescues cognitive flexibility only when deficit is caused by aberrant TH+ VTA→mPFC signaling (not by callosal PV inhibition disrupting interhemispheric gamma) + pathological signal is elevated mPFC→MD activity during 10-90 s post-error exploration + psilocybin attenuates it acutely AND 24 h later + potentiated MD-thalamic input to L5 SC neurons + abolished D2R+NMDA-dependent afterdepolarization = mechanism-specific precision-medicine target; bioRxiv Yale/Rutledge SSRI+EMA — 8 days citalopram vs placebo + 11 days smartphone EMA yielding real-world RPEs + citalopram amplifies reward-outcome AND RPE impact on positive affect (positive-RPE-selective) + attenuates negative-affect impact on reward expectation + RPE-mood coupling during dosing predicts washout PA = EMA as early on-target readout before symptom-scale change; bioRxiv Amilhon MRR→vHP 5-HT projection as sex-specific anxiety substrate — female 5-HTvHP neurons show heightened intrinsic excitability + elevated in-vivo recruitment + delayed disengagement from aversive contexts + ChETA activation increases anxiety-like behavior only in females; bioRxiv Liston lab ACC spatial-transcriptomics of stress motivational-impairment susceptibility vs resilience — enhanced inhibitory neuropeptide signaling + astrocytic contact signaling mark resilience vs ACC pyramidal hyperexcitability marks susceptibility (abstract-only); bioRxiv Mahajan+Seymour entropy-regularised RL reconciles aversive-PE vs action-PE views of tail-of-striatum dopamine — threat belief gates aversive value + default-policy term generates action-PE + both coexist in same TD error (abstract-only); bioRxiv UNC operant ethanol selectively reduces mGlu2/3 protein in NAc + LY379268 agonist reduces binge + LY487379 mGlu2 PAM ineffective = mGlu3 does the work (male-only); bioRxiv Jagust lab dopamine buffers preclinical AD DMN — amyloid impairs learning independent of tau + higher dorsolateral striatal DA rescues + rebalances DMN-frontostriatal connectivity = DA network rebalancing as cognitive resilience mechanism (abstract-only); bioRxiv SEA-AD DREAM Challenge snRNA-seq→AD neuropathology + first AI Agent Track + QWK=1.0 top score via donor-metadata leakage + scVIP transcriptomics-only reaches QWK≈0.65 = model-to-data privacy-preserving framework; bioRxiv MAPLE LLM-QSP calibration schemas + value-in-snippet + DOI resolution + code execution validators + 0/18 batch extractions passed first attempt + modeler still edits 65% of forward-model types = validator-heavy hybrid; arXiv MentalHospital virtual env for LLM SOAP-workflow psychiatric encounters + 1,193 EHR cases + 76 ICD-11 disorders + MentalEval QWK=0.944 with experts + strongest LLM trails clinicians by 37.28 pp on objective competence + mental status assessment is the bottleneck. 938 Daily roundup for July 13, 2026: bioRxiv Sohal lab psilocybin selectively rescues cognitive flexibility only when the deficit is caused by aberrant TH+ VTA→mPFC signaling (not by callosal PV inhibition) + attenuates elevated mPFC→MD post-error activity acutely AND 24 h later + potentiated MD-thalamic input to L5 SC neurons + abolished D2R+NMDA afterdepolarization = mechanism-specific precision-medicine target; bioRxiv Yale/Rutledge 8-day citalopram + 11-day smartphone EMA — citalopram amplifies reward-outcome + RPE impact on positive affect + attenuates negative-affect impact on reward expectation + RPE-mood coupling during dosing predicts washout PA = EMA as early on-target readout before symptom scales move; bioRxiv Amilhon MRR→vHP 5-HT projection is sex-specific anxiety substrate — female 5-HTvHP neurons show delayed disengagement from aversive contexts + ChETA activation increases anxiety-like behavior only in females; bioRxiv Liston lab ACC spatial-transcriptomics of stress motivational-impairment — resilience marked by inhibitory neuropeptide + astrocytic contact signaling vs susceptibility marked by ACC pyramidal hyperexcitability (abstract-only); bioRxiv Mahajan+Seymour entropy-regularised RL reconciles aversive-PE vs action-PE views of tail-of-striatum dopamine (abstract-only); bioRxiv UNC operant ethanol selectively reduces mGlu2/3 in NAc + LY379268 reduces binge + LY487379 PAM ineffective = mGlu3 does the work; bioRxiv Jagust lab dopamine buffers preclinical AD DMN — striatal DA synthesis capacity rescues amyloid-impaired learning + rebalances DMN-frontostriatal connectivity (abstract-only); bioRxiv SEA-AD DREAM Challenge snRNA-seq→AD neuropathology + first AI Agent Track + top team QWK=1.0 via donor-metadata leakage + scVIP transcriptomics-only reaches QWK≈0.65; bioRxiv MAPLE LLM-QSP calibration schemas — value-in-snippet + DOI resolution + code execution validators + 0/18 batch extractions passed first attempt = validator-heavy hybrid workflow; arXiv MentalHospital 1,193 EHR cases + 76 ICD-11 disorders + MentalEval QWK=0.944 with experts + strongest LLM trails clinicians by 37.28 pp + mental status assessment is bottleneck. false 2026-07-12-receptor-and-reason Sun, 12 Jul 2026 12:00:00 +0000 https://arxiv.org/abs/2607.07666 Daily roundup for July 12, 2026: arXiv Ensemble QSP multi-agent QSP framework — three-layer hierarchical memory keeps injected mid-term project state bounded (median 301 tokens, max 4050 across 104 runs) + five specialist worker agents under PI agents with physics-based checklists + robust autonomous PK/PD model selection + improved PK parameter recovery vs single-agent baselines + result quality stable across lower-cost and frontier LLMs + linguistically diverse prompts + domain-agnostic architecture = agentic scaffolding engineered for QSP; bioRxiv VTA acetylcholine is necessary for BOTH appetitive AND aversive states to become motivationally relevant + intracranial pharmacology dissects receptor mechanism = VTA ACh as valence-general motivational gate not affect-sign, implications for muscarinic/nicotinic modulators; bioRxiv calbindin upregulation in adult V1 inhibitory neurons reactivates juvenile ocular dominance plasticity + AAV-driven Calb1 manipulation causally determines OD plasticity extent = single-molecule reopener of cortical critical periods, psychedelic-plasticity adjacent, Juvian Inc spinout; bioRxiv 73-participant within-subject psychedelic vs everyday vs lab insight phenomenology + perceived meaning was strongest predictor of belief change + context itself NOT independently associated after phenomenology = REBUS-consistent, psychedelic efficacy runs through meaning-making not drug-context; bioRxiv sEEG + roving auditory oddball + mutual/co-information PE encoding + thalamic stable reduction under distraction + temporal cortex opposing learning trajectories + attention reorganizes redundant-vs-synergistic balance + Hebbian-learning network model reproduces = attention changes informational geometry not just gain; bioRxiv Balloon Analogue Risk Task in monkey + FEP decision model with OP-bias latent state + ML inverse-inference method + rule distinct from reward-dependent regulation = tool for time-varying OP-bias extraction, computational depression biomarker substrate; bioRxiv maternal separation + adult stress + PFC mitochondrial respirometry + MS animals resistant to negative bias shift + greater PFC mitochondrial respiratory capacity + uncoupling = ELS produces resilient recalibrated phenotype not damaged one, prefrontal bioenergetics as buffer + entry point for mito-modulating candidates; bioRxiv Trem2 R47H v2 reframe posted today walks back v1 increased-synaptic-density finding — with better variance modeling basal transmission + mEPSCs + evoked release + CA1/CA3 density unchanged + reframed R47H acts through context-dependent mechanisms activated by aging/inflammation/AD pathology = methodological reminder about early-postnatal density measurements; bioRxiv retrosplenial cortex engram tagging + longitudinal calcium imaging + engrams enriched for egocentric/boundary cells + recruited from pre-existing spatial scaffold not generated de novo + scaffold-silencing vs engram-silencing produce dissociable memory deficits = scaffold-engram architecture, two targets not one for memory-based interventions; arXiv four-analyst LLM workflow across 68 commercially compatible physiological corpora + 4 independent commercial LLM families + 695 markers deduplicated to 649 + threshold-bounds audit flagging 51 sanity violations = cross-model agreement + threshold sanity checks as lightweight audit layer, pattern transferable to pharmacology mining. 775 Daily roundup for July 12, 2026: arXiv Ensemble QSP multi-agent QSP framework — three-layer hierarchical memory keeps injected mid-term project state bounded (median 301 tokens, max 4050 across 104 runs) + five specialist worker agents under PI agents with physics-based checklists + improved PK parameter recovery vs single-agent baselines + result quality stable across lower-cost and frontier LLMs + domain-agnostic architecture = agentic scaffolding engineered for QSP; bioRxiv VTA acetylcholine necessary for BOTH appetitive AND aversive states to become motivationally relevant = valence-general motivational gate, implications for muscarinic/nicotinic modulators; bioRxiv calbindin upregulation in adult V1 inhibitory neurons reactivates juvenile ocular dominance plasticity + AAV manipulation causally determines OD plasticity extent = single-molecule reopener of cortical critical periods, psychedelic-plasticity adjacent; bioRxiv 73-participant within-subject psychedelic insight phenomenology + perceived meaning was strongest predictor of belief change + context NOT independently associated after phenomenology = REBUS-consistent, psychedelic efficacy runs through meaning-making; bioRxiv sEEG PE encoding + thalamic stable reduction under distraction + temporal cortex opposing learning trajectories + attention reorganizes redundant-vs-synergistic balance = attention changes informational geometry not just gain; bioRxiv BART monkey + FEP with OP-bias latent state + inverse-inference method + rule distinct from reward-dependent regulation = tool for time-varying OP-bias extraction; bioRxiv maternal separation + adult stress + MS animals resistant to negative bias shift + greater PFC mitochondrial capacity + uncoupling = ELS produces resilient recalibrated phenotype, prefrontal bioenergetics as buffer; bioRxiv Trem2 R47H v2 reframe walks back v1 density finding — basal transmission and density unchanged with better variance modeling = R47H acts through context-dependent mechanisms activated by aging/inflammation/AD pathology; bioRxiv retrosplenial cortex engrams enriched for egocentric/boundary cells recruited from pre-existing spatial scaffold + scaffold vs engram silencing produce dissociable memory deficits = scaffold-engram architecture with two targets; arXiv four-analyst LLM workflow across 68 physiological corpora + 695 markers deduplicated to 649 + threshold audit flagged 51 violations = cross-model agreement + sanity checks as lightweight audit layer, transferable to pharmacology mining. false 2026-07-11-receptor-and-reason Sat, 11 Jul 2026 12:00:00 +0000 https://arxiv.org/abs/2607.07753 Daily roundup for July 11, 2026: arXiv Transdiagnostic Space of Disorder-Like Phenotypes in RL Agents — 7 disorders as dose-controllable appraisal-signal knobs in appraisal-guided PPO + >1000 runs preregistered assays + graded monotone dose-response + 2D affective space where mania mirrors anxiety + knob-removal remits reward-distortion disorders but not avoidance (needs graded-exposure curriculum) + non-additive knob interactions + 3 knobs transfer to 3D MiniWorld with standard convnet = disciplined induce-then-treat framework for computational psychiatry; bioRxiv SCZ cortical organoids + 24-week clozapine + multiomics reveals SCZ-specific metabolism-independent alternative splicing program in glutamatergic neurons recapitulated in primary tissue + capturing SCZ-risk genetic signal = clozapine mechanism nameable + cleaner TRS target; bioRxiv patient-derived tau-seeded human neuronal chimeras — hPSC-NPC transplant + adult human tau (6 isoforms 1:1 3R:4R) + AD-brain P-tau seeds produce PHFs + NFTs + spread + plasma pTau-217 + memory deficits + 75% AT8+ cells human + WT mice no tangles + snRNA-seq human neurons have higher tau-uptake genes + widespread synaptic suppression while mouse resilient = human-specific vulnerability substrate for tau therapeutics beyond P301L; bioRxiv human MTL single-neuron intracranial recordings — decision-predictive amygdala/hippocampus + delay-encoding entorhinal/hippocampus + prospective temporal coding + impulsive individuals show diminished prospective coding + late decision-prediction = impulsivity as compressed prospective time horizon at population level; bioRxiv zebra finch striatal dopamine at learned motif boundaries — dLight fiber photometry + dip-then-peak at motif onset + developmental backward shift from offset to onset (TD-like shape without external reward) + optogenetic inhibition at motif onset produces adult song deterioration = phasic dopamine does self-referential TD bookkeeping on internally-scored natural skill; bioRxiv astrocyte Mpc2 deletion — motor deficits + hyperexcitability + seizure-associated lethality + failed pyruvate-to-alanine bypass + impaired TCA + imbalanced glutamate/glutamine/GABA = astrocyte mitochondrial pyruvate import is anaplerotic E-I gate not bioenergetic, opens MPC as anti-epileptic target space; bioRxiv DSP-4 noradrenergic depletion + Sholl analysis reveals reduced hippocampal astrocyte branching complexity + isoproterenol fails to rescue morphology but increases SOX9+ density = LC tone required for astrocyte arbor maintenance, β-adrenergic alone insufficient, relevant to LC neurodegeneration in AD/depression; bioRxiv IGF1 tripeptide (trofinetide mechanism) targets Rett astrocytes — degrades elevated IGFBP2 via proteasome + restores astrocyte mitochondrial function + enhances neuronal PI3K/Akt = trofinetide efficacy proximally from astrocyte IGFBP2 knockdown not direct neuronal action; arXiv DrugGen 2 disease-aware generative LM — GPT-2 + disease-ontology conditioning + GRPO on 5 diabetic nephropathy targets + better predicted affinities than DrugGPT/DrugGen baselines + top docking scores exceeding enalapril = disease-context conditioning as underused generative-drug-design signal (predicted docking caveat); bioRxiv ORL-1 (nociceptin receptor) interacts with Cav1.2 L-type channel — co-IP with Cavα1c + Cavα2δ-1 + reduced peak current without biophysical changes + acute nociceptin no further effect (constitutive agonist-independent) + decreased plasma membrane expression via disrupted forward trafficking = new complication for hippocampal ORL-1 programs in cognition/depression. 967 Daily roundup for July 11, 2026: arXiv Transdiagnostic Space of Disorder-Like Phenotypes in RL Agents — 7 disorders as dose-controllable appraisal-signal knobs in appraisal-guided PPO + >1000 runs preregistered assays + graded monotone dose-response + 2D affective space where mania mirrors anxiety + knob-removal remits reward-distortion disorders but not avoidance (needs graded-exposure curriculum) + non-additive knob interactions + 3 knobs transfer to 3D MiniWorld with standard convnet = disciplined induce-then-treat framework for computational psychiatry; bioRxiv SCZ cortical organoids + 24-week clozapine + multiomics reveals SCZ-specific metabolism-independent alternative splicing program in glutamatergic neurons + captured SCZ-risk signal = clozapine mechanism nameable; bioRxiv patient-derived tau-seeded human neuronal chimeras produce mature AD tau pathology + spread + plasma pTau-217 + memory deficits in human neurons + 75% AT8+ cells human + WT mice no tangles + higher tau-uptake gene expression + synaptic suppression in human neurons while mouse resilient = human-specific vulnerability substrate for tau therapeutics; bioRxiv human MTL single-neuron intracranial recordings show impulsive individuals have diminished prospective temporal coding and late decision-prediction = compressed prospective time horizon; bioRxiv zebra finch striatal dopamine at motif onsets developmentally backward-shifts from motif offset + optogenetic inhibition at onset produces adult song deterioration = phasic dopamine does self-referential TD bookkeeping on internally-scored natural skill; bioRxiv astrocyte Mpc2 deletion produces seizure-associated lethality via failed anaplerosis + imbalanced neurotransmitter pools = astrocyte MPC as anti-epileptic target; bioRxiv DSP-4 noradrenergic depletion reduces hippocampal astrocyte branching + isoproterenol doesn't rescue = LC tone required for astrocyte arbor maintenance; bioRxiv IGF1 tripeptide (trofinetide mechanism) rescues Rett astrocytes by degrading IGFBP2 not by direct neuronal action; arXiv DrugGen 2 disease-aware generative LM with GRPO shows better predicted affinities than baselines on 5 diabetic nephropathy targets; bioRxiv ORL-1 (nociceptin receptor) constitutively reduces Cav1.2 surface expression via disrupted forward trafficking = complication for hippocampal ORL-1 cognition/depression programs. false 2026-07-10-receptor-and-reason Fri, 10 Jul 2026 12:00:00 +0000 https://www.biorxiv.org/content/10.64898/2026.07.02.736180v1 Daily roundup for July 10, 2026: bioRxiv SEA-AD DREAM Challenge — first AI Agent Track in a DREAM Challenge for AD neuropathology prediction from snRNA-seq: best QWK = 1.0 for overall ADNC + best CCC = 0.48 for quantitative amyloid/phospho-tau burden + post-hoc perturbation reveals categorical predictions rode donor-metadata leakage while quantitative burden was transcriptome-robust = leakage warning for deeply phenotyped human brain challenges; bioRxiv endogenous mGluR2 cryo-EM from mouse brain resolves 11 assemblies + homodimers + mGluR2/mGluR3 heterodimers (active-state only) + G-alpha-o ternary complexes differ from recombinant = native structural framework for mGluR2/3 modulator design; bioRxiv BBBP-Atlas cross-modal BBB permeability model unifying small molecules + peptides (OmniBBBP 9316 + 902), AUC 0.91, MCC 0.70 external, beats LightBBB by 6 MCC points; bioRxiv nociceptin/orphanin FQ elevated in progressive-MS CSF + NOR antagonist reverses old-mouse EAE clinical scoring + increases myelination = NOR antagonists as remyelination target for progressive MS; bioRxiv opposing GIPR brainstem circuits — AP GIPR neurons permit hyperphagia + NTS GIPR neurons anorectic + GIPR antagonism appetite suppression requires AP Gipr = explains why GIPR agonists AND antagonists both work with GLP-1; bioRxiv rDLPFC vs rPPC cTBS in alcohol approach — DLPFC sustains regulatory signal + PPC gates behavioral expression = PPC as distinct TMS node for AUD; bioRxiv seizure susceptibility scales with continuous-attractor computation — mEC drives acute epileptiform discharges with smoother state trajectories than CA3 + entorhinal silencing abolishes = network geometry biases epileptogenicity; arXiv 187k ChatGPT histories from 766 PHQ-8 participants show high-PHQ late-night + absolutist + first-person-singular usage + high-disclosure without professional redirection + AUROC 0.591 insufficient for screening = LLMs as informal support infrastructure; arXiv MentalHospital S.O.A.P. virtual environment + MentalEval QWK 0.944 vs clinicians shows strongest LLM trails clinicians by 37.28 pp with mental-status assessment as bottleneck; bioRxiv IxIDN + ORDON positive/negative benchmarks for rare-disease drug repurposing (574 + 793 diseases) enabling decision-oriented cross-evidence generalization test. 893 Daily roundup for July 10, 2026: bioRxiv SEA-AD DREAM Challenge — first AI Agent Track in a DREAM Challenge for AD neuropathology prediction from snRNA-seq: best QWK = 1.0 for overall ADNC + best CCC = 0.48 for quantitative amyloid/phospho-tau burden + post-hoc perturbation reveals categorical predictions rode donor-metadata leakage while quantitative burden was transcriptome-robust = leakage warning for deeply phenotyped human brain challenges; bioRxiv endogenous mGluR2 cryo-EM from mouse brain resolves 11 assemblies + homodimers + mGluR2/mGluR3 heterodimers (active-state only) + G-alpha-o ternary complexes differ from recombinant = native structural framework for mGluR2/3 modulator design; bioRxiv BBBP-Atlas cross-modal BBB permeability model unifying small molecules + peptides (OmniBBBP 9316 + 902), AUC 0.91, MCC 0.70 external, beats LightBBB by 6 MCC points; bioRxiv nociceptin/orphanin FQ elevated in progressive-MS CSF + NOR antagonist reverses old-mouse EAE clinical scoring + increases myelination = NOR antagonists as remyelination target for progressive MS; bioRxiv opposing GIPR brainstem circuits — AP GIPR neurons permit hyperphagia + NTS GIPR neurons anorectic + GIPR antagonism appetite suppression requires AP Gipr = explains why GIPR agonists AND antagonists both work with GLP-1; bioRxiv rDLPFC vs rPPC cTBS in alcohol approach — DLPFC sustains regulatory signal + PPC gates behavioral expression = PPC as distinct TMS node for AUD; bioRxiv seizure susceptibility scales with continuous-attractor computation — mEC drives acute epileptiform discharges with smoother state trajectories than CA3 + entorhinal silencing abolishes = network geometry biases epileptogenicity; arXiv 187k ChatGPT histories from 766 PHQ-8 participants show high-PHQ late-night + absolutist + first-person-singular usage + high-disclosure without professional redirection + AUROC 0.591 insufficient for screening = LLMs as informal support infrastructure; arXiv MentalHospital S.O.A.P. virtual environment + MentalEval QWK 0.944 vs clinicians shows strongest LLM trails clinicians by 37.28 pp with mental-status assessment as bottleneck; bioRxiv IxIDN + ORDON positive/negative benchmarks for rare-disease drug repurposing (574 + 793 diseases) enabling decision-oriented cross-evidence generalization test. false 2026-07-08-receptor-and-reason Wed, 08 Jul 2026 09:00:00 +0000 https://www.biorxiv.org/content/10.64898/2026.07.01.735554v1 Daily roundup for July 8, 2026: bioRxiv Scherrer/Schnitzer motor cortex TMS &rarr; RVM endogenous opioid/NMDA plasticity: L5 pyramidal &rarr; RVM circuit shifts ON/OFF neuron balance + NMDA-dep + MOR-dep by endogenous enkephalins jointly mediate 1-2 week analgesia post-TMS + enkephalinase inhibition in RVM during TMS amplifies analgesia amplitude + duration + human chronic-pain data supports opioid amplification of MCS = mechanistic rationale for neurostimulation + enkephalinase-inhibitor combo; bioRxiv non-AgRP/TH Arc GABA&rarr;PVH CRH circuit gates HPA scalability by stressor intensity = therapeutic target for PTSD/anxiety; bioRxiv Fang/Piray volatility vs stochasticity double dissociation: stochasticity-blind &rarr; internalizing (anxiety/depression); volatility-blind &rarr; externalizing (addiction/compulsivity); bioRxiv Tank/Brody HPC engrams configure mPFC context in ~100 ms = memory as prefrontal circuit configurator; bioRxiv Kuner lab IL cortex dual-ensemble: action-outcome predictor + prediction correction reconciles conflicting IL manipulation results; bioRxiv Baunez STN photoinhibition blocks familiarity/hierarchy effects on social motivation + reduces high-effort social motivation = STN social-value circuit; bioRxiv Niell lab DOI psychedelic: increases active gaze shifts + reduces evoked V1 response per shift + behavior-coupled responses selectively affected = corollary-discharge suppression failure; bioRxiv Wu lab astrocyte pannexin-1&rarr;microglia P2Y12 purinergic shielding during neuronal hypoactivity &rarr; circuit rebound; bioRxiv AstraZeneca/Cambridge TfR1-8D3 cryo-EM: pair-level cross-linking avidity + pH-sensitive variants grade brain penetration = structural rules for BBB shuttles; bioRxiv Lein lab SEA-AD 10-region 7M-nuclei AD atlas: Sst/interneurons lost earliest + V1C L4 IT unexpectedly vulnerable + multi-agent AI nominates NMDA hyperexcitability as convergent vulnerability 997 Daily roundup for July 8, 2026: bioRxiv Scherrer/Schnitzer motor cortex TMS &rarr; RVM endogenous opioid/NMDA plasticity: L5 pyramidal &rarr; RVM circuit shifts ON/OFF neuron balance + NMDA-dep + MOR-dep by endogenous enkephalins jointly mediate 1-2 week analgesia post-TMS + enkephalinase inhibition in RVM during TMS amplifies analgesia amplitude + duration + human chronic-pain data supports opioid amplification of MCS = mechanistic rationale for neurostimulation + enkephalinase-inhibitor combo; bioRxiv non-AgRP/TH Arc GABA&rarr;PVH CRH circuit gates HPA scalability by stressor intensity = therapeutic target for PTSD/anxiety; bioRxiv Fang/Piray volatility vs stochasticity double dissociation: stochasticity-blind &rarr; internalizing (anxiety/depression); volatility-blind &rarr; externalizing (addiction/compulsivity); bioRxiv Tank/Brody HPC engrams configure mPFC context in ~100 ms = memory as prefrontal circuit configurator; bioRxiv Kuner lab IL cortex dual-ensemble: action-outcome predictor + prediction correction reconciles conflicting IL manipulation results; bioRxiv Baunez STN photoinhibition blocks familiarity/hierarchy effects on social motivation + reduces high-effort social motivation = STN social-value circuit; bioRxiv Niell lab DOI psychedelic: increases active gaze shifts + reduces evoked V1 response per shift = corollary-discharge suppression failure; bioRxiv Wu lab astrocyte pannexin-1&rarr;microglia P2Y12 purinergic shielding during hypoactivity &rarr; circuit rebound; bioRxiv AstraZeneca/Cambridge TfR1-8D3 cryo-EM structural rules for BBB shuttles; bioRxiv Lein lab SEA-AD 10-region 7M-nuclei AD atlas: NMDA hyperexcitability as convergent vulnerability false Receptor & Reason for July 7, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Ten items today, opening on a spinal excitatory KOR paper out of the Sheahan lab that completes the pair with yesterday's aPVT-KOR dysphoria paper — chemogenetic and pharmacological manipulation of KOR-expressing spinal neurons show they are both necessary and sufficient for acute pain, and these neurons are excitatory and coexpress substance P precursor Tac1 in mouse and human. The circuit logic — dysphoria central, analgesia spinal — makes the peripherally-restricted and biased KOR analgesic programs (nalfurafine, difelikefalin) a rational separation of the two arms, not a lucky accident. Then a genomic-architecture paper cross-referencing schizophrenia and multiple sclerosis identifying a shared structural fracture at the HCAR tandem regulatory domain on chromosome 12q24. High-resolution eQTL in purified immune lineages shows the shared risk allele drives a peripheral-macrophage-specific transcriptomic collapse of HCAR1 (the un-drugged lactate sensor), not HCAR2 or HCAR3, and the locus is unequivocally not associated with Crohn's, lupus, rheumatoid arthritis, or psoriasis — a highly specific neuroimmune ignition switch, not general inflammation. Macrophages become "lactate blind", unable to engage the cAMP-suppressing negative feedback that halts immune proliferation. HCAR1 agonists exist as research tools; the paper is an argument that clinical development is warranted, complementary to the HCAR2/dimethyl fumarate arm already validated in MS. Then a dACC glutamate MRS study in fifty trauma-exposed adolescents showing that trauma exposure degrades glutamatergic recruitment during inhibitory control specifically when negative emotional distractors are present. Dynamic 1H fMRS is starting to work as a within-subject task-block-timescale readout of neurotransmitter levels — this is a candidate stratification biomarker for adolescent anxiety trials and a target-engagement readout for group-II mGluR modulators and ketamine-adjacent psychiatric drug development. Then a chromatin-architecture paper reframing the autism gene POGZ. E13.5 mouse cortex IP-MS identifies the H3K9 methyltransferases G9a and GLP as POGZ's principal in vivo interactors. Loss of POGZ produces bidirectional megabase-scale redistribution of H3K9me3, with newly-repressed regions repositioning to the nuclear lamina, strengthening B-compartment scores, eroding TAD boundaries, weakening insulation, and losing CTCF occupancy. Nascent transcription of neuronal genes within those domains collapses. POGZ is a G9a/GLP-associated boundary keeper protecting neurodevelopmental gene domains from heterochromatinization — one of the cleanest mechanistic explanations to date of how a specific ASD gene disrupts brain development at the transcription-of-neuronal-programs step, and a specific pediatric-neurodevelopmental therapeutic hypothesis for G9a-GLP inhibitors (UNC-0642 as tool compound; oncology and sickle cell development ongoing). Then a large CSF proteomic aging clock from more than ten thousand CSF proteomes across multiple cohorts and platforms. A 249-protein clock predicts age and generalizes across independent datasets; brain-age acceleration is increased across diverse neurological diseases, associated with BBB dysfunction, and predictive of longitudinal cognitive decline, neuroimaging progression, and dementia conversion. A simplified 30-protein panel retains most prognostic performance (clinical-assay tractable). Biology resolves two opposing programs — pro-aging activation of immune/vascular/ECM/coagulation pathways (CHI3L1, CD14, VWF, LRG1, LTBP2) and anti-aging collapse of neuronal/synaptic/axonal/structural-maintenance programs (NPTX2, COL1A2, NID1, CDH8, PENK). Target-engagement handles for the two arms separately and a single-number composite Phase 2 endpoint for CNS neurodegeneration drug development. Then a receptor-engineering item — Trinity College Dublin phage-display against endogenous CB2R conformations on human T cells produces ten constrained peptide antagonists all at IC50 5-10 nM on beta-arrestin recruitment and pERK, with SLKC-09 at 5.4 nM inhibiting gut homing of CD4/CD8 naive/effector/memory T cells in a mouse ileitis model. First low-nanomolar biologic antagonism of a cannabinoid GPCR via constrained peptides (neither small molecule nor full antibody), a modality-generalization signal that the CGRP-mAb story applies to cheaper/smaller/better-penetrating peptide antagonists against CNS-relevant GPCRs (orexin, kappa opioid, trace amine-associated receptors). Then a Kv1.3 allosteric pocket paper reporting MPiN, a state-dependent inhibitor engaging an extracellular pocket framed by the PP1/PP2 turret loops, the pore helix, and the outer S5/S6 helices, acting through bidirectional pore-to-sensor coupling that simultaneously constricts the selectivity filter and facilitates voltage sensor activation. Subtype selectivity is dictated by peripheral G427 and H451 that don't contact the ligand at all but define pocket geometry — a geometry-driven, non-contact selectivity mechanism as a general design principle for paralog selectivity in ion-channel families. In vivo efficacy in mouse psoriasis. Kv1 paralog selectivity is prerequisite for CNS-safe use of Kv-channel modulators (Kv1.1 = Isaac's syndrome, episodic ataxias). Then three agentic-AI items to close. BoltzMol-1 (Boltz PBC) is an optimized Boltz-2 co-folding model packaged into a prospective virtual-screening pipeline with an ADMET-model triage layer (kinetic solubility logS, logD, Caco-2 permeability) inline before synthesis or purchase. On ten targets — most without representation in the underlying affinity training data — functional actives or binders on six out of ten with modest experimental budgets of 28-96 compounds per target. Not a spectacular hit rate, but evidence that co-folding-model-based VS is now workable for prospective use on out-of-distribution targets, and physchem-aware ranking is relevant for BBB-penetrant CNS hit discovery. Then PRECEDE — a precedent-guided co-scientist for side-effect-aware drug redesign. LLM orchestrator with evidence-grounded reasoning over drug-side-effect associations, biomedical knowledge graphs, and medicinal-chemistry precedent cases of safety-driven optimization, with explicit policies and human-review checkpoints. The value is that precedent retrieval is first-class evidence — encoding the inductive move that medicinal chemists actually use ("compound X had the same off-target liability and was fixed by adding a methyl at position 4") rather than treating redesign as isolated generative sampling. Workflow shape rather than benchmark numbers is the take. Closes on EEG-SpikeAgent — an LLM proposes deterministic signal-processing feature modules one at a time, executes code on EEG, and a gradient-boosted-tree classifier evaluates. On the VEPISET public 29-channel scalp EEG dataset (2,516 discharge and 22,933 non-discharge 4-second epochs), auto-generated features achieve AUROC 0.935. Auto-generated features expressed as auditable, inspectable code — a clinically-acceptable path for automated EEG interpretation where explainability has blocked deployment more than raw performance, and a generalizable closed-loop pattern for any biosignal domain that trades interpretability against performance. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-07-receptor-and-reason Tue, 07 Jul 2026 12:00:00 +0000 966 Daily roundup for July 7, 2026: bioRxiv Sheahan lab spinal excitatory KOR neurons gate acute pain (mouse + human) — pharmacological inhibition of spinal KOR-expressing neurons blocks nocifensive behaviors + chemogenetic activation of same neurons drives nocifensive behaviors even w/o noxious input = both necessary + sufficient for acute pain + excitatory KOR neurons coexpress pain-promoting neuropeptides (Tac1 = substance P precursor) in mouse + human spinal cord = KOR agonism reduces pain-neuropeptide release from translationally-relevant excitatory population + concrete spinal-selective analgesic MoA for nalfurafine/difelikefalin peripherally-restricted KOR programs (companion to yesterday's aPVT-dysphoria paper — dysphoria central + analgesia spinal = separation is achievable + not accidental); bioRxiv HCAR1 as un-drugged neuroimmune checkpoint shared between SCZ + MS — cross-referencing genomic architectures identifies shared structural fracture at HCAR tandem regulatory domain (12q24) + eQTL in purified immune lineages shows shared risk allele drives peripheral-macrophage-specific transcriptomic collapse of HCAR1 (not HCAR2 not HCAR3) + locus not associated w/ Crohn's + lupus + RA + psoriasis = highly specific neuroimmune ignition switch not general inflammation + macrophages become "lactate blind" unable to engage cAMP-suppressing negative feedback = HCAR1 agonists (research tools only) as therapeutic checkpoint for SCZ + MS complementary to HCAR2/dimethyl fumarate; bioRxiv trauma-exposed adolescents reduced dACC glutamate modulation during inhibitory control w/ negative emotional stimuli — 50 adolescents + 1H fMRS = trauma degrades glutamatergic recruitment during effortful cognitive control when negative emotional distractors present + dynamic MRS as within-subject task-block-timescale neurotransmitter readout + TE-by-emotional-stimulus interaction on dACC Glu as biomarker for adolescent anxiety trials + group-II mGluR + ketamine-adjacent psychiatric drug development; bioRxiv POGZ safeguards neuronal chromatin — E13.5 cortex IP-MS identifies G9a/GLP H3K9 methyltransferases as POGZ's principal interactors + POGZ loss drives bidirectional megabase-scale H3K9me3 redistribution + repositions to nuclear lamina + strengthens B-compartment + erodes TAD boundaries + reduces CTCF + collapses neuronal-gene nascent transcription = POGZ as G9a/GLP-associated boundary keeper protecting neurodevelopmental gene domains from heterochromatinization + White-Sutton/ASD gene at chromatin-architecture level + G9a-GLP inhibitors (UNC-0642) as testable pediatric neurodevelopmental hypothesis; bioRxiv CSF proteomic aging clock across 10,000+ proteomes — 249-protein clock predicts age + generalizes + brain-age acceleration across neurological diseases + BBB dysfunction + predicts longitudinal decline + neuroimaging progression + dementia conversion + 30-protein simplified panel clinical-tractable + 2 opposing programs = pro-aging activation of immune + vascular + ECM + coagulation (CHI3L1 + CD14 + VWF + LRG1 + LTBP2) vs anti-aging collapse of neuronal + synaptic + axonal (NPTX2 + COL1A2 + NID1 + CDH8 + PENK) = target-engagement handles for 2 arms + composite Phase-2 endpoint for CNS neurodegeneration drug development; bioRxiv Trinity College Dublin CB2R constrained-peptide antagonist library (O'Connell) — phage display against endogenous CB2R conformations on human T cells + 10 protein candidates IC50 5-10 nM + SLKC-09 at 5.4 nM inhibits gut homing of CD4/CD8 T cells in mouse ileitis + first low-nM biologic antagonism of cannabinoid GPCR via constrained peptides = modality-generalization signal for CGRP-mAb-class biologic receptor blockade at cheaper/smaller/better-penetrating scale + should propagate to orexin/KOR/TAAR CNS GPCRs; bioRxiv Kv1.3 novel extracellular allosteric pocket + MPiN inhibitor (Hunan) — pocket framed by PP1/PP2 turret loops + pore helix + outer S5/S6 helices + bidirectional pore-to-sensor coupling constricts selectivity filter + facilitates voltage sensor activation + subtype selectivity via peripheral G427 + H451 that don't contact ligand (define geometry) = geometry-driven non-contact selectivity mechanism as general design principle for ion-channel paralog selectivity + in vivo mouse psoriasis efficacy + Kv1 paralog selectivity prerequisite for CNS-safe Kv-channel modulator use (Kv1.1 = Isaac's + episodic ataxias); bioRxiv BoltzMol-1 hit discovery pipeline (Boltz PBC) — optimized Boltz-2 co-folding + 10 targets (most w/o affinity training data representation) + functional actives or binders on 6/10 w/ 28-96 compounds/target + ADMET triage (logS + logD + Caco-2 permeability) as early developability filter = co-folding-based VS workable prospectively on out-of-distribution targets + physchem-aware ranking relevant for BBB-penetrant CNS hit discovery; arXiv PRECEDE precedent-guided co-scientist for side-effect-aware drug redesign — LLM orchestrator + evidence-grounded reasoning over drug-side-effect associations + biomedical KGs + medicinal-chemistry precedent cases + explicit policies + human-review checkpoints = precedent retrieval as first-class evidence encoding "compound X had same liability + was fixed by adding methyl at position 4" inductive move medicinal chemists use + workflow shape rather than benchmark numbers is the take; arXiv EEG-SpikeAgent agentic closed-loop program synthesis for automated EEG spike detection — LLM proposes deterministic signal-processing feature modules + executes code + gradient-boosted classifier evaluates + VEPISET 29-channel 4-sec-epoch scalp EEG w/ 5-fold CV = AUROC 0.935 + auto-generated features expressed as auditable inspectable code = clinically-acceptable path for automated EEG interpretation where explainability has blocked deployment + generalizable closed-loop pattern for biosignal domains trading interpretability against raw performance. false Receptor & Reason for July 6, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Twelve items today, opening on receptor pharmacology with a run of clean circuit-level manipulations up front. The lead is a paper from the Liu-Chen lab on kappa opioid receptors in the anterior paraventricular nucleus of the thalamus. Conditional AAV-Cre knockdown in Oprk1-floxed mice with tritiated U69,593 receptor autoradiography confirming loss shows that aPVT KOR cKD reduces anxiety-like EPM behavior, cue-induced fear freezing, and naloxone-precipitated morphine withdrawal jumps in both sexes; attenuates KOR agonist U50,488H-induced conditioned place aversion in males only; and spares U50,488H's visceral analgesic and antipruritic effects. That means aPVT KOR is dispensable for the desirable therapeutic effects but essential for the aversive dysphoric ones. Widespread axonal collateralization from aPVT to nucleus accumbens, central amygdala, BNST, prefrontal cortex, and reticular thalamus makes aPVT a single node broadcasting to nearly every stress-and-reward-relevant region — a specific circuit-level rationale for the peripherally-restricted and biased KOR agonist programs (nalfurafine, difelikefalin) working better than expected on pain and pruritus, and a target for centrally-restricted KOR modulators in opioid use disorder. Then a paper from the West lab at Duke on VTA transcriptional plasticity in cocaine seeking. In vivo dCas9/CRISPRi demonstrates a causal role for BDNF transcription in VTA specifically during abstinence — not during exposure — for the incubation of cocaine seeking. snRNA-seq identifies a gene regulation program induced selectively by the prolonged absence of cocaine in dopaminergic and glutamatergic VTA neurons — incubation-of-craving is a transcriptionally-driven active adaptation, not passive memory decay, and any pharmacotherapy for the abstinence window has to engage this program. Then PACAPergic BAC (bed nucleus of the anterior commissure) engages BLA and anterodorsal thalamic networks during looming-threat labyrinth memory. Chemogenetic BAC silencing impairs shelter-directed escape and increases freezing; selective PACAP deletion blocks looming-induced Fos in anterodorsal thalamus specifically. A new PACAP forebrain hub separating the neuron-firing role from the peptide-release role, and a specific circuit substrate for PAC1 modulator PTSD development beyond BNST and amygdala. Then two callback papers. Yesterday we covered a paper showing that neonatal OTR blockade recapitulates maternal-separation phenotypes. Today, the flip side — early postnatal oxytocin administration (P2-P12 daily IP) completely prevents male-specific spatial memory deficits induced by maternal separation, restores CA1 Schaffer collateral LTP, and normalizes neuroinflammatory + GABAergic + synaptic transcript expression in dorsal CA1. Bidirectional oxytocinergic postnatal-window sensitivity with sex + domain specificity — male-specific spatial LTP rescue here vs female-specific object location in yesterday's dOVT paper. Yesterday's dOVT paper also flagged NKCC1/KCC2 chloride homeostasis; today, De Koninck lab paper on how KCC2 is actually regulated. BDNF-TrkB signaling via ryanodine-dependent intracellular Ca release modulates KCC2 function without changing expression; NMDAR activation via extracellular Ca influx internalizes KCC2; calpain-mediated degradation requires voltage-gated Ca influx. Hierarchical scaling from function → internalization → degradation with stimulus strength and time — KCC2 activators for pain, epilepsy, autism should be evaluated on the three arms separately. Then a tauopathy pair. PS19 hippocampal metaplasticity — laminar-specific inflammatory markers (C3 in stratum oriens; CD68 in oriens + lacunosum-moleculare) coincide with occluded transregional metaplasticity where the tauopathic hippocampus is stuck in a state inhibiting future LTP. Complementing that, a v3 super-resolution imaging paper mapping native surface AMPARs in mouse brain slices at <10 nm precision — PS19 disrupts AMPAR clustering + destabilizes nanodomains in hippocampus but not cortex, before overt neurodegeneration, giving a candidate very early biomarker at the synaptic-organization level. Then a comparative vessel-enriched transcriptomics paper across AD, FTD, and HD. Partially conserved vascular dysfunction signature across all three, plus disease-specific endothelial + pericyte + perivascular changes. Two molecularly distinct human pericyte subtypes; matrix-type pericyte fraction consistently reduced across all three diseases — a rare pan-neurodegenerative vascular target. Endothelial remodeling most prominent in capillary + venous segments. Blood-brain-barrier dysregulation as a disease-specific PK modifier for CNS drugs. Then three computational-pharmacology items. First, a class A GPCR paper using coevolution + ML on 447 GPCRdb structures to derive a sequence-independent 2D collective variable of activation, then biasing MD along it to drive 20 receptors across 20 subfamilies through full activation transitions within a microsecond of simulation time. Transfers to orphan GPR183 not in the training set. Converges β2-adrenergic receptor activation free-energy surfaces in presence of different ligand modalities. An enabling tool for structural + dynamical studies of hundreds of orphan/understudied GPCRs. Second, a benchmark study on dynamic-pocket detection across MD ensembles — Fpocket, SiteFerret, AutoSite, SiteHound/EasyMIFs, P2Rank, DeepPocket applied to GLUT1 + aldose reductase (cryptic pocket reference) — energy-based methods more sensitive to transient cryptic regions, geometry-based methods rely on pre-formed cavities. Consensus-oriented framework for identifying druggable cryptic pockets in dynamic systems (GPCR allosteric sites, ion channel state-dependent modulators, kinase back pockets). Third, a fresh-yesterday preprint on target-hypothesis generation for orphan natural products. GNN ligand encoder + ESM-2 protein encoder + bidirectional cross-attention DTI + frequent-hitter bias correction via losartan benchmark + control-anchored docking with matched positive + negative controls nominates orphan GPCR GPR35 as candidate target of ligiamycin A — docks -8.1 kcal/mol at GPR35 agonist pocket vs known agonist zaprinast -8.3, non-binder floor -5.5. FFAR1 fails the negative-control test. Control-anchored DTI + debiased ranking as more principled natural-product target-hypothesis workflow vs naive top-scoring-hit pipelines. Closes on AgentODE — an LLM proposes candidate ODE structures and a tool-augmented inference agent iteratively refines parameter distributions from population-level summary statistics alone. Evaluated on RDEB (231 observations, 46 patients) and two other clinical datasets — summary-statistics-only baseline outperforms individual-level baselines on mechanistic plausibility despite worse predictive performance, suggesting the privacy constraint acts as an inductive bias toward mechanism. New direction for rare-disease pharmacometrics/QSP. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-06-receptor-and-reason Mon, 06 Jul 2026 12:00:00 +0000 1038 Daily roundup for July 6, 2026: bioRxiv aPVT KOR conditional knockdown (Liu-Chen lab) — cKD reduces EPM anxiety + cue-induced fear freezing + naloxone-precipitated morphine withdrawal jumps in both sexes + attenuates U50,488H CPA in males only + spares U50,488H analgesic + antipruritic effects = aPVT KOR is dispensable for therapeutic + essential for dysphoric arm + widespread axonal collateralization to NAc + CeA + BNST + PFC + reticular thalamus = rationale for nalfurafine + difelikefalin biased/peripherally-restricted KOR programs + target for centrally-restricted KOR modulators in OUD morphine withdrawal; bioRxiv VTA BDNF transcription requirement for cocaine-seeking incubation (West lab) — in vivo dCas9/CRISPRi + snRNA-seq shows BDNF transcription in VTA during abstinence (not exposure) causally drives incubation + prolonged-absence-of-cocaine gene program in dopaminergic + glutamatergic VTA neurons = incubation is transcriptionally-driven active adaptation, narrow VTA-and-abstinence-window therapeutic target; bioRxiv PACAPergic BAC engages BLA + anterodorsal thalamic networks in looming-threat labyrinth memory — chemogenetic BAC silencing impairs escape + increases freezing + selective PACAP deletion blocks looming-induced Fos in AD only (not pBLA + not dPAG) + PACAP modulates AD + pBLA intrinsic excitability = new PACAP forebrain hub separating neuron-firing from peptide-release roles + PAC1-modulator PTSD substrate beyond BNST + amygdala; bioRxiv neonatal oxytocin P2-P12 prevents male-specific spatial memory deficits induced by maternal separation — restores CA1 Schaffer collateral LTP + normalizes dorsal CA1 neuroinflammatory + GABAergic + synaptic transcripts = OTR agonism protective early in postnatal window, callback to yesterday's OTR-blockade harmful paper + bidirectional developmental-window pharmacology with sex + domain specificity; bioRxiv KCC2 differential regulation by BDNF-TrkB vs NMDAR via distinct Ca modes (De Koninck) — TrkB via ryanodine intracellular Ca modulates function without expression + NMDAR via extracellular Ca influx internalizes + calpain degradation via voltage-gated Ca influx + function → internalization → degradation scales with stimulus strength + time = KCC2 activators (pain, epilepsy, autism) evaluated on 3 arms separately, callback to yesterday's chloride homeostasis thread; bioRxiv PS19 hippocampal metaplasticity — CA1 increased basal + LTP deficits + dentate opposite + transregional metaplasticity occluded + C3 astrocyte marker only in stratum oriens + CD68 in oriens + lacunosum-moleculare = laminar specificity for anti-inflammatory tauopathy targeting; bioRxiv v3 surface AMPAR nanoscale organization via CAM2-Alexa647 + two-color 3D dSTORM at <10 nm precision — WT CA1 has 2× smaller synaptic AMPAR cluster fraction than cortex + PS19 disrupts hippocampal AMPAR clustering + nanodomains before overt neurodegeneration = early tauopathy biomarker at synaptic-organization level; bioRxiv vessel-enriched transcriptomics AD + FTD + HD — partially conserved vascular dysfunction across all 3 + disease-specific endothelial/pericyte/perivascular changes + capillary + venous endothelial remodeling + matrix-type pericyte fraction consistently reduced across all 3 + endothelial-pericyte comm reorganized around LAMININ + COLLAGEN + FN1 + NCAM + NOTCH + VEGF = BBB dysregulation as disease-specific PK modifier + matrix-type pericyte loss as pan-neurodegenerative vascular target; bioRxiv class A GPCR common evolutionary activation CV — coevolution + ML on 447 GPCRdb structures derives sequence-independent 2D CV + biases MD along CV drives 20 receptors across 20 subfamilies through activation transitions within 1 μs + transfers to orphan GPR183 not in training set + converges β2-AR activation FES in presence of different ligand modalities = enabling tool for structural + dynamical studies of hundreds of orphan/understudied GPCRs; bioRxiv dynamic consensus pocket detection across MD ensembles — 6 tools (Fpocket, SiteFerret, AutoSite, SiteHound/EasyMIFs, P2Rank, DeepPocket) on GLUT1 + aldose reductase + HDBSCAN + IoU + temporal persistence = energy-based more sensitive to transient cryptic + geometry-based needs pre-formed cavities + consensus multi-tool framework for druggable cryptic pockets in dynamic systems; bioRxiv control-validated pan-proteome DTI (GNN + ESM-2 + bidirectional cross-attention + losartan-anchored frequent-hitter debiasing + control-anchored docking) nominates orphan GPCR GPR35 as candidate target of ligiamycin A — -8.1 kcal/mol at GPR35 agonist pocket vs zaprinast -8.3 vs non-binder floor -5.5 + FFAR1 fails negative-control test = control-anchored DTI + debiased ranking as principled natural-product target-hypothesis workflow (posted 2026-07-06); arXiv AgentODE — LLM proposes ODE structures + tool-augmented agent refines parameter distributions from population-summary statistics alone + evaluated on 3 benchmarks + 2 clinical (including RDEB 231 obs, 46 patients) + summary-only outperforms individual-level on mechanistic plausibility despite worse predictive fit = privacy constraint as inductive bias toward mechanism + new direction for rare-disease pharmacometrics/QSP. false Receptor & Reason for July 5, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Twelve items today, biased toward neuromodulator-and-astrocyte biology of learning. Opens with a chemogenetic locus coeruleus study out of the Winstanley lab in TH-Cre rats performing the cued rat gambling task. Inhibiting catecholaminergic LC neurons throughout acquisition accelerates the development of risky choice strategies in both sexes; trial-by-trial the LC-inhibited animals switch more after safe wins and switch less away from risky options after wins and losses, arguing LC normally provides a signal to break the cycle of returning to uncertain outcomes. LC inhibition selectively enhances motor impulsivity in females early in training — choice impulsivity and motor impulsivity appear to have separable LC substrates, and the motor arm may be more sensitive in females, arguing for sex-stratified endpoints in LC-noradrenergic (atomoxetine, guanfacine) trials for pediatric ADHD and gambling disorder. Then a v2 preprint from the Sur lab at MIT on cortical astrocytes extending phasic norepinephrine signals. Phasic LC firing after a false alarm improves next-trial discrimination, but LC bursts last tens of milliseconds — how does that persist? Cortical astrocytes on α1a adrenergic receptors respond with sustained calcium elevations lasting seconds, longer than neurons and spanning the inter-trial interval; astrocyte calcium is causally required for the behavioral gain, and the astrocyte-to-neuron signal is purinergic — blocking neuronal A1 adenosine receptors prevents post-reinforcement behavioral gain. Astrocyte α1a and neuronal A1 adenosine as the actual downstream levers for NE-in-learning drug action in attention deficit, PTSD, and dementia. Then a Turecki/Nagy multi-omics paper on astrocyte subtypes in MDD dorsolateral PFC. Spatial transcriptomics plus matched snRNA-seq and snATAC-seq — the finding that survives all three modalities localizes to deep cortical layers and converges on the PSAP-GPR37L1 signaling axis. GPR37L1 (astrocyte GPCR) plus prosaposin as endogenous ligand coherently downregulated across transcript, chromatin, and spatial context — puts GPR37L1 on the depression target map with a proper multi-omics anchor and adds to the stacking case for astrocyte GPCR signaling as an underappreciated psychopharm surface. Then a projection-specific VTA transcriptomics + circuit paper on fentanyl-context associations. VTA-to-NAc vs VTA-to-PFC projections show discordant fentanyl-induced transcriptional adaptations; fiber photometry and chemogenetics support mesolimbic (VTA-to-NAc) carrying acute reward and mesocortical (VTA-to-PFC) carrying delayed contextual-association updating. Relapse-prevention pharmacology reversing fentanyl-context associations should focus on the mesocortical arm. Then a Stuber/Palmiter paper identifying a cholinergic eligibility trace in central amygdala for conditioned flavor avoidance. Parabrachial Calca+ acetylcholine projections to CeA activate CeA neurons and enhance glutamatergic responsivity on the timescale of minutes — the eligibility trace that keeps synapses primed long after a novel taste. CRISPR-Cas9 knockdown and optogenetic silencing both block CFA. Concrete molecular eligibility trace for temporal credit assignment — a specific hypothesis for how muscarinic pharmacology (M1/M4 agonists in current SCZ trials) affects associative learning. Then a receptor-pharmacology developmental-programming paper: neonatal OTR blockade (dOVT P2-P12) recapitulates the pain-hypersensitivity and cognitive-deficit phenotypes of maternal separation, without any actual stress. Mechanical and cold thermal hypersensitivity match maternal-separation animals; anxiety-like behaviors do not — those need more than OTR blockade. Sex-specific cognitive deficits emerge. Spinal cord GAD65/BDNF/CD11b downregulated and elevated NKCC1/KCC2 ratio in both sexes, indicating compromised chloride homeostasis. Uncomfortable pharmacology takeaway: OTR agonists like carbetocin considered for various neurodevelopmental indications need to reckon with the possibility that either direction of perturbation during this window programs lasting phenotypes. Then a pair of tauopathy microglia stories that go well together. HuMiNAX (Li Gan lab) is a triple-xenograft humanized model — human iPSC-derived neurons, astrocytes, and microglia grafted into adult mouse brain. Tau seeding aggregates only in mutation-carrying human neural grafts; progranulin overexpression in human microglia preserves neurons and restores neuronal gene expression/splicing, anchoring progranulin therapy (in FTD-GRN development) to microglia-mediated resilience in tauopathy generally, not just haploinsufficiency contexts. CRISPRi knockdown of the human-specific lncRNA HNRNPK-AS1 also protects — druggable-by-antisense tau target, adds to the noncoding-first hypothesis. The complementary Bowles/Mancuso paper tests iPSC-microglia carrying MAPT S305N across two independent differentiation protocols — mutant microglia are hypoactive (impaired phagocytosis, reduced cytokine release, diminished regulation of synaptic function), arguing the disease-linked microglial phenotype in tauopathy is early failure-to-respond, not classic pro-inflammatory activation. Therapeutic strategy should boost microglial function early — the opposite of the anti-inflammatory strategies currently in AD-adjacent trials. Then dense CRISPR-tile mutagenesis of the WAVE regulatory complex in human THP-1 myeloid cells with a CCL2-directed migration readout. NCKAP1L nominated for pharmacological targeting; ML-guided compound prediction and wet-lab validation give montelukast (asthma drug, CysLT1 antagonist) as a negative modulator of migration and piperacetazine (historical D2 antagonist antipsychotic) as a positive modulator. Workflow more important than the specific hits: CRISPR-tile the functional protein, ML repurposing screen, wet-lab validate — cheaper than de novo medicinal chemistry and appropriate for target classes where too-selective a modulator isn't wanted. Then FastBindRank — distillation of Boltz-2 into a sequence-based surrogate for ultra-large virtual screening. Trained on ~1% of 122M PubChem compounds; applied to HDAC11, gives 74× hit-rate and 30× discovery-yield increase over direct subset screening under fixed compute; two novel actives experimentally validated. The lightweight sequence model captures structural patterns associated with predicted binding — distillation transfers pocket-selective preferences, not just the score. Viable general recipe for CNS/psychiatric screening pipelines. Then a serotonin aptamer NMR structure. The widely used 44-nt Apt44 turns out to be a G-quadruplex-duplex hybrid with serotonin binding at the G-quadruplex-duplex junction. The parent Apt44 has a long third loop that introduces conformational dynamics, so serotonin triggers a discrete binding-induced conformational switch — the large structural change good sensors want for transduction. Enabling structural biology for rational aptamer engineering that turns FET/fluorescent aptamer sensors into properly quantitative in-vivo tools for psychedelics and SSRI mechanism research. Closes on NeuroCogMap — a cognitive-neuroscience-inspired functional parcellation of LLM internals. Three claims: parcels are stable and semantically coherent across models; major LLM failure modes (hallucination, bias, refusal failure, sycophancy) map to distinct parcel-level disruptions, giving a mechanism-guided intervention handle; and the parcellation improves prediction of human cortical responses during naturalistic language comprehension. The cortex-LLM alignment is more decorative than causal at this point, but the mechanism-guided intervention direction borrows the right stuff from systems neuroscience. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-05-receptor-and-reason Sun, 05 Jul 2026 12:00:00 +0000 1003 Daily roundup for July 5, 2026: bioRxiv chemogenetic LC-NA inhibition in TH-Cre rats + cued rat gambling task — LC inhibition accelerates risky-choice adoption in both sexes + trial-by-trial promotes switches after safe wins + reduces switches away from risky options after wins/losses = LC normally provides break-cycle signal + LC inhibition selectively enhances motor impulsivity in females early in training = choice vs motor impulsivity have separable LC substrates + motor arm more female-sensitive, sex-stratified endpoints for ADHD/gambling LC-NE (atomoxetine/guanfacine) trials; bioRxiv v2 Sur lab (MIT) cortical astrocytes extend phasic NE signals + critically mediate learned behavior — go/no-go graded evidence + astrocytes on α1a AR respond with sustained Ca elevations spanning ITI + astrocyte Ca causally required for post-false-alarm next-trial discrimination gain + astrocyte-to-neuron signal purinergic (block neuronal A1 adenosine prevents behavioral gain) = astrocyte α1a + neuronal A1 as downstream levers for NE-in-learning drug action; bioRxiv Turecki/Nagy multi-omics MDD dlPFC astrocyte subtypes — spatial transcriptomics + snRNA + snATAC converge on deep-cortical-layer PSAP-GPR37L1 axis (astrocyte GPCR + endogenous ligand coherently downregulated) = GPR37L1 on depression target map with proper multi-omics anchor + astrocyte GPCR signaling stacking as underappreciated psychopharm surface; bioRxiv VTA projection-specific transcriptomics + fentanyl-context associations — VTA-to-NAc vs VTA-to-PFC show discordant fentanyl-induced transcriptional adaptations + fiber photometry + chemogenetics support mesolimbic acute-reward + mesocortical delayed-context-updating = relapse-prevention pharmacology reversing fentanyl context associations should target mesocortical arm; bioRxiv Stuber/Palmiter cholinergic eligibility trace in CeA for conditioned flavor avoidance — PB Calca+ ACh projections + 2P Ca imaging shows ACh enhances glutamatergic responsivity in CeA on min-timescale (not ms) + CRISPR + optogenetic silencing block CFA = concrete molecular eligibility trace for temporal credit assignment + M1/M4 muscarinic agonist SCZ trial hypothesis; bioRxiv neonatal OTR blockade (dOVT P2-P12) recapitulates maternal-separation pain-hypersensitivity + cognitive-deficit phenotypes without stress — mechanical+cold hypersensitivity matches NMS + anxiety NOT recapitulated + sex-specific cognitive: NMS-males + dOVT-females impaired + spinal GAD65/BDNF/CD11b downregulated + elevated NKCC1/KCC2 ratio = OTR agonists (carbetocin) for neurodevelopmental indications need to reckon with dev-window sensitivity in both directions (v1 abstract-only); bioRxiv Li Gan HuMiNAX triple-xenograft humanized iPSC neuroimmune model + progranulin OE dampens tau inflammation + preserves neurons + restores gene expression/splicing = progranulin therapy anchored to microglia-mediated tauopathy resilience beyond FTD-GRN + CRISPRi HNRNPK-AS1 lncRNA also protects = druggable-by-antisense tau target; bioRxiv Bowles/Mancuso iPSC-microglia MAPT-S305N cross-protocol hypoactivity (impaired phagocytosis + reduced cytokine + diminished synaptic regulation) = disease-linked microglial phenotype in tauopathy is early failure-to-respond not pro-inflammatory activation + therapeutic strategy should boost microglial function early (opposite of anti-inflammatory AD-adjacent trials); bioRxiv WAVE regulatory complex CRISPR-tile + ML-guided repurposing = montelukast (CysLT1 antagonist) as negative modulator + piperacetazine (D2 antagonist antipsychotic) as positive modulator of chemokine-directed myeloid migration + NCKAP1L nominated as microglia-enriched target = workflow (CRISPR-tile + ML + wet-lab-validate) cheaper than de novo medchem + appropriate where too-selective modulator not wanted; bioRxiv FastBindRank distillation of Boltz-2 into sequence-based surrogate + trained on ~1% of 122M PubChem + HDAC11 target + 74× hit rate + 30× discovery yield vs direct subset screening + 2 novel actives validated = distillation transfers pocket-selective preferences not just score, viable general recipe for CNS screening pipelines; bioRxiv serotonin aptamer NMR structure Apt44/Apt38 = G-quadruplex-duplex hybrid + serotonin binds at G-quad-duplex junction + long-loop conformational dynamics give binding-induced switch = enabling structural biology for rational aptamer engineering turning FET/fluorescent sensors into quantitative in-vivo tools; arXiv NeuroCogMap cognitive-neuroscience parcellation of LLM internals (q-bio.NC + cs.AI + cs.CL) — parcels stable + semantically coherent across models + LLM failure modes (hallucination/bias/refusal/sycophancy) map to distinct parcel disruptions (mechanism-guided intervention handle) + parcellation predicts human cortical responses in naturalistic language comprehension with strongest fit in higher-order association cortex = cortex-LLM alignment more decorative than causal but mechanism-guided-intervention direction borrows the right systems-neuroscience framing. false Receptor & Reason for July 4, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. A structure-and-pharmacology-heavy week, ten items biased toward receptor pharmacology. Opens on a systematic comparison of two 5-HT2A antagonists — MDL 11,939 (glemanserin, mixed 5-HT2A/2C) vs MDL 100,907 (volinanserin, selective 5-HT2A) — on fear extinction in adult C57 mice. Both compounds, dosed at behaviorally comparable levels (1 mg/kg glemanserin, 10 μg/kg volinanserin), acutely enhance freezing on the first CS presentation at extinction, so they agree on acute fear expression. Where they diverge is on extinction itself: repeated glemanserin significantly impairs extinction (more freezing across trials, more trials to criterion), while repeated volinanserin does not. SB 242084 kills the 5-HT2C-mediated confound, leaving ligand-directed signaling at 5-HT2A as the interpretation — the same biased-signaling argument the psychedelic-agonist literature has debated, now holding on the antagonist side. v1 preprint, body not yet rendered — the abstract carries the finding but I want to see the intracellular readouts when they land. Then a cryo-EM series on the most abundant native kainate heteromer, GluK2/GluK5. Apo, glutamate-desensitized, and UBP-310 competitive-antagonist states with engineered affinity-swap mutations let the same tetramer capture GluK5-selective vs GluK2-selective inhibited states. Antagonizing GluK5 (A+C positions, high-affinity subunit) closes the pore dramatically more effectively than antagonizing GluK2 (B+D), because Met648 on GluK5 flips into a "lid" across the pore apex when GluK5 ligand-binding domains rotate outward — a molecular rationale for pushing kainate-receptor drug discovery specifically toward GluK5-selective compounds for migraine, depression, and epilepsy. Then a subunit story worth catching up on — a second revision of the GluN2D dentate paper. GluN2D-containing NMDA receptors in mature dentate granule cells are tonically active, primarily extrasynaptic, and contribute to AP threshold; a physiologically realistic stimulation pattern induces strong GluN2D-dependent LTP of NMDAR-EPSCs at the medial perforant path, attributed to lateral diffusion of extrasynaptic GluN2D-containing receptors into synapses. Conditional knockout impairs hippocampus-dependent memory. Several NMDA-targeting depression/autism/SCZ candidates hit or spare GluN2D depending on subunit profile — this gives the field a functional readout to worry about, and argues open-channel-block vs subunit-selective strategies are quite different bets in dentate circuits. Then a neuropeptide-receptor circuit paper: MCHR1 antagonism (SNAP-94847) in active avoidance extinction. Systemic and prelimbic-mPFC-targeted infusion both accelerate extinction of tone-shock decoupled avoidance while leaving acquisition intact. Fiber photometry from CaMKII+ mPFC neurons shows MCHR1 antagonism enhances excitatory responses on successful avoidance and alters trial-history-dependent activity on avoidance-after-avoidance trials. Endogenous MCHR1 tone in PFC persists an avoidance policy; blockade releases the animal to update. MCHR1 having failed several obesity indications deserves a second look for compulsivity/habit-persistence phenotypes (OCD, addiction, chronic anxiety) where the ability to extinguish a defensive policy is exactly the deficit. Then early-life stress (maternal separation P5-P19) + serial probabilistic reversal learning in adult rats crossed with propranolol, citalopram, and exogenous corticosterone probes. ELS rats over-adjust to noisy feedback — brittle flexibility, not more flexibility. Propranolol and citalopram each restore parts of the phenotype; corticosterone recapitulates parts of the stressed profile in unstressed controls — doubly-modulatory NA+5-HT substrate for the clinical observation that adults with early-life adversity have altered feedback processing under uncertainty, mappable onto drug classes already in use for stress-related mood disorders. Then a mechanism paper that will matter for naloxone-resistance in the fentanyl era. Fentanyl overdose in urethane-anesthetized mice resolves into two phases: Phase 1 is profound RR reduction with rhythmic diaphragmatic coordination preserved; Phase 2 sees partial ventilatory rebound but the diaphragm goes tonic, loses inspiratory-phase EMG dominance, and the two hemi-diaphragms decorrelate. Carotid body denervation removes the tonic component and expiratory-phase EMG elevation but doesn't restore bilateral coordination, and brainstem slice recordings show pre-Bötzinger complex bilateral coordination is disrupted intrinsically. The rescue window is smaller than a naloxone-plus-time model assumes, and adjunct rescue pharmacology (orexin agonists, TRH analogs) needs to target brainstem rhythm coordination, not just chemoreflex drive. Then a noncoding-tauopathy paper. Two lncRNAs — NORAD and MIR22HG — are consistently dysregulated across iPSC-derived neurons, astrocytes, and microglia carrying the MAPT IVS10+16 FTD mutation, and in matched patient brain tissue; NORAD is also altered in AD and PD brains, so it looks like a broader neurodegeneration signature. NORAD sequesters PUM1/PUM2, so unbalancing it de-represses pumilio-targeted transcripts whose protein network converges on RNA regulation, cytoskeleton, proteostasis, and tau interaction networks. lncRNA-pumilio dysregulation as an upstream node for the tau-propagation proteome, with NORAD druggable in principle via antisense or pumilio-competition — a noncoding-first target hypothesis the tauopathy field has been slow to explore (v1 preprint, awaiting causal experiments). Then a repurposing story: "hypoxia in a pill" from the Rutkowski group. GBT-601 (voxelotor cousin, hemoglobin affinity enhancer in sickle-cell development) plus PT-2399 (HIF-2 inhibitor to prevent erythropoietic compensation) delivers tissue hypoxia comparable to chronic 11% O2 breathing without breathing hypoxic gas, and halts progression and reverses neurological symptoms in three preclinical models: Ndufs4 Leigh syndrome, Friedreich's ataxia, and a third neurodegeneration model. Two clinical-stage compounds already in humans — the IND path for Leigh or Friedreich's is much shorter than de novo development, with the pulmonary hypertension/right heart strain worry recontextualized against untreatable pediatric disease. Closes on two computational chemistry items. ConfDock does atom-specific uncertainty quantification for molecular docking via conformal prediction — GNN quantile estimator plus split conformal calibration gives atom-specific prediction intervals with distribution-free finite-sample coverage guarantees, evaluated across 238 protein-ligand complexes and four docking methods, cutting mean interval width by more than half vs a standard conformal baseline without losing target coverage. Tighter intervals for atoms in constrained pockets than in solvent-exposed loops — a drop-in upgrade to any docking pipeline where downstream synthesis prioritization depends on which contacts to trust. And finally, Boltz-2 co-folding for bitter taste receptor pocket prediction. The bitter taste receptors (TAS2Rs) bind ligands in a non-classical intracellular pocket for several members. Boltz-2 nails the correct pocket for 12 of 15 experimentally solved agonist-TAS2R complexes and correctly identifies intracellular binding cases that classical docking on predicted receptor models cannot; aristolochic acid (intracellular at TAS2R14, extracellular at TAS2R43) morphed in silico pins TM3+TM7 as pocket-selectivity determinants. Extending to 1,500 agonist-receptor associations in BitterDB, most TAS2Rs prefer extracellular but several are dual-pocket-competent. Beyond taste (airway smooth muscle, gut chemosensing), co-folding is now a viable early-triage step for any GPCR target where allosteric-vs-orthosteric bias is on the table. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-04-receptor-and-reason Sat, 04 Jul 2026 12:00:00 +0000 849 Daily roundup for July 4, 2026: bioRxiv 5-HT2A antagonist head-to-head glemanserin (MDL 11,939, mixed 2A/2C) vs volinanserin (MDL 100,907, selective 2A) in fear extinction — both acutely enhance CS-1 freezing but only repeated glemanserin impairs extinction + SB 242084 rules out 5-HT2C = ligand-directed signaling at 5-HT2A on the antagonist side; bioRxiv GluK2/GluK5 heteromeric kainate cryo-EM apo + desensitized + UBP-310 states with engineered affinity-swap letting the same tetramer capture GluK5-selective vs GluK2-selective inhibited states — Met648 lid across the pore apex when GluK5 LBDs rotate outward + antagonizing GluK5 closes the pore dramatically more than GluK2 = subunit-selective drug discovery for migraine/depression/epilepsy; bioRxiv GluN2D-containing NMDARs in mature dentate granule cells — tonically active + extrasynaptic + sets AP threshold + physiological stimulation induces GluN2D-dependent LTP at medial perforant path via lateral diffusion + conditional KO impairs memory = functional readout for depression/autism/SCZ candidates depending on subunit profile; bioRxiv MCHR1 antagonism (SNAP-94847) in active avoidance extinction — systemic + prelimbic mPFC infusion both accelerate extinction + fiber photometry shows MCHR1 antagonism enhances CaMKII+ mPFC responses on successful avoidance = MCHR1 deserves a second look for compulsivity/habit-persistence phenotypes (OCD, addiction, chronic anxiety); bioRxiv early-life stress (maternal separation P5-P19) + serial probabilistic reversal learning in adult rats + propranolol + citalopram + corticosterone probes — ELS rats over-adjust to noisy feedback + doubly-modulatory NA+5-HT substrate maps onto drug classes already used for stress-related mood disorders; bioRxiv fentanyl overdose diaphragmatic discoordination — Phase 2 tonic diaphragm + hemi-diaphragm decorrelation + carotid-body-independent + pre-Bötzinger bilateral coordination disrupted intrinsically = naloxone-rescue window smaller than assumed + adjunct rescue pharma (orexin, TRH) needs to target brainstem rhythm coordination not just chemoreflex drive; bioRxiv NORAD/MIR22HG lncRNA-pumilio-tau axis in MAPT IVS10+16 iPSC neurons/astrocytes/microglia + patient brain tissue — NORAD also altered in AD + PD + de-represses pumilio-targeted transcripts converging on RNA regulation + cytoskeleton + proteostasis + tau interaction networks = noncoding-first target hypothesis, NORAD druggable via antisense or pumilio competition; bioRxiv "hypoxia in a pill" (Rutkowski) — GBT-601 (voxelotor cousin) + PT-2399 (HIF-2 inhibitor) delivers tissue hypoxia comparable to chronic 11% O2 + reverses neurological symptoms in Ndufs4 Leigh + Friedreich's + a third neurodegeneration model = repurposing two clinical-stage compounds shortens IND path for pediatric untreatable disease; bioRxiv ConfDock atom-specific uncertainty quantification for docking via conformal prediction — GNN quantile estimator + split conformal calibration + atom-specific intervals with finite-sample coverage guarantees + more-than-half interval-width reduction on 238 complexes + 4 docking methods = drop-in pipeline upgrade where downstream synthesis prioritization depends on which contacts to trust; bioRxiv Boltz-2 co-folding for TAS2R (bitter taste) pocket prediction — 12/15 pocket-choice correct on solved TAS2R-agonist complexes + intracellular binding correctly ID'd where classical docking on predicted models fails + TM3+TM7 as selectivity determinants + BitterDB shows several TAS2Rs dual-pocket-competent = co-folding as viable GPCR pocket-choice hypothesis generator beyond taste (airway smooth muscle, gut chemosensing). false Receptor & Reason for July 3, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Ten items today, agentic-AI-heavier than usual because the corpus at that end was unusually rich. Opens on Copewell, an arXiv architecture paper proposing a multi-agent swarm for equitable mental wellness support — three architectural bets worth arguing about: multi-source assessment fusing self-report, physiological, and contextual signals to mitigate algorithmic bias; Russell Circumplex-based valence-arousal routing to specialized sub-agents; and a dedicated Ethics Supervisor agent embedded in the orchestration layer rather than a filter bolted on. This is a design + beta paper — no controlled efficacy trial vs standard digital CBT and no evidence the affect routing improves outcomes over monolithic agents. Treat it as the architectural proposal the field will be debating as agentic systems enter psychiatry, not as clinical evidence. Then the 2.6-Å cryo-EM structure of the TrkA extracellular domain in complex with the Fab of neutralizing mAb 42F5-15. The tanezumab-class systemic NGF-sequestration antibodies got bruised by rapidly progressive osteoarthritis and pan-Trk kinase inhibitors carry CNS off-target liabilities via TrkB/TrkC; the argument for direct TrkA extracellular-domain blockade is receptor-selective analgesia without either liability. The antibody epitope overlaps the NGF-binding interface — orthosteric, not allosteric as previously proposed — and hits residues conserved in TrkA but divergent in TrkB/TrkC, providing the structural rationale for isoform selectivity. In vivo the antibody attenuates mechanical allodynia. Not a clinical candidate but a clean starting point for a receptor-selective extracellular non-opioid analgesic. Then a paper to read carefully but skeptically about framing — "state-dependent transcriptomic collapse of the brain's lactate and ketone thermodynamic sensors in schizophrenia." The mechanistic core is real: SCZ GWAS summary statistics integrated with Hi-C chromatin conformation and multi-tissue eQTL data localize non-coding structural variance to the HCAR2/HCAR1 tandem locus — hydroxycarboxylic acid receptors for lactate and β-hydroxybutyrate/niacin. Authors claim the shared TAD is fractured to downregulate HCAR1 on the 3′ side and paralyze HCAR2 on the 5′ side, giving a mechanism for known SCZ biomarkers (elevated CSF lactate, absent niacin flush) and linking to the metabolic-psychiatry ketogenic-diet results. The topology finding is worth pursuing; the "evolutionary fuel mismatch" grand-unified framing goes well past the mechanistic data, with no circuit experiment or intervention. Read for the double-dissociation topology, not the theory. Then a real drug discovery paper: MCULE-5095997944, the first small-molecule modulator of the ARF1–C9orf72:SMCR8:WDR41 interaction, for ALS/FTD. Postmortem sporadic and C9orf72-mutant motor cortex shows elevated phospho-ASAP1 Tyr782, a marker for ASAP1 inactivation and therefore for ARF1 stuck in the GTP-bound state. Rational design plus virtual screening of a 40M-compound library targeting the ARF1-CSW interface nominates the top hit, which binds ARF1 in the nanomolar range, reduces GTP-ARF1 on stimulation, and reorganizes ARF1-dependent Golgi in cells. Tool compound, not clinical candidate — no in vivo data, no ADME beyond in silico — but a proof of concept that the CSW-ARF1 interaction is druggable with a coherent target story from human tissue through cells. Then a prenatal cannabinoid mPFC circuit paper: dams exposed to 3 mg/kg CBD or Δ9-THC from GD5-18, offspring studied at P100+. Both compounds abolish endocannabinoid LTD at L5 mPFC neurons, so eCB-LTD loss isn't THC-specific. Compensations are sex-specific: CBD-exposed males show a bidirectional-plasticity saturation ceiling with elevated AMPA/NMDA ratios and slowed NMDAR kinetics; CBD-exposed females preserve net E/I but with a scaled-up architecture; THC-exposed females shift pro-excitatory via inhibitory-tone loss. Only CBD-exposed females show a risk-appraisal deficit dissociated from classical anxiety readouts. Take-home: prenatal CBD isn't neutral, and the receiving-synapse mechanism differs from prenatal THC even at the shared eCB-LTD ceiling. Then StructureSAFE — a structure-aware chemical language model resolving the SBGM trade-off. 3D-graph SBGMs incorporate protein pockets but generate chemically implausible molecules; chemical language models produce plausible molecules but are structure-blind. StructureSAFE fuses protein structural + evolutionary encoders with the SAFE fragment-embedding representation and supports both de novo hit ID and comprehensive lead-optimization subtasks in one model, with the biggest deltas on chemical plausibility vs graph baselines. Lead optimization — scaffold hopping, R-group enumeration, activity-cliff avoidance — is the productive workflow slot that generative models built for from-scratch case have chronically undersupported. Then Active-GRPO — adaptive imitation and self-improving RL for molecular optimization. Active imitate-reinforce lets the policy decide per instance whether to imitate a reference or reinforce its own discoveries; active referencing continually upgrades the reference with the best policy-generated candidate, so the imitation target keeps climbing. Statistically robust gains on TOMG-Bench MOLOPT LogP/MR/QED. The recipe matters more than the absolute numbers — most SFT-then-RL molecular optimization silently caps at reference quality. Then MolSafeEval — an ACL 2026 Findings benchmark for safety risks in AI-generated molecules. Toxicological databases + hazard rules assembled into a molecular safety knowledge graph, then LLM-based reasoning on the graph detects and explains unsafe features across four generation modes (unconditional, property-optimized, target-conditioned, text-conditioned). Current generative models fail on safety in ways narrow toxicity predictors miss. Run this before claiming deployment-ready; open question whether safety supervision at the RL reward level closes the gap or just teaches the model to game the safety graph. Then a normative-modeling neuroimaging paper: cerebellar biotypes predict deep-TMS response in spinocerebellar ataxia type 3. Normative model built from 2,071 healthy controls, applied to 114 genetically confirmed SCA3 patients; unsupervised clustering pulls out two biotypes — Biotype 1 with posterior-cerebellum preservation, Biotype 2 with anterior-motor-cerebellum atrophy centered on lobules I-VI. Biotype 2 is worse at baseline, but responds better to deep TMS, while Biotype 1 responds better to conventional rTMS — counterintuitive stratification for a disease with no disease-modifying therapy. The pediatric-growth-chart-style TMS-response-prediction framework is transferable to TRD and OCD. Closes on a small mechanistic sleep-neurodegeneration study: n=10 healthy older adults, one night of slow-oscillatory tDCS during NREM sleep vs one sham night crossover, then 5 consecutive stimulation nights, with plasma p-tau181, Aβ42, Aβ40, and total tau measured overnight and longitudinally. Stimulation raised SO power and SO-spindle coupling; overnight plasma p-tau went up after stim vs sham; a shift in coupling phase toward the SO up-state was associated with overnight Aβ42 and Aβ40 increases and longitudinal Aβ42 and Aβ42/40 decreases — differential SO-power-vs-coupling-timing associations arguing partially distinct tau and amyloid dynamics. Ten subjects is ten subjects. But so-tDCS is one of the few sleep-microstructure interventions with a plausible non-pharmacological handle on amyloid and tau clearance dynamics that could pair with anti-amyloid immunotherapy without ARIA — a 500-person confirmatory trial with this design would be genuinely informative. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-03-receptor-and-reason Fri, 03 Jul 2026 12:00:00 +0000 736 Daily roundup for July 3, 2026: arXiv Copewell multi-agent swarm for equitable mental wellness support + 3 architectural bets (multi-source self-report+physiological+contextual assessment against algorithmic bias, Russell Circumplex valence-arousal routing to specialized sub-agents, dedicated Ethics Supervisor agent in orchestration) + design/beta paper only + no controlled trial vs digital CBT — architectural proposal to argue about not clinical evidence; bioRxiv 2.6 Å cryo-EM TrkA-ECD with analgesic mAb 42F5-15 Fab + orthosteric NGF-binding-interface overlap (not allosteric as debated) + epitope residues conserved in TrkA but divergent in TrkB/TrkC = structural basis for isoform selectivity + in vivo mechanical-allodynia attenuation, receptor-selective non-opioid analgesic target sidestepping tanezumab-class systemic NGF-sequestration OA + pan-Trk CNS off-target liabilities; bioRxiv SCZ GWAS + Hi-C + multi-tissue eQTL localize non-coding structural variance to HCAR1/HCAR2 TAD (lactate + BHB/niacin sensors) + double-dissociation topology model + tie to metabolic-psychiatry ketogenic-diet results — receptor topology worth follow-up but "evolutionary fuel mismatch" framing goes past mechanistic data; bioRxiv ARF1 small-molecule MCULE-5095997944 for C9orf72:SMCR8:WDR41 ALS/FTD + elevated phospho-ASAP1 Tyr782 in postmortem sALS + C9ALS motor cortex + 40M-compound rational + virtual screen against ARF1-CSW interface + nanomolar binding + reduces GTP-ARF1 + reorganizes ARF1-dependent Golgi, tool compound not clinical candidate but proof CSW-ARF1 interface is druggable; bioRxiv prenatal CBD vs Δ9-THC GD5-18 3 mg/kg in adult offspring mPFC L5 — both compounds abolish eCB-LTD (not THC-specific) + sex-specific compensations: CBD-males bidirectional-plasticity ceiling + elevated AMPA/NMDA + slowed NMDAR kinetics + CBD-females intact E/I but scaled-up architecture + THC-females pro-excitatory via inhibitory-tone loss + CBD-females alone risk-appraisal deficit dissociated from classical anxiety, prenatal CBD is not neutral + mechanism at receiving synapse differs from prenatal THC; bioRxiv StructureSAFE structure-aware chemical language model resolving 3D-graph-vs-CLM SBGM trade-off + protein structural+evolutionary encoders fused with SAFE fragment-embedding representation + unified de novo hit ID + comprehensive lead-optimization subtasks + biggest gains on chemical plausibility vs graph baselines — productive workflow slot for lead-opt scaffold hopping, R-group enumeration, activity-cliff avoidance chronically undersupported by generative models; arXiv Active-GRPO active imitate-reinforce (per-instance choose imitate vs reinforce once policy beats reference) + active referencing (upgrade reference with best policy candidate) breaks reference-quality performance ceiling + statistically robust TOMG-Bench MOLOPT LogP/MR/QED gains — recipe matters more than absolute numbers, most SFT-then-RL molecular optimization silently caps at reference quality; arXiv MolSafeEval (ACL 2026 Findings) molecular safety knowledge graph fusing toxicological databases + hazard rules + LLM-based reasoning over 4 generation modes (unconditional, property-optimized, target-conditioned, text-conditioned) surfaces safety vulnerabilities narrow toxicity predictors miss + open question whether safety RL supervision closes gap or teaches gaming the safety graph; bioRxiv cerebellar normative modeling of 2,071 controls + 114 SCA3 patients + unsupervised clustering yields 2 biotypes (posterior preservation vs anterior lobule I-VI atrophy) + counterintuitive result Biotype 2 worse at baseline but greater dTMS response while Biotype 1 responds better to rTMS — stratification biomarker for disease with no DMT + framework transferable to TRD/OCD; bioRxiv slow-oscillatory tDCS during NREM sleep + AD biomarker clearance n=10 healthy older adults + one so-tDCS + one sham night crossover + 5 stim nights + increased SO power + SO-spindle coupling + overnight plasma p-tau181 increase after stim + SO-spindle coupling-phase shift toward SO up-state associated with overnight Aβ42/Aβ40 increases + longitudinal Aβ42/Aβ42/40 decreases — differential SO-power-vs-coupling-timing associations argue partially distinct tau + amyloid dynamics, small sample but rare non-pharmacological handle that could pair with anti-amyloid immunotherapy without ARIA. false Receptor & Reason for July 2, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Ten items today. Opens on the Seattle AD Brain Cell Atlas team extending SEA-AD to ten cortical regions — approximately seven million nuclei from 84 donors profiled by snRNA-seq, ATAC-seq, and Multiome alongside quantitative neuropathology and WGS, mapped to an expanded 207-cell-type BRAIN Initiative taxonomy with a hierarchical pseudo-progression model jointly modeling amyloid-beta and phospho-tau. Only ~30% of cell types shift but move coherently across regions; earliest losses are SST, LAMP5, VIP, SNCG, and PVALB inhibitory interneuron subsets plus myelinating oligodendrocytes alongside AD-associated microglia emergence, with L2/3 and deep-layer excitatory losses, sharper microglial increases, and reactive astrocytes following. Unexpectedly, primary visual cortex L4 IT excitatory neurons are also lost. Replication across three cohorts covering 700+ additional donors. A multi-agentic AI workflow constructed literature-grounded hypotheses from DE and nominated hyperexcitability — mediated in part by high NMDAR expression — as the convergent vulnerability phenotype for L4 IT neurons; SST interneurons converged on hyperexcitability through partly distinct pathways and were enriched for AD GWAS-prioritized genes. Then a female-cohort ketamine study in WKY rats — 10 mg/kg drove strongest delayed anxiolytic effect at 7d, 15 mg/kg strongest immediate at 24h; PV+ interneuron counts unchanged but 8-oxo-dG oxidative stress elevated globally in BLA and prelimbic PFC and specifically inside BLA PV+ interneurons — delayed anxiolysis co-occurs with elevated oxidative stress, arguing for region-specific stress-circuit modulation rather than interneuron preservation. Then the classical striatal CIN pause — optogenetic CIN synchronization elicits GABAA-mediated feedback inhibition requiring β2-nAChR + recruited by thalamostriatal input + D2-modulated but dopamine-independent, and cell-type-specific silencing + local β2-nAChR deletion + retrograde β2-nAChR deletion together localize the inhibitory source outside the striatum, with nicotinic gating on the projection cell not the CIN. Then hippocampal-microcircuit epilepsy — focal β3-GABAA-R KO in CA1 pyramidals selectively impairs PV inhibition onto deep but not superficial PCs + robust IEDs at KO site + CA2-to-deep-CA1 generation mechanism + optogenetic CA2 stimulation reliably evokes IEDs in KO. Then a developmental thalamus paper mapping fine-scale nucleus-specific maturation trajectories to schizophrenia risk zones and a unique neuroreceptor fingerprint that relates to symptom severity — nucleus-resolved neurodevelopmentally-timed pharmacology handle. Then a clinical sensorimotor paper — submovement analysis in 17 psychosis patients + 21 heavy cannabis users + 17 controls shows both groups have less frequent + more variable submovements + attenuated responses to observed movements + effect more pronounced in psychosis, giving a subclinical shared sensorimotor-integration phenotype. Then hippocampal DA — GRAB-DA fiber photometry shows nutrient-dependent post-meal DA elevation in dorsal HPC (chow/HFD/liquid sucrose but not low-calorie sweetener) + pharmacological HPC D2R blockade increases food intake + impairs meal-related episodic memory. Then PKA regulatory subunits — RI vs RII double-KO cell lines show distinct + often opposite proteome + phosphoproteome remodeling under basal + glucose-perturbed conditions, with RI elevating and RII depressing glycolytic signaling + PKA catalytic C-isoform abundance/activity, framing PKA modulation as two orthogonal dials rather than one — direct relevance to AD-context pharmacology. Then F.A.D.E. — an open-source multi-agent conversational drug-discovery platform converting natural-language queries into candidates through a three-branch target-adaptive architecture (structure prediction + pocket detection + equivariant-diffusion de novo ligand generation + affinity estimation), validated on EGFR kinase domain and cellular retinol-binding protein. Closes on a Claude Sonnet 4.6-driven meta-audit of 72,644 bioRxiv-to-publication pairs from 2018-2025 — primary + two secondary claims parsed per pair + content/hedging classified + validation Cohen's kappa 0.63-0.66 vs two domain experts + primary claim unchanged in 39.9% + minor in 50.0% + substantial in 10.2% + hedging shifts asymmetric with 2× more claims becoming more cautious than more confident. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-02-receptor-and-reason Thu, 02 Jul 2026 12:00:00 +0000 1147 Daily roundup for July 2, 2026: SEA-AD extended to 10 cortical regions + ~7M nuclei + 84 donors + 207-cell-type BRAIN Initiative taxonomy + hierarchical pseudo-progression model jointly modeling Aβ + pTau + only ~30% of cell types shift but coherently across regions + earliest losses SST + LAMP5 + VIP + SNCG + PVALB interneurons + myelinating oligodendrocytes + AD-associated microglia + later L2/3 + deep-layer excitatory losses + reactive astrocytes + unexpected V1C L4 IT excitatory neurons + replication across 3 cohorts + 700 additional donors + multi-agentic AI workflow lifts hyperexcitability (partly high NMDAR) as convergent vulnerability phenotype for L4 IT + SST interneurons enriched for AD GWAS-prioritized genes; ketamine in female WKY rats + 10 mg/kg strongest delayed anxiolytic at 7d + 15 mg/kg strongest immediate at 24h + PV+ count unchanged + 8-oxo-dG oxidative stress elevated globally in BLA + prelimbic PFC + specifically inside BLA PV+ interneurons, delayed anxiolysis co-occurs with elevated oxidative stress → region-specific stress-circuit modulation not interneuron loss; striatal CIN pause + optogenetic CIN synchronization + GABAA-mediated feedback + β2-nAChR-required + thalamostriatal recruitment + D2-modulated dopamine-independent + local β2-nAChR + cell-type-specific silencing excludes local GABAergic sources + retrograde β2-nAChR deletion abolishes inhibition, canonical pause generated by long-range circuit + nicotinic gating on projection cell; hippocampal microcircuits + focal β3-GABAA-R KO in CA1 PCs impairs PV inhibition to deep but not superficial PCs + IEDs + CA2-to-deep-CA1 generation + optogenetic CA2 evokes IEDs, epilepsy as microcircuit-selective; thalamus fine-scale developmental portrait + most protracted-development nuclei at greatest SZ risk + neuroreceptor fingerprint correlates with symptom severity; submovements in 17 psychosis + 21 heavy cannabis users + 17 controls + shared reduced/variable submovements + attenuated response to observed movements, subclinical shared sensorimotor-integration phenotype; nutrient-dependent HPC DA via GRAB-DA + post-meal elevation with chow/HFD/liquid sucrose but not low-calorie sweetener + HPC D2R blockade increases food intake + impairs meal memory, HPC D2 signaling in food intake through memory-encoding pathway not classical reward; PKA RI vs RII double-KO cells + distinct + often opposite proteome/phosphoproteome remodeling + RI elevates + RII depresses glycolytic signaling + PKA-C isoform activity, two orthogonal dials for AD-context PKA modulation; F.A.D.E. Fully Agentic Drug Engine + open-source multi-agent conversational platform + three-branch target-adaptive architecture + structure prediction + pocket detection + equivariant-diffusion de novo ligand generation + affinity estimation + validated on EGFR + CRBP; Claude Sonnet 4.6-driven audit of 72,644 preprint→publication pairs + validation Cohen's kappa 0.63-0.66 vs experts + primary claim unchanged 39.9% + minor 50.0% + substantial 10.2% + hedging asymmetric with 2× more cautious than more confident. false Receptor & Reason for July 1, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Ten items today. Opens on a Bryan Roth lab paper that directly challenges the 5-HT2A-mGluR2 heterodimer hypothesis — they replicate mGluR2 agonist attenuation of DOI-driven head-twitch, then use tagged-receptor engineered mice, radioligand binding, and kinetic analyses to find no colocalization or oligomerization at baseline or under 5-HT2A agonist exposure. The surviving mechanism for mGluR2 modulation of psychedelic-relevant 5-HT2A signaling is presynaptic glutamate control, and the receptor-heterodimer target concept needs a hard reread. Then a big-cohort cerebrovascular snRNA-seq of postmortem ITG comparing AD, CBD, Pick's, and PSP — largely disease-specific transcriptional programs across all vascular cell types but heat-shock genes consistently upregulated across every proteinopathy, AD-specific vascular remodeling, and AD GWAS risk loci dysregulated in endothelial cells, with tissue-clearing plus light-sheet validation. Then NPTX2-centered cognitive resilience — 575-sample bulk RNA-seq middle temporal gyrus across four cohorts plus PRM-MS proteomics on 135 of them, showing NPTX2 sitting inside a tight plasticity cluster with BDNF, VGF, SST, and SCG2 that's inversely coupled to a lysosomal-and-chromatin-stress axis. In AD, transcript-level plasticity-cluster correlations weaken while stress-linked transcription regulators FOXJ1, ZHX3, SMAD5, JDP2, and ZIC4 over-correlate with NPTX2 protein loss — a coordinated collapse rather than a single-protein hit. Then ROSMAP mitochondrial Hsp60/10 client proteins — clients decline much more at protein than RNA in late-stage AD, more than abundance-matched non-clients, most-central clients decline most, client abundance inversely tracks Braak/tau and positively tracks preserved cognition, and mitochondrial translation plus TCA-pyruvate-redox clients emerge as highest-priority subnetworks. Then Doeller lab virtual-reality navigation fMRI in controls, preclinical AD, and early AD with CSF Aβ/tau and APOE — dissociable amyloid-β-linked hippocampal hyperactivity during memory-demanding navigation, tau-linked reductions in entorhinal hexadirectional (grid-cell) modulation, and altered object-centered representations in posterior cingulate and entorhinal-adjacent regions. Hippocampal hyperactivity was the earliest and only signal distinguishing preclinical individuals from controls — a navigation-based functional readout for early-intervention trials. Then Paul Kenny's lab: miR-206 in the postnatal brain is Purkinje-cell-restricted, dispensable for cell fate/morphology/motor coordination but regulates translational programs controlling intrinsic excitability. Deficiency increases tonic firing; elevation shifts Purkinje cells to burst firing; either constitutive or PC-specific miR-206 deletion impairs prepulse inhibition, and rescue is bidirectional — restoration reverses PPI deficits, elevation in WT PCs impairs PPI. A schizophrenia-linked microRNA tunes cerebellar Purkinje firing dynamics to control sensorimotor gating. Then STING-does-the-opposite: STING deficiency in the kainic-acid TLE model amplifies microglial activation, drives lysosomal expansion, enhanced phagocytic engulfment of neurons, hippocampal loss, and cognitive impairment through elevated STIM1 and dysregulated store-operated calcium entry. Pharmacological SOCE inhibition rescues neuron survival and cognition — STING repositioned as a negative regulator of microglial activation in this context, with microglial calcium signaling as a defined therapeutic axis in TLE. Then ventral striatal astrocytes: PMCA-based astrocyte-calcium attenuation across DMS, DLS, and VS shows VS-specific decrease in inverse temperature (noisier decisions) and reduced win-stay behavior in a probabilistic reversal-learning task; fiber-photometry shows VS astrocyte calcium correlates with model-inferred reward-prediction errors for many seconds post-outcome; trial-by-trial calcium predicts choice variability; slice ephys and computational modeling implicate presynaptic E-I regulation plus RPE-signal-sharing across MSN populations — astrocytes as a component of the algorithmic apparatus of value-based decision-making. Then social isolation: 10-week adult isolation in BALB/c males increases ventral hippocampal miR-30e-5p, decreases target Neurod1, and impairs object pattern separation without anxiety- or depression-like phenotypes, with a blunted acute-stress response — a cognition-specific hit through post-transcriptional gene regulation, providing a mouse mechanism for the human miR-30e SNP linked to cognition, EEG waveform latency, depression, and schizophrenia. Closes on Microsoft's HealthAgentBench — 54 agentic healthcare tasks across 7 categories with unique environments and end-to-end clinical workflows. Codex GPT-5.5 is the strongest and most cost-effective at ~42% overall success; frontier agents promising on EHR pipeline construction, weak on medical imaging especially for Claude Code models; large-search-space compositional reasoning remains hard for all current agents. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-07-01-receptor-and-reason Wed, 01 Jul 2026 12:00:00 +0000 1110 Daily roundup for July 1, 2026: Roth lab tagged-receptor engineered mice + radioligand binding + kinetic analyses show no direct 5-HT2A-mGluR2 physical interaction in vitro or in vivo despite mGluR2 agonism attenuating DOI head-twitch — presynaptic-glutamate-inhibition model is the surviving mechanism, 5-HT2A-mGluR2 heterodimer target concept needs a hard reread; MGH/MADRC cerebrovascular snRNA-seq of postmortem ITG in AD + CBD + Pick's + PSP — largely disease-specific vascular transcriptomes but heat-shock upregulation shared across all four proteinopathies + AD-specific vascular remodeling + AD GWAS risk loci dysregulated in endothelial cells; NPTX2-centered cognitive resilience — 575-sample MTG RNA-seq + PRM-MS proteomics show NPTX2 in tight plasticity cluster (BDNF+VGF+SST+SCG2) inversely coupled to lysosomal-and-chromatin-stress axis + stress-linked FOXJ1+ZHX3+SMAD5+JDP2+ZIC4 over-correlate with NPTX2 protein loss in AD, coordinated plasticity-and-mitochondria program collapse rather than single-protein hit; ROSMAP Hsp60/10 mitochondrial chaperonin client protein decline stronger than RNA + stronger than abundance-matched non-clients + network-centrality-linked + inverse Braak/tau + positive cognition + mitochondrial translation + TCA-pyruvate-redox highest-priority clients; Doeller lab VR-navigation fMRI in controls + preclinical AD + early AD — amyloid-β-linked hippocampal hyperactivity + tau-linked entorhinal hexadirectional modulation reductions + altered object-centered representations, hippocampal hyperactivity earliest + only preclinical-vs-control signature; Kenny lab cerebellar miR-206 Purkinje-restricted expression + regulates translational programs controlling intrinsic excitability + deficiency increases tonic firing + elevation drives burst firing + bidirectional PPI rescue in KO + PPI impairment on WT overexpression, schizophrenia-linked microRNA tunes cerebellar Purkinje firing dynamics to control sensorimotor gating; STING restrains calcium-dependent microglial phagocytosis in kainic-acid TLE — STING deficiency amplifies microglial activation via elevated STIM1 + dysregulated SOCE + neuronal loss + cognitive impairment, pharmacological SOCE inhibition rescues, microglial calcium signaling as therapeutic axis in TLE; ventral striatal astrocyte calcium encodes reward-prediction errors for seconds post-outcome + attenuation increases decision noise + reduces win-stay + slice + in-silico modeling implicate presynaptic E-I regulation + RPE-signal-sharing across MSNs, astrocytes as component of the algorithmic apparatus of value-based decision-making; social isolation 10-week adult BALB/c males increases ventral hippocampal miR-30e-5p + reduces Neurod1 + impairs pattern separation without anxiety/depression + blunted acute-stress response — mouse mechanism for human miR-30e SNP linked to cognition/EEG/depression/schizophrenia; Microsoft HealthAgentBench 54 agentic healthcare tasks across 7 categories with Codex GPT-5.5 at ~42% overall + EHR pipeline strengths + medical imaging weakness especially Claude Code + large-search-space compositional reasoning still hard. false Receptor & Reason for June 30, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Ten items today. Opens on a Johnson lab Kentucky paper using EFAD mice — humanized APOE on the 5xFAD amyloidosis background — to compare anti-amyloid therapy across E2, E3, and E4 isoforms head to head. Nine-month-old mice get twelve weeks of chimeric aducanumab or IgG. E4 mice clear plaques like the others but accumulate cerebral microhemorrhages, with microglia and astrocyte reactivity concentrated in the perivascular compartment. Single-cell and spatial transcriptomics on the neurovascular unit show loss of vascular plasticity plus coordinated inflammatory and immune upregulation that E2 and E3 do not show. The conclusion: the E4 cerebrovasculature is uniquely susceptible to antibody-mediated vascular damage, and perivascular immunological targets are where ARIA risk mitigation has to live. Then a complementary mechanistic story in Tg2576 cerebral amyloid angiopathy — mTOR activation drives fibrillar vascular Aβ accumulation, endothelium-dependent cerebrovascular reactivity loss, tight-junction-mediated BBB breakdown, increased microhemorrhages, and nNOS-dependent neurovascular uncoupling. Rapamycin reverses cognitive deficits in contextual memory, fibrillar Aβ, vasodilation, microhemorrhages, and neurovascular coupling in the same model — connecting mTOR mechanistically to both ARIA-relevant vascular pathology and the cognitive phenotype. Then the Zoghbi group's CHIP-LMO7-BAG5 paper — cross-species genetic screen identifies LMO7 as a tau-level regulator. LMO7 bridges the E3 ligase CHIP and its inhibitory cochaperone BAG5, holding CHIP from ubiquitinating tau. Knocking down LMO7 in adult tauopathy mice unleashes CHIP, drops total and phospho-tau, attenuates gliosis, and rescues memory. AlphaFold-derived structural models of the complex plus AI-driven virtual screening nominate two FDA-approved drugs that disrupt the LMO7-BAG5 interaction: telmisartan, the angiotensin II receptor blocker that's been at the edge of CNS repurposing, and an optimized mycophenolic acid derivative, MPA-2. Oral telmisartan or MPA-2, or viral delivery of a CHIP-derived competing peptide, reduces pathogenic tau in tauopathy mice without overt toxicity. CHIP-LMO7-BAG5 is a druggable node relevant beyond AD to STUB1-associated ataxias. Then cryo-EM of hippocampal AMPA receptors showing how PRRT1, also known as SynDIG4, gates GluA1's role in LTP. PRRT1 associates preferentially with GluA1 via its membrane-embedded CD225 domain and sequesters the GluA1 C-terminal tail through a previously unappreciated lipid modification at cysteine 825 — physically blocking the CaMKII binding site phosphorylated at Ser831. Overexpressing PRRT1 impairs LTP; mutating Cys825 to serine rescues LTP. The lipid switch at Cys825 is the cleanest structural picture yet for bidirectional tuning of the LTP-mobilizable AMPA pool. Then CaMKIIα phase separation is shown to be required for structural LTP, with an I205K KI mouse exhibiting complete loss of structural LTP despite normal spine morphology, marked hyperactivity, and profound aversive memory deficits. Atomoxetine ameliorates the hyperactivity phenotype. The authors identify a human patient with an I205N variant presenting with ADHD and mild intellectual disability, and additional pocket variants destabilize the CaMKII-GluN2B condensate through distinct MD-trajectory mechanisms. CaMKII phase separation is now causally upstream of spine plasticity, hyperactivity, and memory — with a defined molecular handle for an ADHD and intellectual disability variant subset. Then Btbd11 — interneuron-specific — forms a phase-separated condensate with PSD-95 and TARP-γ2, stabilizing TARP-γ2 and GluA1 within glutamatergic interneuron synapses. Btbd11 deletion shrinks the PSD, reorganizes PSD-95 nanoclusters, and drastically reduces glutamatergic input onto interneurons. Interneuron-specific PSD architecture as a therapeutic target distinct from principal-neuron synapses — relevant because most schizophrenia hypotheses run through interneuron hypofunction. Then the Tseng-Akil-Watson meta-analysis of antipsychotic frontal-cortex transcriptomic effects across rodent and NHP datasets — 63 DEGs in rodents (117 with NHP) enriched for oligodendrocyte/myelination, cellular stress, and cardiovascular pathways. The headline result: comparison to post-mortem schizophrenia signatures shows the antipsychotic-induced signature overlaps a fraction of what's been attributed to schizophrenia biology — treatment status has to be modeled, not just adjusted for. Then PTSD GWAS fine-mapping landing on FOXP2 in amygdala intercalated cells — knocking FOXP2 down in mouse ITCs increases excitability and reduces K+ channel conductance, producing less freezing during and after fear conditioning. Bulk RNA-seq confirms decreased K+ channel transcription and broader CRH/Wnt/neurokinin regulation; downstream gene set is enriched for orthologs of tier-one PTSD GWAS risk genes; human PTSD post-mortem medial amygdala with low FOXP2 shows matching K+ channel downregulation. The directionality is opposite of a naive risk-equals-more-fear model. Then a methodological caution: RS67333, widely used as a 5-HT4 partial-agonist tool compound, drives striatal DA and ACh effects in mouse slices through acetylcholinesterase inhibition rather than 5-HT4 — FSCV plus GRABACh3.0 plus a structurally-distinct 5-HT4 ligand (BIMU8) without AChE activity confirm this. The RS67333 striatal literature needs revisiting, and RS67333 may be useful as a CNS-penetrant AChE-inhibitor scaffold distinct from donepezil and rivastigmine. Closes on UCSF Clinical Reasoning Graphs — structured graph representations extracted from 750 frontier-LLM diagnostic traces across 50 NEJM CPC cases. Graph-similarity is 0.488 for correct pairs vs 0.484 for incorrect pairs; no comparison survives multiple-testing correction. Discriminating-feature reflection prompting moves things — +33% on that subtask. The honest summary, and the title of the paper, is competence without consistency — the accuracy is real but the reasoning underneath isn't stable across clinically similar cases. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. https://github.com/andrewsu/ai-nuggets/tree/main/podcasts/receptor-and-reason 2026-06-30-receptor-and-reason Tue, 30 Jun 2026 12:00:00 +0000 1095 Daily roundup for June 30, 2026: Johnson lab Kentucky EFAD-mouse anti-amyloid therapy comparing E2/E3/E4 — APOE4-specific ARIA from perivascular microglia + astrocyte reactivity + neurovascular-unit single-cell + spatial transcriptomics showing E4-only vascular-plasticity loss + inflammatory/immune upregulation, perivascular immunological targets as next-gen ARIA mitigation axis; mTOR drives Tg2576 CAA cerebrovascular dysfunction + BBB breakdown + neurovascular uncoupling all reversed by rapamycin with contextual-memory rescue, mTOR linked mechanistically to both ARIA vascular pathology + cognitive deficit; Zoghbi cross-species screen + AlphaFold + virtual screen — LMO7-CHIP-BAG5 disruption by oral telmisartan + MPA-2 + CHIP-derived viral peptide clears tau + rescues memory in tauopathy mice; PRRT1/SynDIG4 cryo-EM — Cys825 lipid modification sequesters GluA1 C-terminal tail blocking CaMKII Ser831 site, C825S rescues LTP, PRRT1 overexpression impairs LTP — bidirectional handle on LTP-mobilizable AMPA pool; CaMKIIα I205K phase-separation-deficient KI mouse — loss of structural LTP + hyperactivity + memory deficits + human I205N ADHD-ID variant + atomoxetine rescue + additional pocket variants destabilize CaMKII-GluN2B condensate; Btbd11 interneuron-specific phase-separated condensate with PSD-95 + TARP-γ2 at glutamatergic interneuron synapses — KO shrinks PSD + reorganizes nanoclusters + reduces glutamatergic input onto interneurons, scaffold biology distinct from principal-neuron synapses; Tseng-Akil-Watson antipsychotic frontal-cortex transcriptome meta-analysis — 63→117 DEGs enriched for oligodendrocyte/myelination + cellular stress + cardiovascular, overlaps post-mortem schizophrenia DEGs → treatment status must be modeled not adjusted; FOXP2 GWAS fine-mapping to amygdala ITCs — FOXP2 KD increases excitability + decreases K+ channel transcription + reduces freezing (counter-intuitive direction), downstream gene set enriched for PTSD GWAS orthologs + matched in human post-mortem medial amygdala; RS67333 5-HT4 tool compound drives striatal DA + ACh effects via AChE inhibition not 5-HT4 (FSCV + GRABACh3.0 + BIMU8 control) — striatal RS67333 literature requires reinterpretation + alternative reading as CNS-penetrant AChE-inhibitor scaffold; closes on UCSF Clinical Reasoning Graphs from 750 frontier-LLM diagnostic traces across 50 NEJM CPC cases — graph-similarity 0.488 correct vs 0.484 incorrect + no comparison survives MTC + discriminating-feature reflection prompting yields +33%, competence-without-consistency as operational warning for diagnostic-aide deployment. false Receptor & Reason for June 29, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens on a Mayo Clinic transcriptomic dissection from the Petrucelli, Prudencio, and Keith Josephs groups: TDP-43 pathology shows up in roughly half of advanced AD cases but always co-occurs with tau and amyloid. They generate regionally resolved transcriptomic profiles across pure AD, AD with limbic-predominant age-related TDP-43 encephalopathy (the AD/LATE phenotype with types α and β), FTLD-TDP (types A and B), and cognitively normal controls, using quantitative phospho-TDP-43 and phospho-tau as continuous covariates to decouple pathology-specific signals. TDP-43 in AD/LATE drives transcriptomic programs largely uncoupled from tau burden, and those programs are not the same as the FTLD-TDP programs. The amygdala is the regional convergence point of TDP-43-associated remodeling across both diseases; frontal cortex remodeling is largely restricted to FTLD. Subtype stratification within each disease unmasks immune-activation and cellular-vulnerability trajectories that completely disappear when subtypes are lumped together. The take for AD trial design: any program stratifying by amyloid and tau alone is collapsing biologically distinct patient subsets, and TDP-43 morphology should be a stratification variable for limbic-onset cognitive decline. Then an ALS-associated TARDBP K181E mutation in iPSC-derived forebrain organoids: no TDP-43 overexpression, no exogenous stress, and the mutant organoids spontaneously develop hyperphosphorylation, cytoplasmic accumulation of p-TDP-43, RNA dysregulation, pro-inflammatory signaling, and apoptotic activation. scRNA-seq plus enhanced CLIP shows that K181E alters TDP-43's RNA-binding specificity — meaning the mutant protein actively mis-targets transcripts before any aggregation manifests, a gain-of-altered-binding model rather than the dominant loss-from-aggregation model. The downstream is widespread mis-splicing including cryptic exon inclusion, and the new biomarker candidate is PRDM2 — cryptic-exon-derived PRDM2 peptide immunoreactivity is detectable in ALS spinal motor neurons co-localized with p-TDP-43. Then a CO2 sensor in the dorsal raphe-to-VTA pathway: Connexin 26 on serotonergic terminals projecting DR→VTA, brief PCO2 elevations open Cx26 and modulate VTA-DA excitability through serotonergic signaling without touching glutamatergic transmission, and a GRAB5HT-sensor experiment shows Cx26 is directly permeable to serotonin — a channel co-synapse. Selectively removing CO2 sensitivity from DR-serotonergic Cx26 in vivo delays hypercapnic arousal from sleep and reduces VTA-DA activation. Implications across SUDEP, sleep apnea pharmacology, and a non-canonical neurotransmitter release mode worth thinking about for selective serotonergic modulation. Then a Parkinson's circuit-pathology paper on the Synj1 R258Q mutation that selectively impairs the Sac1 phosphatase domain: synaptic vesicle trafficking defects across the brain but selective dystrophic phenotype — focal axonal dilations with onion-like DAT-enriched plasma-membrane infoldings in a vulnerable subset of dorsolateral striatal DA axons, reconstructed by correlative light microscopy plus FIB-SEM, and a regional dopamine-release deficit. Early-onset familial Parkinsonism as endocytic-recycling catastrophe — drug-target axis in the endocytic machinery rather than downstream of α-synuclein. Then mitochondrial turnover at central GABAergic synapses governing epilepsy vulnerability: MIRO1 or TRAK1 conditional knockout stalls mitochondrial turnover at axon terminals of somatostatin-positive interneurons, drives cristae loss, reduces synaptic GABA, destabilizes network oscillations, and produces recurrent seizures and premature death — and post-weaning gene therapy reverses the mitochondrial alterations and rescues the epileptic phenotype. Mitochondrial transport as a defined drug-target axis for genetic epilepsies, with a developmental-window reversibility. Then prenatal circadian rhythm disruption from a light-cycle reversal during gestation in C57BL/6J: female offspring develop an anhedonic phenotype with decreased food self-administration + cocaine intake + reinforcement + increased anxiety, male offspring develop the opposite — increased cocaine preference + reinforcement + decreased anxiety + increased risk-taking + decreased FST immobility, with sex-divergent corticosterone rhythm disruption. Pregnancy-window shift work as a developmental psychiatric risk + sex-divergent HPA-axis programming as a cleaner hypothesis than maternal-stress-equals-bad. Then discriminative Pavlovian goal-tracking memory reconsolidation in Lister hooded rats: propranolol (10 mg/kg) impairs subsequent goal-tracking regardless of reminder type and consistently across sexes, MK-801 (0.1 mg/kg) impairs discrimination with both non-reinforced and reinforced reminders. Restores the rodent foundation for the cue-exposure-plus-propranolol clinical program in addiction cravings. Then turning to agentic AI — the Sinitskiy et al. evaluation of five advanced AI research frameworks (Kosmos, K-Dense, ToolUniverse, BioAgents from bio.xyz, AI Scientist-v2 from Sakana) on three real-life tasks (uncertainty quantification for molecular property predictions, ML on Therapeutic Data Commons, agent-based modeling), including two recently published papers as the gold standard. Genuine strengths: hypothesis generation, routine data acquisition + coding, statistical confidence often absent from originals, well-formatted reports. Real failures: no framework matched the scope or depth of the original studies, results varied across multiple runs with the same prompt, severe hallucinations in final reports, literature-coverage gaps, overconfident conclusions, and verification required substantial domain expertise. Useful for prototyping directions and stress-testing completed studies, not a substitute for a competent group. Then the pairing piece: Clair3-Connect, a client-server re-architecture of the long-read variant caller. Schema-defined agentic tools with feature tensors crossing the boundary while identifiable data stays on the client, mutual TLS plus AES-256-GCM. 60/60 APOE diplotyping runs correct. 12K tokens in 3 turns vs 81K-163K for shell-driven baselines (6.8-14× fewer tokens), about a quarter the wall-clock, 4% vs 35% token-usage variation, 7.2% encryption overhead, SNP F1 within 0.1-0.3 points of standard Clair3 at 50× coverage. Agentic interfaces as a first-class deliverable of bioinformatics tool development, not a wrapper retrofit. Closes on ISTP Tech's DPISO DrugEngine — state-space compression plus a Discrete Phase-Interference Search Operator targeting intrinsically disordered α-synuclein NAC residues 61-95. Compresses an ~8.46×10⁸-molecule mirror to ~10⁷ active set (~85× reduction) on a single desktop workstation, then to a commercially-available short list. 3/3 prospective inhibitor-call concordance — 2-D08, uralenol, and herbacetin (all natural-product-adjacent polyphenols) all suppress α-synuclein aggregation in a ThT assay at 100 µM with EGCG as positive control, two candidates achieving ≥80% plateau reduction. Proof-of-concept that a state-space-compressed virtual screen against a hard intrinsically-disordered target can deliver experimentally validated hits without HPC. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-29-receptor-and-reason Mon, 29 Jun 2026 12:00:00 +0000 1155 Daily roundup for June 29, 2026: Petrucelli/Prudencio/Josephs Mayo regionally-resolved transcriptomics decoupling TDP-43 subtypes (AD/LATE α/β + FTLD-TDP A/B) from tau burden in AD — autonomous subtype-dependent transcriptional landscapes + amygdala shared remodeling + frontal cortex FTLD-restricted + immune activation + cellular vulnerability trajectories unmasked only by subtype stratification, argues TDP-43 morphology as trial-enrichment variable for limbic-onset cognitive decline; TARDBP K181E iPSC forebrain organoid model — spontaneous p-TDP-43 + cytoplasmic accumulation + altered RNA-binding specificity (eCLIP + scRNA-seq) + cryptic exon inclusion + PRDM2 cryptic-peptide immunoreactivity in ALS spinal motor neurons, gain-of-altered-binding-specificity model beyond loss-from-aggregation; CO2-sensitive Cx26 hemichannels on dorsal-raphe-to-VTA serotonergic terminals — channel co-synapse directly permeable to 5-HT (GRAB5HT sensor) modulates VTA-DA excitability + selective DR-Cx26 CO2-sensitivity removal delays hypercapnic arousal + reduces VTA-DA activation, SUDEP + sleep apnea pharmacology target; Synj1 R258Q dystrophic nigrostriatal axons — Sac1-phosphatase-impaired endocytic catastrophe + onion-like DAT-enriched plasma membrane infoldings in vulnerable DA axons (FIB-SEM) + regional dopamine-release deficit, endocytic-recycling drug-target axis for familial PD upstream of α-synuclein; MIRO1/TRAK1 mitochondrial-transport KO stalls turnover at somatostatin GABAergic terminals — cristae loss + reduced synaptic GABA + recurrent seizures + premature death, post-weaning gene therapy rescues, mitochondrial transport as drug-target for genetic epilepsies; prenatal CRD light-cycle reversal C57BL/6J — anhedonic phenotype in females + SU-like phenotype in males + sex-divergent corticosterone, pregnancy-window shift work as sex-divergent HPA-axis psychiatric risk; Pavlovian goal-tracking memory reconsolidation in Lister-hooded rats — propranolol 10 mg/kg + MK-801 0.1 mg/kg both impair subsequent discrimination regardless of reminder type, restores rodent foundation for cue-exposure-plus-propranolol clinical addiction program; Sinitskiy-et-al evaluation of five AI research frameworks (Kosmos + K-Dense + ToolUniverse + BioAgents bio.xyz + AI Scientist-v2 Sakana) on three real-life tasks — hypothesis generation + routine code + statistical confidence + formatting strengths, run-to-run variance + hallucinations + literature gaps + overconfidence + verification-needs-expertise weaknesses, prototyping/stress-test tools not researcher replacement; Clair3-Connect client-server agentic interface to long-read variant calling — 60/60 APOE diplotyping correct + 12K tokens in 3 turns vs 81K-163K shell baselines (6.8-14× fewer) + 25% wall-clock + 4% vs 35% token variation + 7.2% encryption overhead + SNP F1 within 0.1-0.3 of standard Clair3 at 50× coverage, agentic interfaces as first-class bioinformatics-tool deliverable; closes on ISTP DPISO DrugEngine state-space compression on intrinsically-disordered α-synuclein NAC residues 61-95 — ~8.46×10⁸→~10⁷ active set ~85× reduction on a single desktop + 3/3 prospective inhibitor-call concordance for 2-D08 + uralenol + herbacetin + EGCG positive in ThT aggregation assay at 100 µM + two compounds ≥80% plateau reduction, virtual screen against hard IDP target validated experimentally without HPC. false Receptor & Reason for June 28, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens on a Carhart-Harris/Deco/Kringelbach methodological reframing of antidepressant comparison using data from the Imperial psilocybin-vs-escitalopram MDD trial: whole-brain models fit to pre/post resting fMRI yielding generative effective connectivity matrices, then perturbed systematically and locally, re-optimized into response GEC matrices, with a network reconfiguration index quantifying how far the brain's effective wiring moves under stress. Global NRI rises after psilocybin (more reorganizable, more flexible) and falls after escitalopram (more stabilized). Both arguably therapeutic but not the same kind — destabilize a stuck attractor vs. fortify an alternative one. Portable methodology applicable to any treatment with pre/post imaging. Then the Jun Wang group's circuit map for fentanyl withdrawal's hypodopaminergic state: naloxone-precipitated withdrawal recruits CRF neurons in central amygdala and BNST, those CRF neurons make monosynaptic excitatory connections onto dorsostriatal striosomal medium spiny neurons, CRF acting through CRFR1 strengthens glutamatergic drive onto striosomes, in vivo dopamine photometry shows withdrawal amplifies striosome-mediated suppression of dopamine release prevented by CRFR1 antagonism. CRFR1 antagonists have repeatedly missed in stress-disorder trials but a striosome-specific opioid-withdrawal indication with this circuit map is a sharper retry. Then circuit-level pharmacology of levodopa-induced dyskinesia: optogenetic dissection of MSN-MSN lateral inhibition in dorsolateral striatum reveals D2-to-D1 lateral inhibition as the strongest of four connection types, weakened in 6-OHDA parkinsonian state and restored by chronic levodopa, and chemogenetic suppression of D2-originating lateral inhibition lowers the threshold for dyskinetic involuntary movements. Action selection as a local microcircuit-driven competition where dyskinesia is what happens when the suppression machinery breaks — adds a local circuit handle to LID pharmacology beyond canonical D1 supersensitivity and cortical-thalamic remodeling. Then the Harkany lab on methamphetamine-induced disruption of neuropeptide expression: high-temporal-resolution monitoring distinguishes infant hyperactivity from adult biphasic response, Fos-maps it to goal-directed cortical areas, and shows differential age-specific alteration of somatostatin, cholecystokinin, and galanin in corticolimbic areas — co-released inhibitory neuropeptides framed as candidate circuit-protective buffering against stimulant-induced hyperexcitability during critical developmental windows. Then a serotonergic-arousal story: in mouse models of anxiety and depression, exogenous 5-HTP shortens both induction and recovery from sevoflurane, TPH2-promoter-targeted DREADD activation in CeA or mPFC accelerates emergence, optogenetic activation of the CeA-to-mPFC serotonergic projection mimics the chemogenetic effect, and both pharmacological and circuit-level engagement restore delayed emergence toward baseline. Defined limbic-prefrontal 5-HT arousal pathway as target for perioperative delayed-emergence management. Then a structure-based drug design study at platelet P2Y1R that is methodologically squarely about biased agonism at a GPCR: docking against the human P2Y1R crystal identifies Val194/Lys196/Thr205 ECL2 residues that biased endogenous nucleotides engage but MRS2500 does not, rational design of nucleotide analogue KMR-82-13 selectively inhibits Gα12/13-mediated chemotaxis and inflammation while sparing Gαq-PLC aggregation and tail-bleeding, and the binding mode is transferred onto a non-nucleotide scaffold KSN-159-27 with 200ns MD validation as a pathway-selective inverse agonist. The same logic generalizes to next-generation 5-HT2A psychedelic and kappa-opioid analgesia programs. Then a comp-pharm theory paper that should sharpen PKPD endpoint selection: Ehrlich occupancy time, the time-integral of fractional target occupancy from Ehrlich's 1913 principle, with analytical bounds for open systems with first-order drug elimination, induced-fit kinetic trapping through reduced effective Kd, and Copeland's 1/koff residence time recovered as the rebinding-free limit. Joint affinity-and-clearance optimization, not k-off alone. Then the Dorrestein/Wang/Northen GNPS2 agentic AI for drug-metabolite structural elucidation from LC-MS/MS: an LLM probabilistically hypothesizes structures beyond training data, wrapped in spectral alignment, molecular formula inference, rule-based structural enumeration, ML spectrum prediction, and natural-language expert hypotheses to constrain hallucination. Experimentally confirmed novel metabolites: a phosphorylated hydroxyzine metabolite, an acetaminophen-p-coumaric-acid ester, and two oxidative ibuprofen-carnitine conjugates from public repositories. LLM proposes, domain tool validates, expert constrains — working agentic template for structural bioinformatics. Then an OECD-anchored pharmacological- and assay-readout-stratified curation framework for opioid bioactivity databases: 115,232 raw records reconciled into 50,977 curated measurements across 19,585 compounds spanning MOR, DOR, KOR, and NOP from ChEMBL, BindingDB, PubChem, and the IUPHAR Guide to Pharmacology. KOR pKi R²=0.79, DOR pKi R²=0.77, MOR-antagonist functional inhibition R²=0.86. Pooled baselines silently train on binding-displacement records while reporting functional IC50 — predicting affinity while pretending to predict function. The dataset contract, not model architecture, defines what a QSAR claim can mean. Closes on BrainAgent, a multi-agent LLM framework for autonomous brain-signal understanding: an orchestrating LLM decomposes EEG/BCI workflow into specialized preprocessing, feature-extraction, task-identification, and classification agents, competitive with paradigm-specific traditional ML across motor-imagery, P300, and emotion-classification benchmarks and substantially more flexible at workflow assembly. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-28-receptor-and-reason Sun, 28 Jun 2026 12:00:00 +0000 1007 Daily roundup for June 28, 2026: Carhart-Harris/Deco/Kringelbach perturbation-induced brain reconfiguration in MDD — whole-brain generative effective connectivity models fit to pre/post resting fMRI + systematic + local artificial perturbations + re-optimized response GEC + network reconfiguration index, global NRI rises after psilocybin (more reorganizable) + falls after escitalopram (more stabilized), destabilize-vs-stabilize antidepressant-action framing using Imperial psilocybin-vs-escitalopram trial data; Jun Wang-lab fentanyl-withdrawal extended-amygdala CRF → dorsostriatal striosome → dopamine pathway — naloxone-precipitated withdrawal recruits CeA + BNST CRF neurons + monosynaptic excitation of striosomal MSNs + CRFR1-dependent postsynaptic glutamate strengthening + in-vivo DA photometry confirms amplified striosome-mediated suppression prevented by CRFR1 antagonism, sharper retry indication for CRFR1 antagonists; striatal MSN-MSN lateral inhibition + LID action-selection — optogenetic dissection of four connection types in 6-OHDA mouse, D2-D1 strongest + weakened parkinsonian + restored chronic levodopa, chemogenetic D2-lateral-inhibition suppression lowers dyskinesia threshold, local circuit handle on LID beyond canonical D1 supersensitivity; Harkany-lab methamphetamine-induced neuropeptide disruption — high-temporal-resolution infant hyperactivity monitoring + Fos-mapped goal-directed cortical activation + differential age-specific somatostatin + cholecystokinin + galanin alteration in corticolimbic areas, co-released inhibitory-neuropeptide buffering as developmental-vulnerability mechanism; 5-HT CeA-mPFC sevoflurane-emergence pathway — exogenous 5-HTP + TPH2 hM3Dq/hM4Di + ChETA optogenetic activation in anxiety/depression mouse models restores delayed emergence, perioperative serotonergic limbic-prefrontal pathway; King's-College P2Y1R pathway-selective inverse agonist — docking against PDB 4XNW + Val194/Lys196/Thr205 ECL2 residue engagement + nucleotide analogue KMR-82-13 + non-nucleotide MD-validated scaffold KSN-159-27 selectively inhibit Gα12/13 chemotaxis + LPS lung inflammation while sparing Gαq aggregation + bleeding, biased-agonism rational-design loop generalizing to 5-HT2A + KOR; Ehrlich occupancy time framework — Ehrlich 1913 "Corpora non agunt nisi fixata" as time-integral of fractional occupancy + analytical bounds for open systems with elimination + induced-fit kinetic trapping + Copeland's 1/koff recovered as rebinding-free limit, joint affinity-and-clearance PKPD endpoint; Dorrestein/Wang/Northen GNPS2 agentic AI for drug-metabolite LC-MS/MS structural elucidation — LLM-proposes + domain-tool-validates + expert-constrains workflow + experimental confirmation of phosphorylated hydroxyzine + acetaminophen-p-coumaric-acid ester + ibuprofen-carnitine conjugates from public repositories; OECD-anchored opioid bioactivity database stratification (MOR/DOR/KOR/NOP) — 50,977 curated measurements across 19,585 compounds + KOR pKi R²=0.79 + MOR-antagonist functional R²=0.86, pooled-baseline silently trains on binding-displacement records while reporting functional IC50, dataset contract defines QSAR claim validity; closes on BrainAgent — multi-agent LLM EEG/BCI workflow with orchestrating LLM decomposing into preprocessing + feature-extraction + task-identification + classification agents, competitive with paradigm-specific ML + workflow-assembly flexibility for clinical BCI deployment. false Receptor & Reason for June 27, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens on a clean within-subject crossover from the Jocham group: 62 healthy participants did a Bayesian-style perceptual decision-making task under biperiden (muscarinic ACh antagonist), propranolol (β-adrenergic antagonist), and placebo. Neither drug changed how much choices were governed by sensory evidence; both sped updating of prior beliefs in response to new evidence. Propranolol's effect was specific to reward outcomes — noradrenaline gating a reward-specific channel; biperiden sped updating after both reward and non-reward, producing less stable beliefs over time — muscarinic ACh setting the overall prior-updating learning rate. Then a Rauch-lab Alzheimer's drug-discovery story: a quantitative platform around LRP1 ligand-binding-domain 4 (BD4) — the tau-interaction region of the endocytic receptor that mediates tau uptake and propagation — with three orthogonal binding assays (fluorescence polarization, split luciferase complementation, TR-FRET) agreeing on nanomolar tau affinity and competitive displacement by tau/RAP/peptide. HTS pulls candidate inhibitors that phenocopy competitive blockade in a cellular tau-uptake assay. ADDF-funded; upstream-of-aggregation drug-discovery handle. Then the Akassoglou-lab fibrin-inflammatory-domain-blockade story extended to 5XFAD: the Fggγ390-396A knock-in rescues high-risk cost-benefit-conflict decision-making, in-vivo 2P + correlative light/EM restore microglia-neuron contacts near Aβ plaques, snRNA-seq on 165,252 nuclei recovers SLC1A3 glutamate sensing + CD39-adenosine + GAS6-Tyro3 ligand-receptor pathways between microglia and neurons, and in 57 AD patient CSF samples profiled with SomaScan, fibrinogen correlates with complement + coagulation + inflammation + synaptic-marker proteins. Translational bridge complete: vascular-derived driver of the microglial state disrupting neuronal networks in AD without breaking fibrin's clotting function. Then a CSF1R-targeted molecular MRI probe — binds extracellular Ig domain, actively taken up by murine + human microglial lines, separates A53T α-synuclein PD mouse model from controls via radiomics-assisted MR analysis with IHC confirmation. Candidate non-invasive PD biomarker avoiding TSPO PET-tracer pharmacogenetic and specificity issues. Then a fast variant-specific epilepsy assay: mRNA overexpression of SCN2A R1882Q + R853Q + SCN8A R1872Q gain-of-function variants in 3-dpf zebrafish larvae produces measurable sporadic seizure activity that is knocked down by topiramate, GS967, and PF-04856264 — the latter two being persistent-sodium-current-selective blockers, exactly the right mechanism for these channelopathies. Then in-vivo CRISPR/Cas9 KO screening in chicken nucleus magnocellularis with an all-in-one triple-guide vector: of fourteen microtubule-associated candidates, GSK3β and Tau are required for developmental axon-initial-segment shortening, constitutively active GSK3β overexpression drives shortening, and microtubule stabilization abolishes the effect — a mechanistic answer to what lithium's primary target does at the AIS-mediated intrinsic excitability machinery. Then VTA dopamine projection-specific reward prediction: fiber photometry + cell-type/pathway-specific optogenetic inhibition + Pavlovian + decision-making probes in rats show cue-evoked VTA-DA→BLA activity encodes predicted-reward value, mediates cue-driven action bias + current-value tracking (goal-directed control), and constrains new learning from cue-paired outcomes. Then a class-B1 GPCR methods piece: SMA 2000 vs. DIBMA-12 vs. electroneutral sulfo-DIBMA copolymer encapsulation of CGRP and PTH1 receptors in native-lipid nanodiscs via NanoBRET ligand-binding + mini-Gα recruitment — all three preserve extracellular ligand binding but only sulfo-DIBMA preserves mini-Gαs engagement + G-protein-dependent allosteric modulation with prolonged residence time. Polymer charge + backbone chemistry, not extraction yield, determines whether the active-state conformational landscape survives solubilization. New default polymer for class-B1 GPCR biophysics and biased-agonist programs. Then an arXiv audit of framing-sensitive behavioral instability in instruction-tuned LLMs for mental-health interactions: matched semantically-equivalent prompts under different contextual framings systematically alter interpretive response, layer-wise probing shows behavior-associated information decodable through transformer depth, and activation-steering experiments suggest the framing-associated representational directions are partially causal. Robustness to framing is a deployment criterion for psychiatry-adjacent conversational AI, and current models don't have it. Closes on a multisite resting-state EEG dataset (n=2,874, ages 5-18, ADHD/ASD/anxiety/LD/controls): repeated case-control comparisons across sample sizes show small samples produce inflated, unstable effect sizes; large samples converge on uniformly small but robust effects; power rises steeply with sample size; false positives are stable. The EEG counterpart of the BWAS sample-size lesson — case-control biomarker discovery in pediatric psychiatry needs sample sizes the field has not been routinely using. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-27-receptor-and-reason Sat, 27 Jun 2026 12:00:00 +0000 887 Daily roundup for June 27, 2026: Jocham-lab biperiden vs. propranolol within-subject crossover (n=62) — muscarinic ACh sets the overall prior-updating learning rate, β-adrenergic noradrenaline gates a reward-specific channel of the same updating process, neither drug touches sensory-evidence weighting; Rauch-lab tau-LRP1 small-molecule HTS — LRP1-BD4 + three orthogonal nanomolar assays + cellular tau-uptake-reducing hits as upstream-of-aggregation drug-discovery handle; Akassoglou-lab Fggγ390-396A fibrin-inflammatory-domain block in 5XFAD — rescues cost-benefit-conflict risk-taking + restores microglia-neuron contacts near Aβ plaques + snRNA-seq (165k nuclei) recovers SLC1A3 + CD39-adenosine + GAS6-Tyro3 microglia-neuron ligand-receptor signaling + AD CSF SomaScan correlates fibrinogen with complement + coagulation + inflammation + synaptic markers; CSF1R-targeted microglia-specific molecular MRI probe separates A53T α-synuclein PD mouse from controls via radiomics — non-invasive PD biomarker avoiding TSPO PET issues; mRNA-overexpression zebrafish assay for SCN2A R1882Q/R853Q + SCN8A R1872Q gain-of-function epilepsy variants — topiramate + GS967 + PF-04856264 persistent-Na-current-selective blockers all knock activity down, variant-specific ASM triage front end; in-vivo CRISPR/Cas9 KO screen in chicken nucleus magnocellularis — GSK3β + Tau required for developmental AIS shortening + constitutively active GSK3β OE drives shortening + microtubule stabilization abolishes it, mechanistic answer for lithium's primary target at AIS-mediated intrinsic excitability; VTA-DA→BLA projection-specific fiber photometry + optogenetics + Pavlovian/decision tasks — cue-evoked dopamine encodes predicted-reward value + mediates cue-driven action bias + current-value tracking + constrains new learning; sulfo-DIBMA vs. SMA 2000 vs. DIBMA-12 NanoBRET comparison on CGRP + PTH1 class-B1 GPCRs — only electroneutral sulfo-DIBMA preserves mini-Gαs engagement + G-protein-dependent allosteric modulation, polymer charge + backbone chemistry not extraction yield determine active-state preservation; arXiv framing-sensitive behavioral instability audit of LLMs in mental-health interactions — matched semantically-equivalent prompts systematically alter interpretive response, layer-wise probes + activation steering suggest framing-associated representational directions are partially causal, robustness to framing as deployment criterion; closes on multisite resting-state EEG dataset (n=2,874, ages 5-18) repeated case-control subsampling — small samples produce inflated unstable effects, large samples converge on uniformly small robust effects, EEG counterpart of BWAS sample-size lesson for pediatric-psychiatry biomarker discovery. false Receptor & Reason for June 26, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens on a small but tactically interesting Allison-lab CNS polypharmacy paper in a larval zebrafish TBI model: levetiracetam given immediately post-injury blocks tau aggregation and cell death (IC50 ≈3.17 µM), JNJ-42153605 (mGluR2 PAM, Janssen schizophrenia program tool compound) is itself more potent (IC50 ≈87 nM), and a subeffective dose of the PAM (10⁻⁵ mM) shifts LEV potency ~16-fold — repurposing a stalled mGluR2 PAM as an acute synaptic-vesicle-AED sensitizer in the post-TBI window, warranting closed-head mammalian-model + longer-disease-course follow-up. Then a vitamin-D3 presupplementation paper in the MK-801 mouse model of schizophrenia: 500 IU/kg/day VD3 normalizes prefrontal NR1/NR2A/NR2B + α7nAChR + PSD-95 + calcineurin + BDNF/NGF, rescues hyperlocomotion + anxiety + working memory, calcitriol (1 µM) directly potentiates α7nAChR single-channel current amplitude in HEK293T — direct allosteric ligand-gated-ion-channel modulation by a nutraceutical is the testable claim, fits the NMDA-hypofunction + α7 cognitive-symptom framework, Indian SCZ-cohort VD-deficiency translational hook. Then GPR65 trafficking-required signaling in fibroblast-like synoviocytes: pharmacological and genetic internalization blockade abolishes nuclear ERK + the transcriptional cytokine program, and conditioned media from internalization-blocked FLS loses ability to sensitize DRG sensory neurons — endosomal signaling, not plasma-membrane signaling, drives GPR65 pro-nociceptive output; biased-agonist design at this proton-sensing GPCR has to target the trafficking step itself. Then a Schmidt-lab single-nucleus RNA-seq atlas of the rat nucleus accumbens after 10-day volitional IV cocaine self-administration vs. yoked saline (36,766 nuclei): D1 Ebf1+ MSNs account for ~40% of all differentially-expressed events, followed by D2 Stk32a+ MSNs + astrocytes + D1 Ppm1e+ MSNs — cell-type-specific RAS/MAPK + NMDA-adjacent + CREB-related programs, not a uniform accumbens reward signature, atlas-of-record for subtype-selective CUD pharmacotherapy. Then a fiber-photometry sensor-saturation caveat: under acute escalating morphine in rats GCaMP6f-in-VTA-DA + dLight1.3b-in-NAc-lateral-shell show expected transient increases but high-affinity GRABDA2h transients are suppressed while whole-stream signal rises with reduced variability — GRABDA2h saturates under sustained MOR-driven release, inverting the apparent direction. Lesson: photometry pharmacology requires orthogonal-sensor cross-validation at the doses you actually use. Then allosteric β1-integrin antibody JB1a in unilateral 6-OHDA mouse PD model — pre-lesion underpowered trend, 3 days post-lesion null, 7 days post-lesion ~50% reduction in apomorphine-induced circling to sham levels — temporal specificity rules out simple acute neuroprotection but apomorphine reflects D2 supersensitivity not dopaminergic preservation and there's no histology; proof-of-concept behavioral evidence requiring α-syn-driven model + stereology before any disease-modification claim. Then TCR-repertoire-based deep-learning diagnosis of idiopathic Parkinson's: three etiologically-distinct mouse models (α-syn genetic, MPTP-like toxin, third) used as supervised anchors to molecularly stratify human iPD into two subtypes, subtype-specific multimodal multi-instance classifiers reach AUC>0.8 from peripheral blood RNA-seq — animal-model-anchored stratification as a template for trial enrichment in etiologically-heterogeneous syndromes. Then AD patient iPSC cerebral organoids (neurons + astrocytes + microglia) benchmarked on MS + SomaScan + Olink against 121 prior studies + 17,000+ plasma + CSF + cortex samples: detect almost every clinically-nominated candidate, recapitulate disease-associated change in ~1/4 of the most reproducible — spans synaptic + mitochondrial + proteostatic biology, the first quantitative multi-platform proteomic-mirror benchmark for an AD organoid system. Then a provenance-aware multi-agent system for psychiatric medication safety: Neo4j knowledge graph grounded in ATC-N + ICD-10 + MedDRA unifies 466,525 Reddit posts + 60,782 WebMD reviews + 20 years of FAERS records for 9 antidepressants; entity-recognition F1=0.97 vs. physician annotation; Reddit/WebMD Jaccard concordance up to 0.905 (more concordant with each other than with FAERS); sertraline adverse events appear in patient sources hundreds of days before corresponding FDA reports — patient-generated data as a leading-indicator independent safety signal with traceable claim provenance. Closes on iClima — ratiometric genetically-encoded Cl⁻ sensor with Kd=3.5 mM at pH 7.2, pKa=7.84 (pH-insensitive in physiological range), photostable enough for sustained in vivo imaging — detects GABA-A-evoked Cl⁻ transients in vitro, sensory-evoked transients in mouse-V1 pyramidal neurons in vivo, and simultaneous spectral-separated Cl⁻ + GCaMP6f Ca²⁺ in the same cell. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-26-receptor-and-reason Fri, 26 Jun 2026 12:00:00 +0000 804 Daily roundup for June 26, 2026: Allison-lab JNJ-42153605 mGluR2 PAM + levetiracetam polypharmacy in larval zebrafish TBI — LEV alone blocks tau aggregation (IC50 ≈3.17 µM), JNJ-42153605 alone more potent (IC50 ≈87 nM), subeffective JNJ shifts LEV potency ~16-fold, repurposes a schizophrenia-program PAM as a post-TBI synaptic-vesicle-AED sensitizer to blunt CTE/AD-like tauopathy; bioRxiv vitamin-D3 presupplementation in MK-801 SCZ mouse rescues hyperlocomotion + anxiety + working memory, normalizes PFC NR1/NR2A/NR2B + α7nAChR + BDNF/NGF + PSD-95 + calcineurin, calcitriol directly potentiates α7nAChR single-channel currents — direct allosteric ligand-gated-channel modulation fits NMDA-hypofunction + α7 cognitive-symptom framework + Indian patient VD-deficiency translational hook; bioRxiv endosomal GPR65 signaling required for nuclear ERK + cytokine transcription + DRG-sensory-neuron sensitization in fibroblast-like synoviocytes — endosomal-not-membrane signaling drives pro-nociceptive output, biased-agonist design must target trafficking; bioRxiv Schmidt-lab volitional cocaine self-administration NAc snRNA-seq (36,766 nuclei) — D1 Ebf1+ MSNs absorb ~40% of differential-expression events, D2 Stk32a+ MSNs + astrocytes + D1 Ppm1e+ MSNs follow, RAS/MAPK + NMDA-adjacent + CREB-related programs, atlas-of-record for subtype-selective CUD pharmacotherapy; bioRxiv photometry caveat — morphine raises GCaMP6f-in-VTA-DA + dLight1.3b-in-NAc-lateral-shell transients but suppresses GRABDA2h transients while raising whole-stream signal — GRABDA2h sensor saturation under sustained MOR-driven release, photometry pharmacology requires orthogonal-sensor cross-validation; bioRxiv allosteric β1-integrin antibody JB1a in unilateral 6-OHDA mouse PD — 7-day post-lesion ~50% reduction in apomorphine-induced circling to sham levels (p<0.01), temporal specificity rules out acute neuroprotection but apomorphine reflects D2 supersensitivity not preservation + no histology, proof-of-concept behavioral data; bioRxiv TCR-repertoire deep-learning iPD diagnosis — three mouse-model anchors stratify human iPD into two subtypes, subtype-specific multimodal multi-instance classifiers AUC>0.8 from peripheral blood RNA-seq — animal-model-anchored stratification as trial-enrichment template; bioRxiv AD patient iPSC organoids benchmarked on MS + SomaScan + Olink vs. 121 studies + 17,000+ plasma/CSF/cortex samples — detect almost every clinically-nominated candidate + recapitulate ~1/4 most reproducible disease-associated changes (synaptic + mitochondrial + proteostatic), first quantitative multi-platform proteomic-mirror for AD organoid systems; arXiv provenance-aware multi-agent psychiatric medication safety system — Neo4j graph unifies 466,525 Reddit posts + 60,782 WebMD reviews + 20-year FAERS for 9 antidepressants on ATC-N + ICD-10 + MedDRA, entity-recognition F1=0.97, Reddit/WebMD Jaccard up to 0.905 vs. lower FAERS overlap, sertraline AEs in patient sources hundreds of days before FDA — patient-generated data as leading-indicator independent safety signal with traceable provenance; closes on iClima ratiometric Cl⁻ sensor (Kd=3.5 mM at pH 7.2, pKa=7.84 pH-insensitive, photostable) detects GABA-A Cl⁻ transients in vitro + sensory-evoked transients in mouse-V1 pyramidal neurons in vivo + simultaneous Cl⁻ + GCaMP6f Ca²⁺ in same cell — opens KCC2/NKCC1-driven inhibitory-drive + seizure Cl⁻ dysregulation imaging. false Receptor & Reason for June 25, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens on an Emory lifespan plasma-proteomic analysis of APOE-ε4/ε4 across ages 20-90 (n=413 vs. n=2,764 ε3/ε3 controls) that identifies metabolism, GLP-1/IGF, mitochondrial, microtubule, proteostasis, and synaptic pathway alterations starting in young adulthood. STEP1 semaglutide post-hoc (n=1,311, 68-week SC semaglutide proteomics) shows 44/153 semaglutide-modulated proteins overlap with ε4/ε4-shifted proteins (p=1.6e−8) and directionally reverse the ε4 phenotype, concentrated at preclinical + clinical AD stages — IGFBP2, GDF15, TAGLN among the consistent reversers. Caveats: post-hoc cross-trial analysis, no pathway-level score survives FDR. Lines up biologically with EVOKE/EVOKE+ Phase 3 semaglutide-in-AD readouts and argues APOE-ε4 homozygosity should be a target-enrichment stratum for AD-modification trials touching metabolic biology. Then a Galovic-group NACC observational study (n=52,537) using inverse-probability-of-treatment weighting with gradient-boosted propensity scores — antiseizure-medication users show lower amyloid burden across regions (Centiloid β=−25.7 in APOE-ε4 carriers, p=0.007) + lower temporal tau-PET (META-temporal composite β=−0.05, p=0.01). Amyloid signal replicates independently in ADNI (Centiloid β=−8.6, p=0.01). Four comparator drug classes show no amyloid effect — clean negative control. Hyperexcitability-as-AD-driver hypothesis tested by quasi-experimental observational design; next step is a prospective ASM trial in AD-prodromal ε4 carriers with subclinical epileptiform activity. Then a Neuropsychopharmacology paper on GNE-5729, a GluN2A-selective positive allosteric modulator, tested in mice on Y-maze working memory + attentional set-shifting at 0.75/1.5/3.0 mg/kg. Working memory rescue is state-dependent (negative baseline-performance-by-drug-effect correlation, inverted-U signature) — significant at 1.5 mg/kg in males, baseline-modulated in females. Rescues dl-amphetamine + scopolamine challenges (neuromodulatory perturbation models) but fails to rescue MK-801 (direct channel block) — mechanistically consistent. No set-shifting effect at tested dose despite genetic implication of GluN2A in cognitive flexibility. 3 mg/kg sedates without enhancing — narrow therapeutic window. Females more sensitive (estrous-cycle not monitored). Translational read: GluN2A PAMs suited to reduced-but-preserved-glutamatergic-signaling populations (age-related cognitive decline, preclinical AD), not direct-NMDA-blockade models or full schizophrenia hypofunction. Cleanest in-vivo behavioral data on a GluN2A PAM to date. Then a paired-pulse TMS + IV midazolam study in 15 younger (18-35) vs. 15 older (50-69) adults — younger show baseline short-interval intracortical inhibition (conditioned/unconditioned MEP ratio=0.73), older lack inhibition (ratio=1.25) at baseline. Midazolam (GABA-A PAM) reduces single-pulse MEPs in both groups but abolishes paired-pulse inhibition only in younger adults. Older adults' lack of baseline inhibition + minimal midazolam response evidence age-related GABA-A-mediated cortical-inhibition decline — limits intracortical-target benzodiazepine pharmacology in elderly + invalidates pooled-age TMS biomarker work; age-stratified normative data non-negotiable. Then a Niv-lab paper on three pharmaceutical TAS2R14 agonists — Tamoxifen, Carbimazole, Lidocaine — interrogated via BRET2 G-protein recruitment + IP-One downstream signaling + targeted intracellular- and extracellular-pocket mutations. Carbimazole shows strong intracellular-pocket dependence, Lidocaine shows extracellular dependence, Tamoxifen is mostly mutation-insensitive (mixed binding mode). EC50s low-micromolar for Tamoxifen + Carbimazole; Lidocaine substantially higher. Practical implication: TAS2R14-directed extra-oral drug repurposing depends on route-of-access matching the binding pocket (inhaled for extracellular-pocket ligands, systemic for intracellular), not just receptor occupancy. Then RXFP1-relaxin: AlphaFold2 complex model (ipTM=0.93, pTM=0.76) validated against decades of mutagenesis + crosslinking data reveals extracellular-linker-trapping as the activation switch + 63-GDNNGW-68 motif as the critical interface. RFdiffusion + BindCraft yield two classes of de novo mini-protein binders. Antagonists target LRR domain — 8 leads, top two with KD 0.46-1.0 nM + functional IC50 ~5 nM. Agonists engage both LRR + linker — 4 functional activators, top one with EC50=1.49 nM + full agonism in native-receptor-expressing cells. Folds entirely unrelated to relaxin (convergent-function-divergent-architecture). High RXFP1 selectivity vs. RXFP2. Therapeutic relevance: heart failure, pulmonary fibrosis, hepatic fibrosis. Workable template for orphan GPCR design — high-confidence AlphaFold model + identify activation switch + design binders that engage it. Then biomeStat — autonomous AI agent constrained as strict deterministic orchestrator (LLM writes code that drives established bioinformatics tools rather than generating biological outputs directly). End-to-end genomic-epidemiology demonstration on 1,000 Asian dengue genomes from 16 countries (2000-2025): IQ-TREE + TreeTime phylogenetics, BEAST2 phylodynamics on NVIDIA H200 GPU, HyPhy selection-pressure analysis, PyMOL structural mapping — orchestrated in <24 h wall-clock. Findings: Re~1.0 endemic stability, 1,869 candidate immune-escape sites structurally colocalized with B/T-cell epitopes, 176 highly conserved drug-target residues across the viral replication complex, JNJ-1802 + NITD-688 antiviral binding-site residues absolutely conserved across all four DENV serotypes. Hallucination problem sidestepped by deterministic-output constraint. Then SP-Mind — arXiv paper introducing an autonomous reasoning agent for spatial proteomics that converts natural-language queries into end-to-end multiplexed-tissue-imaging-to-phenotype-discovery workflows without per-task fine-tuning. SP-Bench: 102 tasks × 18 categories spanning diverse tissue types. SOTA vs. open-source biomedical-agent baselines. Pairs with biomeStat as exemplars of the orchestrator pattern extending across omics modalities. Then RaDaR — open-source 32B-parameter reasoning LLM specialized for rare-disease diagnosis, trained on 49,170 real + 104,666 synthetic phenotype-anchored cases. Randomized AI-physician-assistance trial: RaDaR-assisted physicians show +21.44pp diagnostic-accuracy improvement over internet search. Retrospective replay: model prioritizes final diagnosis ahead of documented clinical suspicion in 61.06% of cases (~1.87mo potential lead time). Outperforms 671B DeepSeek-R1 across public benchmarks + 4 external validation centers — domain-specialization beats raw scale. Caveats: synthetic-data monotonic-scaling suggests data ceiling; AI-assistance trials have known effect-modification blind spots. Strongest clinical-trial-grade endorsement of specialized reasoning models for diagnostic support published this year. Closes on CellOS — 12B-parameter multi-view single-cell foundation model from Vitaura (Beijing) trained on 390.5M single-cell transcriptomes via 3-stage strategy: causal cell-sentence language modeling + function-preserving dense-to-mixture-of-experts expansion + latent-space alignment via LLM-JEPA objective between paired expression + perception views. Reportedly beats SOTA single-cell foundation models on cell-state annotation + batch integration + perturbation-response prediction. Multi-view + JEPA-style latent alignment as a principled response to last week's VCBench finding that linear + NN baselines beat foundation models on 4/5 testable virtual-cell dimensions; the field-relevant question is whether CellOS holds up against the same VCBench baselines, not just against other foundation models. Patent pending, weights + model card not yet released — usual industry-paper reproducibility caveats apply. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-25-receptor-and-reason Thu, 25 Jun 2026 12:00:00 +0000 1322 Daily roundup for June 25, 2026: Emory APOE-ε4/ε4 lifespan plasma proteomics (n=413 vs. n=2,764 ε3/ε3 controls) identifies GLP-1/IGF + mitochondrial + synaptic + proteostasis pathway alterations from young adulthood, STEP1 semaglutide post-hoc reverses 44/153 ε4-shifted proteins (p=1.6e−8, no pathway score survives FDR) concentrated at preclinical/clinical AD stages — biological substrate for genotype-stratified GLP-1 AD modification aligned with EVOKE/EVOKE+; Galovic NACC observational study (n=52,537) with IPTW propensity weighting links antiseizure-medication use to lower amyloid (Centiloid β=−25.7 in APOE-ε4 carriers, p=0.007) + lower temporal tau-PET (META-temporal β=−0.05, p=0.01), amyloid signal replicates in ADNI (β=−8.6, p=0.01), four comparator drug classes show no signal — quantitative framework for ASMs as AD disease-modifying candidates testing hyperexcitability hypothesis; Neuropsychopharmacology GNE-5729 GluN2A-selective PAM rescues working memory state-dependently (inverted-U signature) + rescues amphetamine + scopolamine challenges but not MK-801, no set-shifting effect, narrow dose window — GluN2A PAMs suited to reduced-but-preserved-glutamatergic-signaling populations (aging, preclinical AD), not direct-NMDA-blockade models; bioRxiv paired-pulse TMS + IV midazolam in 15 younger (18-35) vs. 15 older (50-69) adults — younger show baseline SICI (conditioned/unconditioned MEP=0.73), older lack inhibition (ratio=1.25), midazolam abolishes paired-pulse inhibition only in younger — age-related GABA-A-mediated cortical-inhibition decline + age-stratified TMS biomarker normative data non-negotiable; bioRxiv Niv-lab pharmaceutical TAS2R14 agonists Tamoxifen + Carbimazole + Lidocaine show distinct binding modes via BRET2 + IP-One + targeted mutations — Carbimazole intracellular-pocket dependent, Lidocaine extracellular, Tamoxifen mixed/insensitive — TAS2R14-directed repurposing constrained by route-of-access matching binding pocket; bioRxiv RXFP1-relaxin AlphaFold2 complex (ipTM=0.93, pTM=0.76) identifies extracellular-linker-trapping as activation switch + 63-GDNNGW-68 motif as critical interface, RFdiffusion + BindCraft yield de novo mini-protein agonists (EC50=1.49 nM full agonism native cells) + antagonists (KD 0.46-1.0 nM, IC50 ~5 nM) RXFP1-selective vs. RXFP2 — orphan-GPCR design template relevant to heart failure + pulmonary + hepatic fibrosis; bioRxiv biomeStat strict-deterministic-orchestrator agentic AI runs end-to-end IQ-TREE + TreeTime + BEAST2 + HyPhy + PyMOL on 1,000 Asian DENV genomes in <24 h — Re~1.0, 1,869 immune-escape sites colocalized with epitopes, JNJ-1802 + NITD-688 binding-site residues absolutely conserved across all 4 serotypes — agentic biology done right by constraining LLM output to deterministic tool orchestration; arXiv SP-Mind autonomous reasoning agent for spatial proteomics converts NL queries to end-to-end multiplexed-tissue-imaging-to-phenotype workflows without per-task fine-tuning, SOTA on SP-Bench (102 tasks × 18 categories) — orchestrator pattern extending to spatial-omics modality; arXiv RaDaR 32B specialized rare-disease-diagnosis reasoning LLM in randomized AI-physician-assistance trial gives +21.44pp diagnostic-accuracy improvement vs. internet search, prioritizes final diagnosis ahead of documented clinical suspicion in 61% of cases (~1.87mo lead time), outperforms DeepSeek-R1 671B across benchmarks + 4 external validation centers — domain-specialization beats raw scale; closes on CellOS 12B multi-view single-cell foundation model trained on 390.5M cells via 3-stage causal-cell-sentence + dense-to-MoE + LLM-JEPA-latent-alignment strategy reportedly beats SOTA single-cell foundation models on cell-state + batch integration + perturbation-response prediction — multi-view + JEPA-style latent alignment as a principled response to VCBench, pending validation against the same linear + NN baselines. false Receptor & Reason for June 24, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with MapLight's ML-004 Phase 2 IRIS topline in autism spectrum disorder — selective 5-HT1B/1D agonist, n=161 (102 adolescents, 59 adults), 12-week double-blind placebo-controlled, missed primary endpoint of change-from-baseline on caregiver-reported ABI–Social Communication Domain. Pre-specified analysis in adolescents with moderate-to-severe baseline irritability (ABC-I ≥16) showed care-partner-reported ABC-I effect size 1.33 (nominal p=0.013) + clinician-rated CGI-I effect size 1.08 (nominal p=0.036). Safety profile clean: no SAEs on active, no extrapyramidal events, lower mean weight gain than placebo. Heading into End-of-Phase-2 with FDA — failed-social-communication framing puts the program at risk; positive-irritability-subgroup framing positions ML-004 as a more selective alternative to aripiprazole + risperidone for ASD-associated irritability with cleaner side-effect profile. Honest read: trial was powered for a social-communication signal that didn't materialize, irritability subgroup is hypothesis-generating and needs a prospective ABC-I-primary confirmatory trial. Selective 5-HT-1B/1D agonism is emerging as a possible irritability-circuit modulator without the metabolic + motor liabilities of dopamine blockers. Then a Schnitzer-lab bioRxiv preprint that resolves the long-standing puzzle of amantadine's mechanism — dual-color two-photon Ca²⁺ imaging in 6-OHDA mice simultaneously monitors dSPN + iSPN dynamics across healthy, hypokinetic (parkinsonian), and hyperkinetic (dyskinetic) states, separating pathway-balance analysis from action-specific-ensemble analysis. L-DOPA rescues dSPN/iSPN imbalance but fails to restore the locomotion-coding ensemble; amantadine rescues the locomotion ensemble without normalizing pathway balance. Forelimb dyskinesias correspond to a distinct ensemble (not locomotion); amantadine selectively suppresses the dyskinesia ensemble's resting activity while leaving locomotion-coding cells alone. Classical pathway-balance model is the wrong lens for amantadine + probably for anti-dyskinetic pharmacology — what matters is which action-coding ensembles fire when. Drug-discovery question becomes which receptor targets selectively modulate dyskinesia vs. locomotion ensembles. Then a Surmeier-lab paper on the cholinergic-interneuron-to-GABAergic-interneuron-to-SPN arm of striatal circuitry — optogenetic ChI stimulation in healthy mice evokes GABA-A currents in both direct + indirect SPNs via nAChR-mediated activation of GABAergic interneurons, simulations suggest state-dependent control of SPN dendritic integration modulated by concurrent muscarinic signaling. In prodromal + overt parkinsonian models the circuit collapses — because the GABAergic interneurons downregulate nicotinic AChRs (not because ChIs fail). Nicotinic agonists or PAMs targeted to the relevant nAChR subunit on striatal GIs could restore lost gain control independent of dopaminergic preservation, explaining long-noted protective effects of nicotinic pharmacology in PD + nominating GI nAChRs as early-biomarker + early-intervention substrates because the disruption appears prodromally. Then the first scRNA-seq of an HD-positive human fetal striatum + age/sex-matched control — CAG-repeat expansion confirmed, 2,032 DEGs across 9 cellular clusters, gene-enrichment shows disrupted key biological processes with cluster-specific pathway dysregulation, overlap with publicly available adult HD postmortem signatures. Molecular HD pathology present during early human striatal development — supports neurodevelopmental component + earlier-intervention timeline for HTT-lowering + CAG-targeting therapies (tominersen, branaplam-style splicing modulators) including potentially prenatally in families with known expansions. n-of-one caveat, additional fetal samples needed for confirmation, but direction consistent with imaging literature. Then a tour-de-force exercise + tau paper — multi-organ spatial metabolomic atlas of long-term exercise in WT mice across 6 organs (brain, heart, lung, liver, kidney, skeletal muscle) catalogues 224 metabolic features with coordinated inter-organ remodeling; brain shows most pronounced region-specific adaptation. Extended to PS19 tauopathy: exercise reduces >70%% of observable tau pathology in PS19 hippocampus + restores mitochondrial-related metabolome. Integrated proteomic + spatial metabolomic analysis identifies NADH dehydrogenase Complex I as convergent rescue node. Biological test: AAV-Ndi1 (yeast NADH dehydrogenase that stands in for Complex I) expression in PS19 neurons without exercise increases cerebral antioxidants, restores malate-aspartate-shuttle metabolites, reduces tau pathology — recapitulates anti-tau effects of exercise. Provides molecular substrate for the human-cohort literature linking exercise to delayed cognitive decline, nominates Complex I augmentation as drug-tractable exercise-mimetic axis (NAD precursors, MAS-component modulators) constrained to be cell-autonomous in neurons rather than systemic. Then the Matthews group at Imperial on ABCA7 rs3752231 risk-variant effects on glial responses + amyloid plaque maturation — quantitative neuropathology on 4G8-immunostained mid-temporal gyrus from 99 donors (Braak 0-VI) + glial-enriched snRNA-seq on 54. Carriers exhibit increased Aβ burden + larger plaques in late AD explained almost entirely by selective expansion of diffuse plaques + relative reduction in compact plaques (consistent with impaired microglial plaque maturation). Transcriptional responses to increasing Aβ are largely genotype-specific: non-carriers show canonical disease-associated microglial activation (complement, phagocytic, inflammatory) + astrocyte responses preserving synaptic support; carriers show distinguishable activation state with carrier-specific ligand-receptor signaling suggesting non-cell-autonomous suppression of microglial phagocytosis. Therapeutic nominations: microglial TREM2 activation, CD33 inhibition, astrocytic EAAT2 induction. Methodological template: pair quantitative neuropathology with genotype-stratified glial snRNA-seq in 100-donor cohorts to turn ambiguous common variants into tractable phenotypes mapping onto immune-modulating drug programs. Then a chemical-biology piece — ADLumin-5 self-photosensitizing chemiluminescence compound (molecularly produced light) induces photo-oxidation + photodegradation of misfolded proteins (Aβ, tau, α-synuclein, TDP-43) validated by LC-MS, MALDI-MS, Western; Aβ photo-oxidation reduces oligomer toxicity. In vivo: 4-month 5xFAD treatment with longitudinal molecular imaging shows reduced Aβ accumulation. Chemistry insight: chemiluminescence-driven photosensitization generates light at the compound location, removing the external-light bottleneck that has kept conventional photodynamic therapy out of brain targets sitting centimeters below skull. Photo-theranostic dual-functional: same molecule images + treats. Open questions on selectivity (preference for misfolded over native forms at therapeutic concentrations) + specificity (ROS damage to nearby lipids/proteins) — novel enough that safety story needs work before leaving rodents. Then a Kravitz + Creed collaboration identifying ventral pallidal GABAergic neurons as preferential controllers of hedonic feeding — optogenetic activation drives robust consumption of high-fat diet + palatable liquids but not regular chow; neurons express few hunger-hormone receptors, not activated by ghrelin or fasting; single-cell Ca²⁺ imaging shows stronger engagement during long vs. short feeding bouts (palatability-linked bout-duration control), confirmed by closed-loop optogenetics. taCasp3-mediated ablation of VP-GABA reduces palatable-liquid intake + blocks high-fat-diet-induced obesity without impairing homeostatic chow intake. Clinical context: GLP-1 + dual-incretin agonists (semaglutide, tirzepatide) suppress homeostatic hunger but a fraction of patients respond modestly because their overeating is hedonically driven; a preferentially hedonic circuit with a clean ablation phenotype rationalizes next-generation anti-obesity therapeutics targeting the hedonic axis specifically + could explain a chunk of GLP-1 nonresponse. Translation gated by VP being hard to target with small molecules — surface markers on VP-GABA would open chemogenetic + biologic-targeting approaches. Then VCBench — single-cell foundation models (Geneformer, scGPT, UCE, TranscriptFormer, Arc State) benchmarked across 7 virtual-cell capability dimensions synthesized from 4 frameworks: perturbation response, cross-species universality, GRN inference, modality integration, temporal dynamics, multi-scale integration, in silico experimentation. 5 dimensions admit direct or proxy evaluation; multi-scale + in silico experimentation are structurally untestable. Pre-registered linear + nearest-neighbor baselines match or exceed every foundation model on 4 of 5 testable dimensions (perturbation, cross-species, GRN, temporal ordering). TranscriptFormer alone beats baselines on cross-modal RNA→protein (+53%% Pearson, with documented contamination caveat) + reaches Level 2 on pre-registered Virtual Cell rubric, but architectural choice behind this advantage causes spectral collapse destroying temporal-ordering performance — tradeoff invisible to single-task benchmarks. No foundation model publishes complete cell-level training manifest, leaving contamination undetectable. Methodological contributions: Contamination Reporting Schema, common-label-set protocol controlling cross-species class-count confounds, spread-error correlation probe for epistemic calibration — likely outlast the specific benchmark. Interpretation: scaled-up models not useless for single-cell, but the current crop has not earned the virtual-cell framing. Closes on a Bittremieux-group methods paper — SIMBA, a transformer-based model that infers two interpretable graph-based structural distances (maximum common edge subgraph + substructure edit distance) directly from tandem-MS spectrum pairs. Reframes the analog-discovery problem from spectral similarity (weakly correlated with structure) to direct prediction of chemistry-grounded structural distances. Consistently retrieves structurally closer analogs than existing methods + the distances are interpretable as chemistry rather than opaque embedding distances. Use case: drug-discovery-adjacent analog discovery for hit-expansion against natural-product + microbial congeners absent from spectral libraries. The kind of small focused methods paper that quietly changes day-to-day workflow — interpretable-distance framing is the right move for regulated hypothesis-driven discovery. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-24-receptor-and-reason Wed, 24 Jun 2026 12:00:00 +0000 1106 Daily roundup for June 24, 2026: MapLight ML-004 (selective 5-HT1B/1D agonist) misses Phase 2 IRIS primary on caregiver-reported ABI social communication in ASD (n=161), but pre-specified adolescent ABC-I ≥16 subgroup shows ABC-I effect size 1.33 (p=0.013) + CGI-I 1.08 (p=0.036), with no EPS + lower mean weight gain than placebo — repositioning toward ASD-irritability vs. aripiprazole/risperidone pending End-of-Phase-2 with FDA; bioRxiv Schnitzer-lab dual-color two-photon Ca²⁺ imaging in 6-OHDA dissociates amantadine from L-DOPA — L-DOPA rescues dSPN/iSPN balance but not locomotion-coding ensemble, amantadine rescues locomotion ensemble without restoring balance + selectively suppresses dyskinesia-ensemble resting activity — pathway-balance model wrong for amantadine, ensemble-specific pharmacology opens new target landscape; bioRxiv Surmeier-lab ChI-driven nAChR-mediated GABAergic-interneuron→SPN circuit collapses in prodromal + parkinsonian mice via GI nicotinic-AChR downregulation, not ChI failure — nicotinic agonist/PAM strategy targeting striatal GIs rescues motor symptoms independent of dopaminergic preservation; bioRxiv first scRNA-seq of HD-positive human fetal striatum (CAG-confirmed) identifies 2,032 DEGs across 9 clusters with overlap to adult HD signature — molecular HD pathology present in early human striatal development supports earlier-intervention timeline for HTT-lowering therapies; bioRxiv multi-organ spatial metabolomic atlas + brain proteomics in long-term exercising mice nominates neuronal Complex I as convergent rescue node — exercise reduces >70%% PS19 hippocampal tau, AAV-Ndi1 in neurons without exercise recapitulates anti-tau effect — drug-tractable Complex I axis as cell-autonomous exercise mimetic for tauopathies; bioRxiv Matthews-group ABCA7 rs3752231 quantitative neuropathology + snRNA-seq on 99/54 donors links carrier risk to selective diffuse-plaque expansion + impaired microglial plaque maturation + carrier-specific glial activation state — genotype-stratified targets: microglial TREM2 activation, CD33 inhibition, astrocytic EAAT2 induction; bioRxiv ADLumin-5 self-photosensitizing chemiluminescence agent photo-degrades Aβ + tau + α-synuclein + TDP-43 in vitro + reduces in vivo Aβ in 5xFAD over 4mo — molecularly produced light removes the external-light bottleneck on PDT for CNS misfolded-protein targets, photo-theranostic dual-functional, selectivity + safety open; bioRxiv Kravitz-and-Creed labs identify ventral pallidal GABA as preferential hedonic-feeding controller — optogenetic activation drives high-fat-diet + palatable-liquid consumption but not chow, taCasp3 ablation blocks DIO without impairing chow — hedonic-axis substrate rationalizing next-gen anti-obesity therapeutics for GLP-1 nonresponders; bioRxiv VCBench shows pre-registered linear + nearest-neighbor baselines match or beat 5 single-cell foundation models on 4/5 testable virtual-cell dimensions, TranscriptFormer wins only on cross-modal RNA→protein (+53%% with contamination caveat + spectral-collapse tradeoff invisible to single-task benchmarks), no model publishes complete training manifest — Contamination Reporting Schema + common-label-set protocol + spread-error correlation probe are the methodological contributions; closes on Bittremieux-lab SIMBA transformer infers max-common-edge-subgraph + substructure-edit distances directly from tandem-MS spectrum pairs for structurally-grounded analog discovery with interpretable graph-based distances — drug-discovery-adjacent workflow for natural-product + microbial congeners absent from spectral libraries. Receptor & Reason for June 24, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with MapLight's ML-004 Phase 2 IRIS topline in autism spectrum disorder — selective 5-HT1B/1D agonist, n=161 (102 adolescents, 59 adults), 12-week double-blind placebo-controlled, missed primary endpoint of change-from-baseline on caregiver-reported ABI–Social Communication Domain. Pre-specified analysis in adolescents with moderate-to-severe baseline irritability (ABC-I ≥16) showed care-partner-reported ABC-I effect size 1.33 (nominal p=0.013) + clinician-rated CGI-I effect size 1.08 (nominal p=0.036). Safety profile clean: no SAEs on active, no extrapyramidal events, lower mean weight gain than placebo. Heading into End-of-Phase-2 with FDA — failed-social-communication framing puts the program at risk; positive-irritability-subgroup framing positions ML-004 as a more selective alternative to aripiprazole + risperidone for ASD-associated irritability with cleaner side-effect profile. Honest read: trial was powered for a social-communication signal that didn't materialize, irritability subgroup is hypothesis-generating and needs a prospective ABC-I-primary confirmatory trial. Selective 5-HT-1B/1D agonism is emerging as a possible irritability-circuit modulator without the metabolic + motor liabilities of dopamine blockers. Then a Schnitzer-lab bioRxiv preprint that resolves the long-standing puzzle of amantadine's mechanism — dual-color two-photon Ca²⁺ imaging in 6-OHDA mice simultaneously monitors dSPN + iSPN dynamics across healthy, hypokinetic (parkinsonian), and hyperkinetic (dyskinetic) states, separating pathway-balance analysis from action-specific-ensemble analysis. L-DOPA rescues dSPN/iSPN imbalance but fails to restore the locomotion-coding ensemble; amantadine rescues the locomotion ensemble without normalizing pathway balance. Forelimb dyskinesias correspond to a distinct ensemble (not locomotion); amantadine selectively suppresses the dyskinesia ensemble's resting activity while leaving locomotion-coding cells alone. Classical pathway-balance model is the wrong lens for amantadine + probably for anti-dyskinetic pharmacology — what matters is which action-coding ensembles fire when. Drug-discovery question becomes which receptor targets selectively modulate dyskinesia vs. locomotion ensembles. Then a Surmeier-lab paper on the cholinergic-interneuron-to-GABAergic-interneuron-to-SPN arm of striatal circuitry — optogenetic ChI stimulation in healthy mice evokes GABA-A currents in both direct + indirect SPNs via nAChR-mediated activation of GABAergic interneurons, simulations suggest state-dependent control of SPN dendritic integration modulated by concurrent muscarinic signaling. In prodromal + overt parkinsonian models the circuit collapses — because the GABAergic interneurons downregulate nicotinic AChRs (not because ChIs fail). Nicotinic agonists or PAMs targeted to the relevant nAChR subunit on striatal GIs could restore lost gain control independent of dopaminergic preservation, explaining long-noted protective effects of nicotinic pharmacology in PD + nominating GI nAChRs as early-biomarker + early-intervention substrates because the disruption appears prodromally. Then the first scRNA-seq of an HD-positive human fetal striatum + age/sex-matched control — CAG-repeat expansion confirmed, 2,032 DEGs across 9 cellular clusters, gene-enrichment shows disrupted key biological processes with cluster-specific pathway dysregulation, overlap with publicly available adult HD postmortem signatures. Molecular HD pathology present during early human striatal development — supports neurodevelopmental component + earlier-intervention timeline for HTT-lowering + CAG-targeting therapies (tominersen, branaplam-style splicing modulators) including potentially prenatally in families with known expansions. n-of-one caveat, additional fetal samples needed for confirmation, but direction consistent with imaging literature. Then a tour-de-force exercise + tau paper — multi-organ spatial metabolomic atlas of long-term exercise in WT mice across 6 organs (brain, heart, lung, liver, kidney, skeletal muscle) catalogues 224 metabolic features with coordinated inter-organ remodeling; brain shows most pronounced region-specific adaptation. Extended to PS19 tauopathy: exercise reduces >70%% of observable tau pathology in PS19 hippocampus + restores mitochondrial-related metabolome. Integrated proteomic + spatial metabolomic analysis identifies NADH dehydrogenase Complex I as convergent rescue node. Biological test: AAV-Ndi1 (yeast NADH dehydrogenase that stands in for Complex I) expression in PS19 neurons without exercise increases cerebral antioxidants, restores malate-aspartate-shuttle metabolites, reduces tau pathology — recapitulates anti-tau effects of exercise. Provides molecular substrate for the human-cohort literature linking exercise to delayed cognitive decline, nominates Complex I augmentation as drug-tractable exercise-mimetic axis (NAD precursors, MAS-component modulators) constrained to be cell-autonomous in neurons rather than systemic. Then the Matthews group at Imperial on ABCA7 rs3752231 risk-variant effects on glial responses + amyloid plaque maturation — quantitative neuropathology on 4G8-immunostained mid-temporal gyrus from 99 donors (Braak 0-VI) + glial-enriched snRNA-seq on 54. Carriers exhibit increased Aβ burden + larger plaques in late AD explained almost entirely by selective expansion of diffuse plaques + relative reduction in compact plaques (consistent with impaired microglial plaque maturation). Transcriptional responses to increasing Aβ are largely genotype-specific: non-carriers show canonical disease-associated microglial activation (complement, phagocytic, inflammatory) + astrocyte responses preserving synaptic support; carriers show distinguishable activation state with carrier-specific ligand-receptor signaling suggesting non-cell-autonomous suppression of microglial phagocytosis. Therapeutic nominations: microglial TREM2 activation, CD33 inhibition, astrocytic EAAT2 induction. Methodological template: pair quantitative neuropathology with genotype-stratified glial snRNA-seq in 100-donor cohorts to turn ambiguous common variants into tractable phenotypes mapping onto immune-modulating drug programs. Then a chemical-biology piece — ADLumin-5 self-photosensitizing chemiluminescence compound (molecularly produced light) induces photo-oxidation + photodegradation of misfolded proteins (Aβ, tau, α-synuclein, TDP-43) validated by LC-MS, MALDI-MS, Western; Aβ photo-oxidation reduces oligomer toxicity. In vivo: 4-month 5xFAD treatment with longitudinal molecular imaging shows reduced Aβ accumulation. Chemistry insight: chemiluminescence-driven photosensitization generates light at the compound location, removing the external-light bottleneck that has kept conventional photodynamic therapy out of brain targets sitting centimeters below skull. Photo-theranostic dual-functional: same molecule images + treats. Open questions on selectivity (preference for misfolded over native forms at therapeutic concentrations) + specificity (ROS damage to nearby lipids/proteins) — novel enough that safety story needs work before leaving rodents. Then a Kravitz + Creed collaboration identifying ventral pallidal GABAergic neurons as preferential controllers of hedonic feeding — optogenetic activation drives robust consumption of high-fat diet + palatable liquids but not regular chow; neurons express few hunger-hormone receptors, not activated by ghrelin or fasting; single-cell Ca²⁺ imaging shows stronger engagement during long vs. short feeding bouts (palatability-linked bout-duration control), confirmed by closed-loop optogenetics. taCasp3-mediated ablation of VP-GABA reduces palatable-liquid intake + blocks high-fat-diet-induced obesity without impairing homeostatic chow intake. Clinical context: GLP-1 + dual-incretin agonists (semaglutide, tirzepatide) suppress homeostatic hunger but a fraction of patients respond modestly because their overeating is hedonically driven; a preferentially hedonic circuit with a clean ablation phenotype rationalizes next-generation anti-obesity therapeutics targeting the hedonic axis specifically + could explain a chunk of GLP-1 nonresponse. Translation gated by VP being hard to target with small molecules — surface markers on VP-GABA would open chemogenetic + biologic-targeting approaches. Then VCBench — single-cell foundation models (Geneformer, scGPT, UCE, TranscriptFormer, Arc State) benchmarked across 7 virtual-cell capability dimensions synthesized from 4 frameworks: perturbation response, cross-species universality, GRN inference, modality integration, temporal dynamics, multi-scale integration, in silico experimentation. 5 dimensions admit direct or proxy evaluation; multi-scale + in silico experimentation are structurally untestable. Pre-registered linear + nearest-neighbor baselines match or exceed every foundation model on 4 of 5 testable dimensions (perturbation, cross-species, GRN, temporal ordering). TranscriptFormer alone beats baselines on cross-modal RNA→protein (+53%% Pearson, with documented contamination caveat) + reaches Level 2 on pre-registered Virtual Cell rubric, but architectural choice behind this advantage causes spectral collapse destroying temporal-ordering performance — tradeoff invisible to single-task benchmarks. No foundation model publishes complete cell-level training manifest, leaving contamination undetectable. Methodological contributions: Contamination Reporting Schema, common-label-set protocol controlling cross-species class-count confounds, spread-error correlation probe for epistemic calibration — likely outlast the specific benchmark. Interpretation: scaled-up models not useless for single-cell, but the current crop has not earned the virtual-cell framing. Closes on a Bittremieux-group methods paper — SIMBA, a transformer-based model that infers two interpretable graph-based structural distances (maximum common edge subgraph + substructure edit distance) directly from tandem-MS spectrum pairs. Reframes the analog-discovery problem from spectral similarity (weakly correlated with structure) to direct prediction of chemistry-grounded structural distances. Consistently retrieves structurally closer analogs than existing methods + the distances are interpretable as chemistry rather than opaque embedding distances. Use case: drug-discovery-adjacent analog discovery for hit-expansion against natural-product + microbial congeners absent from spectral libraries. The kind of small focused methods paper that quietly changes day-to-day workflow — interpretable-distance framing is the right move for regulated hypothesis-driven discovery. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-24-receptor-and-reason Wed, 24 Jun 2026 12:00:00 +0000 1065 Daily roundup for June 24, 2026: MapLight ML-004 (selective 5-HT1B/1D agonist) misses Phase 2 IRIS primary on caregiver-reported ABI social communication in ASD (n=161), but pre-specified adolescent ABC-I ≥16 subgroup shows ABC-I effect size 1.33 (p=0.013) + CGI-I 1.08 (p=0.036), with no EPS + lower mean weight gain than placebo — repositioning toward ASD-irritability vs. aripiprazole/risperidone pending End-of-Phase-2 with FDA; bioRxiv Schnitzer-lab dual-color two-photon Ca²⁺ imaging in 6-OHDA dissociates amantadine from L-DOPA — L-DOPA rescues dSPN/iSPN balance but not locomotion-coding ensemble, amantadine rescues locomotion ensemble without restoring balance + selectively suppresses dyskinesia-ensemble resting activity — pathway-balance model wrong for amantadine, ensemble-specific pharmacology opens new target landscape; bioRxiv Surmeier-lab ChI-driven nAChR-mediated GABAergic-interneuron→SPN circuit collapses in prodromal + parkinsonian mice via GI nicotinic-AChR downregulation, not ChI failure — nicotinic agonist/PAM strategy targeting striatal GIs rescues motor symptoms independent of dopaminergic preservation; bioRxiv first scRNA-seq of HD-positive human fetal striatum (CAG-confirmed) identifies 2,032 DEGs across 9 clusters with overlap to adult HD signature — molecular HD pathology present in early human striatal development supports earlier-intervention timeline for HTT-lowering therapies; bioRxiv multi-organ spatial metabolomic atlas + brain proteomics in long-term exercising mice nominates neuronal Complex I as convergent rescue node — exercise reduces >70%% PS19 hippocampal tau, AAV-Ndi1 in neurons without exercise recapitulates anti-tau effect — drug-tractable Complex I axis as cell-autonomous exercise mimetic for tauopathies; bioRxiv Matthews-group ABCA7 rs3752231 quantitative neuropathology + snRNA-seq on 99/54 donors links carrier risk to selective diffuse-plaque expansion + impaired microglial plaque maturation + carrier-specific glial activation state — genotype-stratified targets: microglial TREM2 activation, CD33 inhibition, astrocytic EAAT2 induction; bioRxiv ADLumin-5 self-photosensitizing chemiluminescence agent photo-degrades Aβ + tau + α-synuclein + TDP-43 in vitro + reduces in vivo Aβ in 5xFAD over 4mo — molecularly produced light removes the external-light bottleneck on PDT for CNS misfolded-protein targets, photo-theranostic dual-functional, selectivity + safety open; bioRxiv Kravitz-and-Creed labs identify ventral pallidal GABA as preferential hedonic-feeding controller — optogenetic activation drives high-fat-diet + palatable-liquid consumption but not chow, taCasp3 ablation blocks DIO without impairing chow — hedonic-axis substrate rationalizing next-gen anti-obesity therapeutics for GLP-1 nonresponders; bioRxiv VCBench shows pre-registered linear + nearest-neighbor baselines match or beat 5 single-cell foundation models on 4/5 testable virtual-cell dimensions, TranscriptFormer wins only on cross-modal RNA→protein (+53%% with contamination caveat + spectral-collapse tradeoff invisible to single-task benchmarks), no model publishes complete training manifest — Contamination Reporting Schema + common-label-set protocol + spread-error correlation probe are the methodological contributions; closes on Bittremieux-lab SIMBA transformer infers max-common-edge-subgraph + substructure-edit distances directly from tandem-MS spectrum pairs for structurally-grounded analog discovery with interpretable graph-based distances — drug-discovery-adjacent workflow for natural-product + microbial congeners absent from spectral libraries. Receptor & Reason for June 24, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with MapLight's ML-004 Phase 2 IRIS topline in autism spectrum disorder — selective 5-HT1B/1D agonist, n=161 (102 adolescents, 59 adults), 12-week double-blind placebo-controlled, missed primary endpoint of change-from-baseline on caregiver-reported ABI–Social Communication Domain. Pre-specified analysis in adolescents with moderate-to-severe baseline irritability (ABC-I ≥16) showed care-partner-reported ABC-I effect size 1.33 (nominal p=0.013) + clinician-rated CGI-I effect size 1.08 (nominal p=0.036). Safety profile clean: no SAEs on active, no extrapyramidal events, lower mean weight gain than placebo. Heading into End-of-Phase-2 with FDA — failed-social-communication framing puts the program at risk; positive-irritability-subgroup framing positions ML-004 as a more selective alternative to aripiprazole + risperidone for ASD-associated irritability with cleaner side-effect profile. Honest read: trial was powered for a social-communication signal that didn't materialize, irritability subgroup is hypothesis-generating and needs a prospective ABC-I-primary confirmatory trial. Selective 5-HT-1B/1D agonism is emerging as a possible irritability-circuit modulator without the metabolic + motor liabilities of dopamine blockers. Then a Schnitzer-lab bioRxiv preprint that resolves the long-standing puzzle of amantadine's mechanism — dual-color two-photon Ca²⁺ imaging in 6-OHDA mice simultaneously monitors dSPN + iSPN dynamics across healthy, hypokinetic (parkinsonian), and hyperkinetic (dyskinetic) states, separating pathway-balance analysis from action-specific-ensemble analysis. L-DOPA rescues dSPN/iSPN imbalance but fails to restore the locomotion-coding ensemble; amantadine rescues the locomotion ensemble without normalizing pathway balance. Forelimb dyskinesias correspond to a distinct ensemble (not locomotion); amantadine selectively suppresses the dyskinesia ensemble's resting activity while leaving locomotion-coding cells alone. Classical pathway-balance model is the wrong lens for amantadine + probably for anti-dyskinetic pharmacology — what matters is which action-coding ensembles fire when. Drug-discovery question becomes which receptor targets selectively modulate dyskinesia vs. locomotion ensembles. Then a Surmeier-lab paper on the cholinergic-interneuron-to-GABAergic-interneuron-to-SPN arm of striatal circuitry — optogenetic ChI stimulation in healthy mice evokes GABA-A currents in both direct + indirect SPNs via nAChR-mediated activation of GABAergic interneurons, simulations suggest state-dependent control of SPN dendritic integration modulated by concurrent muscarinic signaling. In prodromal + overt parkinsonian models the circuit collapses — because the GABAergic interneurons downregulate nicotinic AChRs (not because ChIs fail). Nicotinic agonists or PAMs targeted to the relevant nAChR subunit on striatal GIs could restore lost gain control independent of dopaminergic preservation, explaining long-noted protective effects of nicotinic pharmacology in PD + nominating GI nAChRs as early-biomarker + early-intervention substrates because the disruption appears prodromally. Then the first scRNA-seq of an HD-positive human fetal striatum + age/sex-matched control — CAG-repeat expansion confirmed, 2,032 DEGs across 9 cellular clusters, gene-enrichment shows disrupted key biological processes with cluster-specific pathway dysregulation, overlap with publicly available adult HD postmortem signatures. Molecular HD pathology present during early human striatal development — supports neurodevelopmental component + earlier-intervention timeline for HTT-lowering + CAG-targeting therapies (tominersen, branaplam-style splicing modulators) including potentially prenatally in families with known expansions. n-of-one caveat, additional fetal samples needed for confirmation, but direction consistent with imaging literature. Then a tour-de-force exercise + tau paper — multi-organ spatial metabolomic atlas of long-term exercise in WT mice across 6 organs (brain, heart, lung, liver, kidney, skeletal muscle) catalogues 224 metabolic features with coordinated inter-organ remodeling; brain shows most pronounced region-specific adaptation. Extended to PS19 tauopathy: exercise reduces >70%% of observable tau pathology in PS19 hippocampus + restores mitochondrial-related metabolome. Integrated proteomic + spatial metabolomic analysis identifies NADH dehydrogenase Complex I as convergent rescue node. Biological test: AAV-Ndi1 (yeast NADH dehydrogenase that stands in for Complex I) expression in PS19 neurons without exercise increases cerebral antioxidants, restores malate-aspartate-shuttle metabolites, reduces tau pathology — recapitulates anti-tau effects of exercise. Provides molecular substrate for the human-cohort literature linking exercise to delayed cognitive decline, nominates Complex I augmentation as drug-tractable exercise-mimetic axis (NAD precursors, MAS-component modulators) constrained to be cell-autonomous in neurons rather than systemic. Then the Matthews group at Imperial on ABCA7 rs3752231 risk-variant effects on glial responses + amyloid plaque maturation — quantitative neuropathology on 4G8-immunostained mid-temporal gyrus from 99 donors (Braak 0-VI) + glial-enriched snRNA-seq on 54. Carriers exhibit increased Aβ burden + larger plaques in late AD explained almost entirely by selective expansion of diffuse plaques + relative reduction in compact plaques (consistent with impaired microglial plaque maturation). Transcriptional responses to increasing Aβ are largely genotype-specific: non-carriers show canonical disease-associated microglial activation (complement, phagocytic, inflammatory) + astrocyte responses preserving synaptic support; carriers show distinguishable activation state with carrier-specific ligand-receptor signaling suggesting non-cell-autonomous suppression of microglial phagocytosis. Therapeutic nominations: microglial TREM2 activation, CD33 inhibition, astrocytic EAAT2 induction. Methodological template: pair quantitative neuropathology with genotype-stratified glial snRNA-seq in 100-donor cohorts to turn ambiguous common variants into tractable phenotypes mapping onto immune-modulating drug programs. Then a chemical-biology piece — ADLumin-5 self-photosensitizing chemiluminescence compound (molecularly produced light) induces photo-oxidation + photodegradation of misfolded proteins (Aβ, tau, α-synuclein, TDP-43) validated by LC-MS, MALDI-MS, Western; Aβ photo-oxidation reduces oligomer toxicity. In vivo: 4-month 5xFAD treatment with longitudinal molecular imaging shows reduced Aβ accumulation. Chemistry insight: chemiluminescence-driven photosensitization generates light at the compound location, removing the external-light bottleneck that has kept conventional photodynamic therapy out of brain targets sitting centimeters below skull. Photo-theranostic dual-functional: same molecule images + treats. Open questions on selectivity (preference for misfolded over native forms at therapeutic concentrations) + specificity (ROS damage to nearby lipids/proteins) — novel enough that safety story needs work before leaving rodents. Then a Kravitz + Creed collaboration identifying ventral pallidal GABAergic neurons as preferential controllers of hedonic feeding — optogenetic activation drives robust consumption of high-fat diet + palatable liquids but not regular chow; neurons express few hunger-hormone receptors, not activated by ghrelin or fasting; single-cell Ca²⁺ imaging shows stronger engagement during long vs. short feeding bouts (palatability-linked bout-duration control), confirmed by closed-loop optogenetics. taCasp3-mediated ablation of VP-GABA reduces palatable-liquid intake + blocks high-fat-diet-induced obesity without impairing homeostatic chow intake. Clinical context: GLP-1 + dual-incretin agonists (semaglutide, tirzepatide) suppress homeostatic hunger but a fraction of patients respond modestly because their overeating is hedonically driven; a preferentially hedonic circuit with a clean ablation phenotype rationalizes next-generation anti-obesity therapeutics targeting the hedonic axis specifically + could explain a chunk of GLP-1 nonresponse. Translation gated by VP being hard to target with small molecules — surface markers on VP-GABA would open chemogenetic + biologic-targeting approaches. Then VCBench — single-cell foundation models (Geneformer, scGPT, UCE, TranscriptFormer, Arc State) benchmarked across 7 virtual-cell capability dimensions synthesized from 4 frameworks: perturbation response, cross-species universality, GRN inference, modality integration, temporal dynamics, multi-scale integration, in silico experimentation. 5 dimensions admit direct or proxy evaluation; multi-scale + in silico experimentation are structurally untestable. Pre-registered linear + nearest-neighbor baselines match or exceed every foundation model on 4 of 5 testable dimensions (perturbation, cross-species, GRN, temporal ordering). TranscriptFormer alone beats baselines on cross-modal RNA→protein (+53%% Pearson, with documented contamination caveat) + reaches Level 2 on pre-registered Virtual Cell rubric, but architectural choice behind this advantage causes spectral collapse destroying temporal-ordering performance — tradeoff invisible to single-task benchmarks. No foundation model publishes complete cell-level training manifest, leaving contamination undetectable. Methodological contributions: Contamination Reporting Schema, common-label-set protocol controlling cross-species class-count confounds, spread-error correlation probe for epistemic calibration — likely outlast the specific benchmark. Interpretation: scaled-up models not useless for single-cell, but the current crop has not earned the virtual-cell framing. Closes on a Bittremieux-group methods paper — SIMBA, a transformer-based model that infers two interpretable graph-based structural distances (maximum common edge subgraph + substructure edit distance) directly from tandem-MS spectrum pairs. Reframes the analog-discovery problem from spectral similarity (weakly correlated with structure) to direct prediction of chemistry-grounded structural distances. Consistently retrieves structurally closer analogs than existing methods + the distances are interpretable as chemistry rather than opaque embedding distances. Use case: drug-discovery-adjacent analog discovery for hit-expansion against natural-product + microbial congeners absent from spectral libraries. The kind of small focused methods paper that quietly changes day-to-day workflow — interpretable-distance framing is the right move for regulated hypothesis-driven discovery. Voiced by Voxtral on Garibaldi HPC (mistralai/Voxtral-4B-TTS-2603, neutral_male, 1.2× speed), Mistral voxtral-mini-tts-2603 with en_paul_neutral as API fallback. 2026-06-24-receptor-and-reason Wed, 24 Jun 2026 12:00:00 +0000 1065 Daily roundup for June 24, 2026: MapLight ML-004 (selective 5-HT1B/1D agonist) misses Phase 2 IRIS primary on caregiver-reported ABI social communication in ASD (n=161), but pre-specified adolescent ABC-I ≥16 subgroup shows ABC-I effect size 1.33 (p=0.013) + CGI-I 1.08 (p=0.036), with no EPS + lower mean weight gain than placebo — repositioning toward ASD-irritability vs. aripiprazole/risperidone pending End-of-Phase-2 with FDA; bioRxiv Schnitzer-lab dual-color two-photon Ca²⁺ imaging in 6-OHDA dissociates amantadine from L-DOPA — L-DOPA rescues dSPN/iSPN balance but not locomotion-coding ensemble, amantadine rescues locomotion ensemble without restoring balance + selectively suppresses dyskinesia-ensemble resting activity — pathway-balance model wrong for amantadine, ensemble-specific pharmacology opens new target landscape; bioRxiv Surmeier-lab ChI-driven nAChR-mediated GABAergic-interneuron→SPN circuit collapses in prodromal + parkinsonian mice via GI nicotinic-AChR downregulation, not ChI failure — nicotinic agonist/PAM strategy targeting striatal GIs rescues motor symptoms independent of dopaminergic preservation; bioRxiv first scRNA-seq of HD-positive human fetal striatum (CAG-confirmed) identifies 2,032 DEGs across 9 clusters with overlap to adult HD signature — molecular HD pathology present in early human striatal development supports earlier-intervention timeline for HTT-lowering therapies; bioRxiv multi-organ spatial metabolomic atlas + brain proteomics in long-term exercising mice nominates neuronal Complex I as convergent rescue node — exercise reduces >70%% PS19 hippocampal tau, AAV-Ndi1 in neurons without exercise recapitulates anti-tau effect — drug-tractable Complex I axis as cell-autonomous exercise mimetic for tauopathies; bioRxiv Matthews-group ABCA7 rs3752231 quantitative neuropathology + snRNA-seq on 99/54 donors links carrier risk to selective diffuse-plaque expansion + impaired microglial plaque maturation + carrier-specific glial activation state — genotype-stratified targets: microglial TREM2 activation, CD33 inhibition, astrocytic EAAT2 induction; bioRxiv ADLumin-5 self-photosensitizing chemiluminescence agent photo-degrades Aβ + tau + α-synuclein + TDP-43 in vitro + reduces in vivo Aβ in 5xFAD over 4mo — molecularly produced light removes the external-light bottleneck on PDT for CNS misfolded-protein targets, photo-theranostic dual-functional, selectivity + safety open; bioRxiv Kravitz-and-Creed labs identify ventral pallidal GABA as preferential hedonic-feeding controller — optogenetic activation drives high-fat-diet + palatable-liquid consumption but not chow, taCasp3 ablation blocks DIO without impairing chow — hedonic-axis substrate rationalizing next-gen anti-obesity therapeutics for GLP-1 nonresponders; bioRxiv VCBench shows pre-registered linear + nearest-neighbor baselines match or beat 5 single-cell foundation models on 4/5 testable virtual-cell dimensions, TranscriptFormer wins only on cross-modal RNA→protein (+53%% with contamination caveat + spectral-collapse tradeoff invisible to single-task benchmarks), no model publishes complete training manifest — Contamination Reporting Schema + common-label-set protocol + spread-error correlation probe are the methodological contributions; closes on Bittremieux-lab SIMBA transformer infers max-common-edge-subgraph + substructure-edit distances directly from tandem-MS spectrum pairs for structurally-grounded analog discovery with interpretable graph-based distances — drug-discovery-adjacent workflow for natural-product + microbial congeners absent from spectral libraries. Receptor & Reason for June 23, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with Definium Therapeutics' Phase 3 Emerge topline on DT120, the lysergide orally-disintegrating-tablet formulation, in major depressive disorder — 149 adults with DSM-5 MDD and baseline MADRS ≥26 across 20 sites, randomized to a single 100 µg dose or matching placebo with 12-week double-blind primary phase + 40-week open-label extension. Hit primary at week 6 cleanly: LS-mean MADRS change of -13.3 on DT120 vs -5.2 on placebo (placebo-adjusted -8.1, p<0.0001), with rapid-onset signature (week 1 placebo-adjusted -14.2 MADRS, bigger than any non-ketamine antidepressant at that timepoint) and durable to week 12 (placebo-adjusted -7.3) from a single dose — no serious adverse events, no suicidality signal. Largest, most rigorous single-dose-psychedelic readout in MDD to date, ODT formulation operational improvement over capsule formulations for dose timing + onset, reshuffles the FDA priority-voucher landscape around psychedelics (Compass for TRD, Usona for MDD, Transcend for methylone-PTSD) regardless of voucher status. Then the natural mechanistic complement from Neuropsychopharmacology — a Basel group's acute-and-post-acute neurobehavioral RCT of LSD in 45 healthy volunteers (24 women), randomized double-blind crossover of the same 100 µg dose vs placebo, with novel post-acute battery (auditory tetanization with EEG, paired-associative stimulation with TMS, serial peripheral BDNF, sequence-typing motor learning at d1, perceived stress + cognitive flexibility at 1 week). Two key takeaways: same 100 µg dose as the Phase 3 (mechanism + clinic at matched exposure, which they often aren't in psychedelic translational science), and the post-acute markers chosen — LTP-like plasticity + motor-cortex plasticity + BDNF — are exactly the substrates hypothesized to do the sustained antidepressant work. Detectable correlates in healthy brains at a single dose argue against pure state-dependent psychological reconsolidation and toward measurable molecular/circuit substrate. Strongest mechanism-to-Phase-3 connection psychedelic neuroscience has assembled (post-acute effect sizes off the abstract pending institutional access). Then a same-issue Neuropsychopharmacology paper on selective orexin receptor cross-over treatment for stress resilience — anterior basolateral amygdala contains pro-stress neuronal population identified by Hcrtr1 expression, sequenced or combined Orx1R + Orx2R interventions produce a different molecular signature than either alone, increases resilience + induces neuroplastic-signaling-gene expression. Cross-over framing implies temporal pharmacology (stage interventions rather than pick one receptor) — different from suvorexant/daridorexant for sleep or seltorexant's Orx2R antagonism for depression. If the gene-expression signature is causal for resilience, it's the staging insight that feeds into next-generation orexin-targeted antidepressants — narcolepsy field already shows Orx2R agonism on a fast track (oveporexton breakthrough designation), pulling resilience + neuroplasticity into the same pharmacological frame extends orexin program well beyond sleep medicine. Then the FDA CRL on Achieve Life Sciences' cytisinicline NDA — PDUFA was June 20, CRL came two days late. Cytisinicline = plant alkaloid α4β2 nAChR partial agonist (same target as varenicline but much shorter half-life + cleaner CV/neuropsychiatric profile). Issue cited is primary manufacturing not efficacy/safety, company resubmits end-2026 with potential first-half-2027 decision. Varenicline supply has been intermittent + NRT options are modest, a clean alpha-4-beta-2 partial agonist with friendlier side-effect profile would meaningfully change addiction-treatment toolkit; CRL slows conventional-smoking indication by 6-9 months but doesn't affect the procedurally-separate Commissioner's National Priority Voucher for e-cigarette/vaping cessation (much bigger market in younger smokers). Then a Bunnett-group bioRxiv preprint on targeting synaptic vesicle endocytosis in nociceptors — persistent pain signaling requires sustained nociceptor→dorsal-horn transmission requires SV recycling requires endocytic machinery (AAK1 + Dnm1 as candidate nociceptor-specific targets). Nav1.8+ DRG nociceptor-specific AAV knockdown of either blocks postoperative + neuropathic hypersensitivity without affecting baseline mechanical/thermal sensitivity, locomotion, or spontaneous behavior. Therapeutic-vehicle contribution: lipid nanoparticles encapsulating CRISPR-dCas9-repressor mRNA for sustained-reversible transcriptional/epigenetic repression of Aak1 or Dnm1 via intrathecal delivery — durable + reversible analgesia across postoperative, inflammatory, neuropathic, and osteoarthritis pain models without impairing acute nociception or locomotion, via reduced neurotransmitter release probability at nociceptor terminals from disrupted SV recycling. Presynaptic + nociceptor-specific + reversibly-tunable epigenetic intervention is structurally different from anything in current chronic-pain pipeline (off the abstract pending full body — questions on single-dose duration, LNP immunogenicity, route translation to chronic-pain patient acceptability). Then a combination-strategy preprint for an SLC6A1 encephalopathy variant — SLC6A1 encodes GAT-1 (principal neuronal GABA transporter), loss-of-function causes DEE through reduced GABA reuptake + impaired trafficking + haploinsufficiency. Newly identified de-novo p.Ala305Val variant in myoclonic-atonic epilepsy patient compared to residue-matched control across 9 structural-stability algorithms (cryo-EM GAT-1 template) + tritiated-GABA uptake (HEK293T + iPSC-derived astrocytes/cortical neurons) + ER colocalization + pharmacologic rescue with 4-PBA, TUDCA, salubrinal. Interesting therapeutic move: combine 4-PBA chemical chaperone with GAT-1 cDNA gene augmentation — each alone partial rescue, combination super-additive on uptake. Lesson for monogenic encephalopathy: two distinct failure modes (folding-trafficking defects + dose insufficiency) call for two distinct interventions, stacking more efficient than maximizing either, chaperone-responsiveness of the variant is the patient-selection criterion. Then an Iino-group bioRxiv on schizophrenia risk gene-enriched prefrontal ensemble — clozapine-responsive ensemble stabilizes value-guided choice, NMDA hypofunction increases value-independent decision noise + weakens pre-choice transient (both clozapine-rescuable), chemogenetic inhibition of the ensemble alone causally increases decision noise. Translational move: integrated receptor-affinity data with mouse + human single-cell transcriptomes from the ensemble cells, rationally designed multi-receptor antagonist cocktail that reactivates ensemble + rescues decision noise without increasing NREM sleep time or NREM delta power (clozapine does increase both). Ensemble-targeted pharmacology — pick the receptor combination that hits the cell type not the symptom — path toward decoupling antipsychotic efficacy from sedation. Structural advance is the cell-type-transcriptome-plus-receptor-pharmacology cocktail design even if this specific combination doesn't translate. Then a bioRxiv preprint introducing VFB-MCP from the Virtual Fly Brain consortium + Cambridge — natural-language access to Drosophila neuroscience grounded by an ontology-led knowledgebase via Model Context Protocol. 30 neuroscience tasks benchmarked against bare LLM + web-search-augmented LLM: MCP-equipped LLM produces precise + verifiable + appropriately quantified answers on 25/30 tasks vs 14/30 web vs 2/30 bare (Wilcoxon p<0.01 Holm-corrected, all pairwise), MCP advantage largest for tasks requiring data quantification (89%% vs 11%% for web) where ontology grounding does real work. Beyond flies: MCP over a domain ontology is now a deployable pattern for any field with a well-curated knowledgebase — generalizes to Allen Brain Atlas, rare-disease registries, curated drug-target databases. Adjacent: arXiv preprint from McGill + Brno on foundation models for epileptogenic-zone identification in drug-resistant epilepsy — about 30%% of epilepsy patients fail medication + become surgical candidates, surgical workup gated by accurate EZ localization. Standard pipeline = feature engineering on intracranial recordings or graph-based seizure-network modeling. Contribution: transformer-based foundation model pretrained self-supervised on large unlabeled interictal sEEG + MEG with cross-modality transfer + explicit sleep-stage context (interictal EZ signatures sleep-stage-dependent, often ignored by feature-engineered methods), benchmarked against classical DL + graph baselines on two open datasets. Clinical thesis: better unsupervised pretraining on multi-modal interictal data is the right path for EZ localization (severe labeled-data bottleneck + high patient-specific variability). Limitations: invasive sEEG/MEG only, scalp-EEG generalization untested, prospective surgical validation absent — but structural point is that DRE EZ localization is exactly the low-data + high-stakes + multi-modal problem where foundation-model pretraining has the biggest payoff. Closes on a Drexel arXiv preprint on persona-conditioned LLM support for people who use drugs — title is "Self-Stigma Is Not a Monolith, But Generic Empathy Is." Current LLM-based mental-health interventions for substance-use populations default to a uniform empathetic tone that misses the structure of self-stigma (internalized shame, externalized resistance, help-seeking willingness, stereotype endorsement, social withdrawal as distinct phenotypes). Conditioned responses across phenotypes compared against generic empathetic baselines using Gemini 3 Flash + Llama 4 Scout: conditioning on high-shame phenotypes improved perceived empathy + relevance, but generic empathy was actually better for individuals with lower internalized stigma — persona-conditioning is not a win on average; it's a win when correctly matched and a loss when mismatched. Clinical-deployment concern: without validated screening to assign phenotype before the conversation starts, persona-conditioning can degrade outcomes for half the target population. Methodological lesson generalizes beyond substance use — average-quality metrics over heterogeneous populations hide phenotype-by-treatment interactions that matter clinically. For deploying mental-health LLMs in real settings, safer default is generic empathy plus a separate validated triage step, not persona conditioning baked into a single endpoint. A useful negative result for the field. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-23-receptor-and-reason Tue, 23 Jun 2026 12:00:00 +0000 958 Daily roundup for June 23, 2026: Definium DT120 (lysergide ODT 100µg) hits Phase 3 Emerge primary in MDD with -13.3 vs -5.2 LSmean MADRS at wk6 (placebo-adjusted -8.1, p<0.0001) + rapid wk1 -14.2 + durable wk12 -7.3 + no SAEs/suicidality — largest single-dose-psychedelic readout in MDD to date with ODT formulation tightening dose timing/onset; Neuropsychopharmacology acute-and-post-acute LSD RCT in 45 healthy volunteers same 100µg dose with TMS-PAS + auditory-tetanization EEG + peripheral BDNF + motor-learning through 1wk post-dose — mechanistic substrate of LTP-like + motor-cortex plasticity for durable antidepressant effect at matched exposure to the Phase 3, strongest mechanism-to-Phase-3 connection psychedelic neuroscience has assembled; Neuropsychopharmacology selective orexin receptor cross-over identifies Hcrtr1+ neurons in anterior BLA as pro-stress circuit + cross-over Orx1R + Orx2R intervention sequence induces resilience + neuroplastic-signaling genes — temporal pharmacological staging extends orexin program beyond sleep into next-gen antidepressants; FDA CRL on Achieve cytisinicline NDA (α4β2 nAChR partial agonist with cleaner profile than varenicline) cites primary manufacturing not efficacy/safety, resubmission end-2026 — slows conventional-smoking indication 6-9mo but vaping CNPV procedurally separate; bioRxiv Bunnett-group targeting synaptic vesicle endocytosis in Nav1.8+ DRG nociceptors via LNP-CRISPR/dCas9-R intrathecal repression of AAK1/Dnm1 produces sustained-reversible analgesia across postoperative + neuropathic + inflammatory + osteoarthritis pain without baseline sensitivity/locomotor effects — presynaptic + nociceptor-specific + reversibly-tunable non-opioid mechanism structurally different from current chronic-pain pipeline (abstract-only); bioRxiv 4-PBA chemical chaperone + GAT-1 gene augmentation as super-additive dual therapy for novel SLC6A1 p.A305V missense DEE — two distinct failure modes (folding-trafficking + dose insufficiency) call for two distinct interventions stacked; bioRxiv schizophrenia risk gene-enriched prefrontal ensemble (clozapine-responsive) stabilizes value-guided choice — rational multi-receptor cocktail designed from cell-type transcriptomes reactivates ensemble + rescues decision noise without NREM/delta-power increases, ensemble-targeted pharmacology decouples antipsychotic efficacy from sedation; bioRxiv VFB-MCP exposes Virtual Fly Brain ontology-backed connectomics knowledgebase via Model Context Protocol — 25/30 vs 14/30 (web) vs 2/30 (bare) on Drosophila neuroscience tasks, MCP advantage largest for quantification tasks (89%% vs 11%%) — deployable architecture for any curated-ontology field; arXiv McGill-Brno foundation model for EZ identification in DRE via self-supervised transformer pretraining on interictal sEEG + MEG with sleep-stage context + cross-modality transfer — addresses labeled-data bottleneck of surgical workup; closes on arXiv Drexel persona-conditioned LLM support for people who use drugs — "Self-Stigma Is Not a Monolith, But Generic Empathy Is" — phenotype conditioning helps high-shame but generic empathy better for low-stigma, mismatch degrades quality — persona-conditioning is not a win on average + clinical deployment needs validated triage step not single-endpoint persona. Receptor & Reason for June 22, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a bioRxiv preprint on ADLumin-5, a self-photosensitizing chemiluminescence compound that emits intramolecular "molecular light" and uses that emission to photo-oxidize + photodegrade beta-amyloid, tau, alpha-synuclein, and TDP-43 in vitro (validated by LC-MS, MALDI-MS, and Western), reduces beta-amyloid toxicity, and over 4 months of dosing lowers plaque load in 5xFAD mice by in-vivo molecular imaging — because ADLumin-5 is both photosensitizer and imaging agent it functions as a dual photo-theranostic, novel modality that sidesteps the deep-brain-illumination problem that has gated optical protein-degradation in vivo (off the abstract; bioRxiv body not yet rendered, mechanistic depth and selectivity awaiting full read). Then a bioRxiv preprint from a Chieti group on amyloid PET tracer off-target binding to brain sulfotransferases — multi-omics on GTEx + Human Protein Atlas identifies SULT1A1, SULT1A3, and SULT4A1 as brain-expressed isoforms, then molecular docking + MD on Pittsburgh Compound-B + flutemetamol + flutafuranol vs each enzyme show energetically comparable but tracer-and-isoform-specific complex stability: PiB + flutemetamol stable in SULT1A1 but PiB drifts in SULT4A1, flutafuranol weak in SULT1A1 but stable in SULT1A3 + SULT4A1, and crucially all three SULTs are expressed in the cerebellum which is the reference region for amyloid PET normalization — so off-target binding sits in the denominator of the signal and potentially accounts for some of the persistent PET-positive-postmortem-negative variance. Then a bioRxiv from the MODEL-AD Dage group treating male + female 5xFAD mice weekly with murine chimeric aducanumab (chAdu) versus matched murine IgG2aκ isotype + saline for 17 weeks starting at 8 months — sex-dependent PK and PD (increased plasma Aβ42:40 + reduced brain Aβ in a sex-dependent way) but only mild behavioral impairments and no actual cognitive improvement, the IgG isotype control reproduces similar brain Aβ reductions (biologically active Fc-receptor-mediated effects independent of amyloid specificity, consistent with the microglial-activation literature), 10%% of treated mice develop treatment-emergent anti-drug antibodies with reduced drug exposure + efficacy, multi-omics reveals disease-associated transcripts + proteins that don't move with chAdu treatment even when amyloid does — the persistent post-amyloid molecular signature anchoring the case that next-generation AD therapeutics need combination strategies addressing pathways beyond Aβ. Then a bioRxiv preprint from Rubicon Biotechnology with Walter Reed + Stanford on Fv-HSP72 — engineered fusion of humanized 3E10 single-chain Fv antibody fragment (which binds extracellular DNA at injury sites + shuttles cargo via ENT2 into stressed cells) with human HSP72 via cleavable linkers (swivel or cathepsin-B-cleavable), screened in rat advanced-blast-simulator TBI model at 10 or 30 mg/kg IV 15 min post-blast. Across three biomarker classes (3-nitrotyrosine for oxidative stress, phosphorylated tau T181/T231/S396 for neurodegeneration, GFAP + S100A10 + C3d for glial state) the cathepsin-B-cleavable variant RBB012-CTB is selected as clinical candidate. A single 30 mg/kg dose at 15 min post-blast preserves novel-object discrimination at d2 + d8 (sham-equivalence) with sustained pTau + GFAP suppression through d9, astrogliosis biased toward neuroprotective S100A10+ A2 state over neurotoxic C3d+ A1 state, plasma clearance is fast (sub-LLOQ by 4h) but brain retention is preserved out to 12h and significantly higher in TBI rats than naïve rats — clean ENT2-mediated targeted-cargo platform for acute neurodegeneration where transient BBB compromise is the entry route. Then a bioRxiv preprint sharpening the mechanism of INPP5D in AD risk — SHIP1 (the INPP5D-encoded protein) localizes to plasma membrane + endolysosomal compartments + binds the CapZ family (required for endosome maturation), CRISPR reduction impairs endosome maturation + lysosomal function, lipid droplets accumulate, cathepsin B leaks from lysosomes into cytosol + activates the NLRP3 inflammasome (so inflammation is downstream of lysosomal failure, not a primary signaling defect). CITE-seq shows SHIP1-deficient microglia shift from immune-responsive to DAM-like phagocytic state with increased TREM2-mediated uptake of synaptic material + apoptotic neurons; Aβ uptake itself is unchanged but Aβ accumulates intracellularly due to defective lysosomal degradation — reframes INPP5D as a lysosomal-function gene, joining APOE and TREM2 in the convergent argument that AD risk genes converge on microglial lipid + lysosomal handling rather than on inflammation per se. Then a bioRxiv preprint with an honest negative-result middle on microglial MyD88 in alcohol use disorder — whole-body MyD88 loss had been shown to increase voluntary drinking, but cell-specific microglial MyD88 KO has minimal effect on voluntary alcohol intake or anxiety-like behavior in a one-bottle binge paradigm, and alcohol doesn't modify MyD88-dependent changes in parvalbumin interneurons + perineuronal nets despite altering microglial morphology in male PFC independent of genotype. Twist: early-life endotoxin challenge induces increased adult drinking in both MyD88-KO + WT males, as does saline injection in MyD88-KO males — microglial MyD88 is a context-dependent modifier of stress sensitivity, not a baseline drug target, methodological lesson on cell-specific-knockout-in-single-context behavioral paradigms. Then an arXiv preprint introducing TxBench-PP (TherapeuticsBench Preclinical Pharmacology), the first verifiable benchmark structured around real preclinical decisions — 100 evaluations across MoA + PD + compound-target engagement + causal target validation + developability + safety + translational efficacy where agents receive realistic workflow snapshots + inspect files in a coding environment + return deterministically-graded structured answers. 16 model-harness configurations across 11 models + 4,800 trajectories show no system reliably recovers preclinical pharmacology decisions: Claude Opus 4.8 with the Pi harness leads at 59.3%% endpoint pass rate, GPT-5.5/Pi at 55.3%% — frontier models in thoughtful harnesses do most of the easy reasoning but ~60%% on binary-ish program decisions is below where you'd want before letting agents make program calls, calibration target for drug-discovery agent builders. Then a bioRxiv on SteerAF — distogram-based inference-time optimization on AlphaFold2 via sparse MSA-feature perturbations generated by block gradient ascent recovers alternative protein conformations matching or exceeding existing methods on 4 benchmarks, with the perturbations correlating to experimentally characterized functional residues at ~50%% precision (interpretable steering knob); also supports decoy selection without experimental structures + MD seeding — natural use case is allosteric + cryptic-site drug discovery where the target's druggable state isn't the AlphaFold-default mode. Then a bioRxiv preprint on SMARTPocket — atomic-level geometric deep learning framework for predicting RNA-small-molecule binding pockets from 3D structure via transfer learning from 110,000+ protein binding interface structures, outperforming existing RNA pocket predictors + generalist biomolecular models on 4 single-chain + 3 broader benchmarks, generalizing to apo RNA when conformational change is modest, finding cryptic ligandable pockets + recapitulating experimentally validated sites in the SARS-CoV-2 frameshifting element + an evolved RNA aptamer, with SMARTPocket-guided docking improving near-native pose recovery + computational efficiency — operational challenge to the assumption that protein-to-RNA pocket-prediction transfer is infeasible, infrastructure for RNA-targeted drug discovery (off the abstract; body not yet rendered). Closes on a bioRxiv preprint from a Seattle group identifying cortisol as the primary pregnancy-related hormone driving hepatic uptake transporter induction in HepaRG cells — induces NTCP + OAT2 + OCT1, with CRISPR knockdown showing the glucocorticoid receptor as primary mediator + HNF4α and HNF1α as selective downstream contributors (HNF4α-KO attenuates OAT2/OCT1, HNF1α-KO enhances NTCP induction + attenuates OCT1), while PXR + CAR are induced but are not the drivers. Direct clinical relevance for antenatal corticosteroid co-administration with psychotropics, anticonvulsants, and analgesics — PBPK models that ignore GR-driven hepatic transporter induction will miss substantial disposition shifts in pregnancy. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-22-receptor-and-reason Mon, 22 Jun 2026 12:00:00 +0000 935 Daily roundup for June 22, 2026: bioRxiv ADLumin-5 self-photosensitizing chemiluminescence compound emits intramolecular light + photo-oxidizes/photodegrades beta-amyloid + tau + alpha-synuclein + TDP-43 in vitro + lowers plaque load in 5xFAD over 4mo by in-vivo imaging — dual-functional photo-theranostic agent for neurodegenerative disease (abstract-only); bioRxiv amyloid PET tracers PiB + flutemetamol + flutafuranol form tracer-and-isoform-specific complexes with brain SULT1A1 + SULT1A3 + SULT4A1 via docking + MD, all three SULTs expressed in the cerebellum reference region used for PET normalization — off-target binding sits in the denominator of amyloid PET signal, structural framework for next-gen tracer evaluation; bioRxiv MODEL-AD Dage lab chronic chimeric aducanumab (chAdu) in 5xFAD shows sex-dependent PK/PD + reduced brain Aβ + no cognitive improvement + 10%% anti-drug antibodies + IgG isotype control alone reproduces Aβ reductions (Fc-mediated nonspecific effects) + multi-omics reveals disease-associated transcripts/proteins unmoved by amyloid lowering — anchors the case for combination approaches in post-aducanumab AD; bioRxiv Rubicon Fv-HSP72 (humanized 3E10 scFv + HSP72 fusion with cathepsin-B-cleavable linker, RBB012-CTB) single 30mg/kg IV at 15min post-blast preserves novel-object memory at d2 + d8 + suppresses pTau + GFAP + biases astrogliosis to S100A10+ A2 over C3d+ A1 through d9, fast plasma clearance + sustained brain retention only in TBI rats — ENT2-mediated targeted-cargo platform for acute neurodegeneration with transient BBB compromise; bioRxiv INPP5D/SHIP1 in human iPSC microglia binds CapZ + regulates endo-lysosomal maturation, CRISPR reduction triggers cathepsin B leak + NLRP3 inflammasome + DAM-like phagocytic shift + intracellular Aβ accumulation from defective lysosomal degradation — reframes top AD risk gene as a lysosomal-function gene joining APOE + TREM2 in microglial lipid/lysosomal convergence; bioRxiv microglia-specific MyD88 KO does not change baseline voluntary alcohol intake (overturning microglial MyD88 hypothesis), early-life endotoxin shifts adult drinking — context-dependent stress-history modifier of AUD risk not baseline drug target, honest negative result; arXiv TxBench-PP verifiable benchmark of 100 preclinical pharmacology evaluations where agents inspect real workflow files + return deterministically-graded answers, 16 model-harness configs across 11 models + 4,800 trajectories show no system reliably recovers preclinical decisions (Claude Opus 4.8/Pi 59.3%%, GPT-5.5/Pi 55.3%%) — calibration target for drug-discovery agent builders; bioRxiv SteerAF distogram-based steering of AlphaFold2 via sparse MSA perturbations recovers alternative conformations + functional residues at ~50%% precision + supports allosteric/cryptic-site drug discovery; bioRxiv SMARTPocket geometric deep learning predicts ligandable + cryptic RNA-binding pockets via transfer from 110K+ protein interfaces, outperforms existing RNA predictors + recapitulates SARS-CoV-2 frameshifting + RNA aptamer pockets (abstract-only); closes on bioRxiv cortisol drives pregnancy-associated hepatic OAT2 + NTCP + OCT1 induction via GR with HNF4α + HNF1α as selective downstream gates, PXR + CAR not the drivers — direct clinical relevance for antenatal corticosteroid co-administration with psychotropics + anticonvulsants + analgesics where PBPK models ignoring GR-driven transporter induction will miss disposition shifts. Receptor & Reason for June 21, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a bioRxiv preprint that puts APOE4/4 microglial biology on a unified continuous manifold in Alzheimer's disease — spatially resolved proteomics with single-nuclear multiomics across APOE3/3 and APOE4/4 brains reveals that APOE4/4 cells lose homeostatic identity but also fail to fully reach the canonical disease-associated microglia program, getting stuck in an intermediate state with inflammatory signaling on, metabolic and phagocytic engagement off, localized to spatial niches of astrogliosis and senescence. The therapeutic implication is that DAM is neither protective nor harmful per se in APOE4 carriers — these cells can't reach DAM in the first place, so the relevant target is the activation threshold transition, not the endpoint. Then a paired bioRxiv from the Trushina group on CP2, a brain-penetrant mitochondrial complex I modulator dosed into APP/PS1 females starting at 19 months (advanced disease, well past the early-intervention window), with response analyzed through iMiceBrain — the first brain-specific genome-scale metabolic model — integrating transcriptomics + lipidomics + targeted metabolomics. CP2 restores mitochondrial substrate flexibility, reactivates fatty acid oxidation, normalizes pyruvate dehydrogenase flux, corrects cholesteryl ester + sphingolipid abnormalities, and tracks with synaptic + cognitive recovery — partial reversible complex I modulation as disease-modifying strategy, mechanistic counterweight to the amyloid/tau frame. Then a bioRxiv on prenatal Δ9-THC exposure in rat nucleus accumbens at P24 via single-nucleus RNA-seq + ATAC-seq — the dominant signal isn't cannabinoid receptor dysregulation but coordinated suppression of mitochondrial oxidative phosphorylation gene programs, plus ribosomal + proteasomal machinery changes, with altered chromatin accessibility at NRF1 + YY2 motif-enriched promoters in MSNs; an acute P24 THC challenge in prenatally-exposed offspring exacerbates the mitochondrial deficit and delays MSN maturation. PCE may act through a mitochondrial-bioenergetic chokepoint that gates reward-circuit MSN maturation rather than primarily through CB1 signaling. Then a bioRxiv from Bachtell's group at Colorado on oxycodone self-administration across seven inbred rat strains from the Hybrid Rat Diversity Panel — acquisition timing, total intake, escalation slope, and within-session distribution are all strain-dependent, with the M520/N strain showing distinctive burst-responding stronger in females. The contribution is a behavioral phenotype taxonomy that decomposes oxycodone use into separable heritable sub-trajectories, letting the next OUD GWAS target the right phenotype rather than aggregate intake. Then a bioRxiv from the Tronson lab on chronic combined oral contraceptives (ethinyl estradiol + levonorgestrel) in mice — basal CORT unchanged but impaired GR-mediated dexamethasone negative feedback, with region-specific divergence: dorsal hippocampus enhanced GR signaling + prolonged Fkbp5 induction, ventral hippocampus enhanced MR signaling + blunted stress-induced CRF, PVN reduced MR expression. OC mood liability is not global cortisol suppression but redistribution of how stress signal is read across hippocampal subfields and the PVN — GR/MR balance in specific subregions is the actionable axis. Then a bioRxiv from the Lerch lab on hippocampal neurovascular insulin signaling in depression — multimodal MRS + resting-state fMRI in humans shows hippocampal GABA + lactate correlate with HbA1c and hippocampal-DMN connectivity tracks mood severity independent of adiposity, but the causal mouse model with hippocampus-specific insulin receptor knockdown enhances neuronal metabolism + attenuates anxiety-like behavior. Brain insulin resistance in depression is cell-type-and-context-dependent, hippocampal metabolic signature is the reliable readout — "fix brain insulin signaling" is too coarse as a therapeutic frame (abstract-only; bioRxiv body not yet rendered). Then a bioRxiv from the Robinson lab applying native mass spec + HDX MS to the MC2R + MRAP complex — MRAP depletion drives heavy C-terminal palmitoylation as a compensatory membrane anchor, ACTH binds only the MC2R-MRAP complex + stabilizes the interface, antagonists shift the equilibrium toward MRAP-independent receptor populations + destabilize TM2 even when MRAP is removed. MRAP-family accessory proteins are not just trafficking chaperones — they remodel the receptor conformational landscape, a pharmacological design axis melanocortin small-molecule programs have generally not engaged. Then a bioRxiv on Tox21mer, a 43-million-parameter transformer foundation model encoding Tox21 concentration-response curves + assay metadata into 768-dim embeddings via masked-response reconstruction with auxiliary outcome + AC50 supervision — frozen-embedding probes hit macro-F1 0.985 on agonist/antagonist/inactive + R² 0.87 on log AC50, and masked-only ablation retains near-baseline performance, indicating the representation is learned from curve shape + assay context not from supervision labels. The curve becomes the primary modality rather than the label, with distillation into chemistry-only student models as the obvious next step. Closes on a bioRxiv on ContinuumCellAgent, an autonomous AI-scientist agent that runs literature review + hypothesis + computational experiments + manuscript + adversarial peer review as a single unattended invocation, with protocols grounded in curated research-method checklists that double as reviewer rubrics + a diagnostics layer recording file-based artifacts, message traces, and state transitions. Validated on open-domain QA + biomedical longevity case studies — for agentic-biology builders, the architectural lesson is making the diagnostic layer first-class + grounding plans in the same rubric the critic uses, the path to debugging long-horizon failures. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-21-receptor-and-reason Sun, 21 Jun 2026 12:00:00 +0000 576 Daily roundup for June 21, 2026: bioRxiv APOE4/4 microglial state manifold in AD — spatial proteomics + snMultiome reveals APOE4/4-enriched intermediate population stuck with inflammatory signaling on but metabolic + phagocytic engagement off, localized to gliosis/senescence niches; therapeutic frame shifts from tuning DAM to lowering the activation threshold itself; bioRxiv CP2 brain-penetrant mitochondrial complex I modulator dosed into 19-month APP/PS1 females restores fatty acid oxidation + pyruvate dehydrogenase flux + cholesterol metabolism + cognition in advanced AD via iMiceBrain genome-scale metabolic model, mild partial complex I modulation as disease-modifying counterweight to amyloid/tau frame; bioRxiv prenatal Δ9-THC suppresses mitochondrial OXPHOS programs in rat nucleus accumbens MSNs at P24 (snRNA-seq + snATAC-seq) via altered NRF1 + YY2 promoter accessibility, acute P24 challenge exacerbates deficit — cannabis developmental impact via mitochondrial bioenergetic chokepoint not primarily CB1; bioRxiv Bachtell lab decomposes oxycodone self-administration into acquisition + total intake + escalation slope + within-session burst-pattern across seven HRDP rat strains with M520/N females showing distinctive burst pattern, phenotype taxonomy for next OUD GWAS; bioRxiv Tronson lab chronic ethinyl estradiol + levonorgestrel impairs GR-mediated dexamethasone negative feedback with region-specific divergence (dHC enhanced GR + Fkbp5, vHC enhanced MR + blunted CRF, PVN reduced MR), OC mood liability is GR/MR subregion redistribution not global cortisol suppression; bioRxiv Lerch lab hippocampal GABA + lactate correlate with HbA1c + DMN connectivity tracks mood severity in depressed humans, hippocampus-specific insulin receptor knockdown in mice enhances neuronal metabolism + attenuates anxiety — brain insulin resistance in depression is cell-type-and-context-dependent, hippocampal metabolic signature is the reliable readout (abstract-only); bioRxiv Robinson lab native MS + HDX MS shows MC2R-MRAP antagonists destabilize TM2 even after MRAP depletion + drive C-terminal palmitoylation as compensatory anchor, MRAP accessory proteins remodel receptor conformational landscape not just trafficking chaperones, pharmacological design axis for melanocortin programs; bioRxiv Tox21mer 43M-parameter transformer foundation model encodes Tox21 concentration-response curves + assay metadata into 768-dim embeddings via masked-curve reconstruction (macro-F1 0.985 + R² 0.87 on log AC50 with frozen probes), representation learned from curve shape + assay context not labels, curve becomes primary modality; closes on bioRxiv ContinuumCellAgent autonomous AI-scientist agent grounds plans + reviewer rubrics in shared curated research-method checklists with file-based artifact + message-trace + state-transition diagnostics, architectural template for debugging long-horizon agent failures in biology. Receptor & Reason for June 20, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Leads with what is probably the most consequential structural-pharmacology paper of the week: a bioRxiv preprint solving the cryo-EM structure of human SLC6A15 (B0AT2), a brain amino-acid transporter linked to major depressive disorder by GWAS years ago, in five states across the transport cycle — apo, with proline + leucine + methionine, and with loratadine + tiagabine — revealing that Phe-308 tunes substrate selectivity at the canonical S1 pocket, that loratadine stabilizes outward-occluded via allosteric inhibition at extracellular S2, and that tiagabine cooperatively locks the transporter inward-open at three sites: S1 plus two previously unrecognized intracellular cavities S3 and S4 that, by sequence and structural analysis, are conserved across the entire SLC6 family (SERT, DAT, NET, GABA + glycine transporters). The state-selective intracellular modulators that SSRI/SNRI/DAT-blocker programs have been pursuing for years now have a structural target list — with the caveat that conservation is inferred and SERT-specific S3/S4 occupancy still needs validation. Then a related preprint, updated this week by the Forrest lab at NINDS, finally settles a decades-old question about how chloride helps SERT: reconstituted-proteoliposome assays show Cl⁻ gradients do not drive 5-HT⁺ accumulation (chloride is not co-transported), and cysteine-accessibility assays plus MD show that Cl⁻ binding instead enhances Na⁺ affinity and stabilizes the conserved Arg104–Glu493 extracellular-gate salt bridge — chloride is an architectural cofactor, not a substrate. Forty years of teaching about chloride co-transport at SERT is now a conformational requirement, and the anion dependence of the NSS family bifurcates between true chloride-symporting amino-acid transporters (GlyT1, GAT1) and the monoamine transporters where chloride is architectural only. Then a bioRxiv preprint introducing AmyBind, a computational pipeline for de novo design of polymorph-specific mini-protein binders against amyloid fibrils — two α-synuclein fibril binders designed (one polymorph-specific, useful for differential diagnosis of Parkinson's versus multiple system atrophy where fibril strains differ), engaging the lateral fibril surface rather than the growing end so neither inhibits elongation in vitro and neither reduces fibril uptake or seeding in cells, making them imaging/diagnostic agents rather than therapeutics for now. The framework should transfer to tau strains (AD vs CTE) and TDP-43 strains in ALS subtypes — infrastructure for the next generation of strain-specific neurodegeneration drugs. Then a bioRxiv preprint on TDP-43's mitochondrial localization sequence M1 — equilibrium all-atom MD shows the β4–β5 bridge hydrogen bonds break early in unfolding, exposing M1 (which is normally solvent-inaccessible), and virtual screening of the FDA-approved ZINC15 subset (~2,100) plus a 515,000-molecule non-approved library against this exposed conformation identifies a hit previously characterized as a MAP-kinase-activated protein kinase 2 (MAPKAPK2) inhibitor — a drug-repurposing candidate whose pre-existing MAPKAPK2 pharmacology could either synergize (anti-inflammatory in microglia) or confound interpretation, clean illustration of transiently exposed disordered motif as druggable interface plus repurposing to enter a hard ALS target class, MD-anchored model awaiting biophysical and cellular validation. Then a bioRxiv preprint identifying the mitochondrial cytochrome CYP27A1 as the missing link between vitamin D signaling and mitochondrial dysfunction in HD models — HD patient-derived cells and 3-nitropropionic acid rat model both show reduced CYP27A1, vitamin D supplementation restores CYP27A1 + mitochondrial fusion + complex II function + cognitive/motor deficits, CYP27A1 overexpression alone is sufficient to recapitulate the protective effect (not just VD pleiotropy), and the mechanism is VDR-dependent transcription of CYP27A1 plus other mitochondrial-biogenesis genes creating a feedback loop — a small-molecule-tractable HD target converging with the long-standing complex II dysfunction story, with the caveat that 3-NP is a metabolic toxin model not a CAG-expansion model so YAC128 or zQ175 validation is the next step. Then a bioRxiv preprint introducing Orion, a computer-using AI agent for biomedical image analysis combining an LLM with terminal execution and graphical-interface control in a shared computing environment — drives desktop scientific software via point-and-click rather than API integration, over 90%% accuracy on biomedical database + literature retrieval tasks, learned to operate CellProfiler and QuPath for cellular and tissue image quantification without bespoke integration code, and in a 100-hour autonomous run on a large-scale perturbation imaging dataset produced 52 research reports of which human review flagged 22 (42%% signal rate) as plausible mechanistic hypotheses — the right shape of automation for the CNS phenotypic-screen analysis bottleneck in iPSC neurons and organoids, useful exploratory-hypothesis-generation funnel with human review remaining essential. Then a bioRxiv preprint on DesignMaster, a diffusion model on top of an E(3)-equivariant graph transformer for rational PROTAC design with gated multi-condition fusion injecting linker length and physicochemical constraints throughout the diffusion process rather than just at the end — 34%% improvement in recovery over previous generative methods on PROTAC-DB benchmarks and 52%% reduction in linker-geometry RMSD on one published ternary-complex case; PROTACs are not yet a big CNS modality due to brain-penetration challenges, but if linker design becomes routine it changes what's possible for tau and TDP-43 targets stuck on classical small-molecule approaches. Then a bioRxiv preprint on parent-of-origin effects in female oxycodone self-administration — reciprocal F1 female crosses of Wistar Kyoto more-immobile and less-immobile substrains (which differ in depression-like behavior and substance-use vulnerability) show that females sired by more-immobile fathers escalated oxycodone intake faster when sessions extended from 1h to 4h and consumed more total oxycodone across expanded-access stages, with lick microstructure showing they're driven more by cue-directed seeking during drug-unavailable timeouts than by drug subjective value — a parent-of-origin effect on female addiction vulnerability that, if it replicates, is a methodological lever for finding the imprinted or epigenetically inherited loci mediating addiction risk. Closes on a bioRxiv preprint testing the attractor-stability hypothesis of OCD in a chronic-quinpirole (D2/D3 agonist) rat model with hippocampus and anterior cingulate recordings — ACC firing rate increases gradually within a session, paralleling the within-session escalation of behavioral stereotypy, and hippocampal temporal (not spatial) firing stability increases at both single-neuron and neuronal-pair levels, linked specifically to theta-rhythm modulation; both behavior and at least one neural signature track with the attractor-stability prediction, making this a methodological template for testing computational psychiatry hypotheses against concrete neural recordings. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-20-receptor-and-reason Sat, 20 Jun 2026 12:00:00 +0000 805 Daily roundup for June 20, 2026: bioRxiv cryo-EM of human SLC6A15 (B0AT2) in five transport-cycle states — Phe-308 tunes S1 substrate selectivity, loratadine allosteric inhibition at extracellular S2, tiagabine cooperative multi-site inhibition at S1 plus two previously unrecognized intracellular cavities S3 + S4 that are conserved across the entire SLC6 family (SERT, DAT, NET, GABA + glycine transporters), structural target list for state-selective intracellular monoamine-transporter modulators with the caveat that SLC6 conservation is inferred and SERT-specific S3/S4 occupancy still needs validation; bioRxiv Forrest lab at NINDS settles decades-old question on how chloride helps SERT — Cl⁻ is not co-transported (no Cl⁻-gradient-driven 5-HT⁺ accumulation in reconstituted proteoliposomes) but enhances Na⁺ affinity + stabilizes the conserved Arg104–Glu493 extracellular-gate salt bridge, an architectural cofactor not a substrate, bifurcates NSS family anion-dependence between true Cl⁻-symporting amino-acid transporters and monoamine transporters where Cl⁻ is architectural only; bioRxiv AmyBind pipeline for de novo polymorph-specific mini-protein binders against amyloid fibrils — two α-synuclein fibril binders (one polymorph-specific, useful for PD vs MSA differential diagnosis), lateral-surface binding mode so no inhibition of elongation/seeding (imaging/diagnostic for now, framework transfers to tau strains + TDP-43 strains); bioRxiv equilibrium all-atom MD shows TDP-43 M1 mitochondrial localization sequence is structurally labile via early loss of β4–β5 bridge, virtual screen identifies a previously-characterized MAPKAPK2 inhibitor as M1 binder — drug-repurposing candidate for ALS with possible synergistic MAPKAPK2-anti-inflammatory effects, MD-anchored model awaiting biophysical/cellular validation; bioRxiv vitamin D – CYP27A1 – VDR feedback rheostat rescues HD mitochondrial + cognitive/motor deficits in patient-derived cells + 3-NP rat model — CYP27A1 overexpression alone recapitulates protection, small-molecule-tractable HD target converging with complex II dysfunction story, caveat 3-NP is metabolic-toxin not CAG-expansion model; bioRxiv Orion computer-using AI agent for biomedical image analysis combines LLM + terminal + GUI control in shared compute environment, drives CellProfiler + QuPath without bespoke integration, >90%% retrieval accuracy + 100h autonomous run on perturbation imaging produced 52 reports with 22 (42%% signal rate) flagged as plausible mechanistic hypotheses, right shape for CNS phenotypic-screen automation; bioRxiv DesignMaster E(3)-equivariant graph-transformer diffusion for rational PROTAC design with gated multi-condition fusion (linker length + physicochemical constraints throughout diffusion) — 34%% recovery improvement + 52%% RMSD reduction on published case, infrastructure for tau + TDP-43 PROTACs if brain-penetration challenges can be solved; bioRxiv reciprocal F1 female cross of Wistar Kyoto more-immobile + less-immobile substrains shows parent-of-origin effect on female oxycodone self-administration — sired-by-more-immobile females escalate faster during 1h→4h transition + cue-directed seeking during timeouts, methodological lever for finding imprinted/epigenetically inherited addiction-risk loci; closes on bioRxiv chronic-quinpirole rat OCD model testing attractor-stability hypothesis — ACC firing rate gradually increases within session paralleling behavioral stereotypy, hippocampal temporal-firing stability increases at single-neuron + pair level linked to theta-rhythm modulation, methodological template for testing computational psychiatry against concrete neural recordings. Receptor & Reason for June 19, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Leads with a bioRxiv preprint that is the cleanest causal dissociation of psilocybin's two main effects this year: cell-type-specific knockout of 5-HT-2A receptors in layer 5 cortical pyramidal neurons abolishes psilocybin-induced synapse formation + maturation + elimination of pre-existing synapses while preserving the head-twitch response, the standard rodent proxy for hallucination — the 5-HT-2A pool that mediates rapid plasticity is anatomically distinct from the pool that mediates the perceptual effect, offering a cell-type wedge alongside the biased-agonism programs in the non-hallucinogenic-psychedelic translational program. Then a bioRxiv preprint providing the molecular grammar to pair with that dissociation — single-nucleus RNA sequencing of mouse dorsal medial frontal cortex at 1/2/4/24/72 hours after psilocybin or ketamine, with baseline cortical excitatory and GABAergic subtypes selectively expressing different serotonin receptor transcripts (substrate for cell-type-specific 5-HT-2A engagement already visible in transcriptome), psilocybin's transcriptional response is bimodal at 1h and 72h with excitatory-neuron synaptic plasticity programs vs GABAergic mitochondrial/metabolic programs, and ketamine peaks distinctly at 2-4h. Then a bioRxiv preprint on trustworthy agentic genomics through versioned skill libraries — 9 frontier LLMs benchmarked across 44,550 scored evaluations on 110 pharmacogenomic cases plus real star-allele diplotypes from 7,000+ individuals in three ancestrally diverse populations, letting the model reason was stochastic and unsafe, grounding it in correct clinical guidelines via retrieval-augmented generation paradoxically increased lethal-class errors, and encoding the validated decision logic as a versioned skill executed as code made the pharmacogenomic mapping exact, auditable, and identical across models while removing the ancestry gradient from the clinical mapping — trustworthiness is a property of pipeline architecture not the underlying model, and the versioned-skill-plus-execution pattern is the one to copy anywhere touching pharmacogenomics. Then a bioRxiv preprint introducing FlowBench, decomposing agentic bioinformatics performance into planning, fault recovery, biological interpretation, and end-to-end output fidelity across 23 models from 3 providers via a provider-agnostic FlowAgent harness — workflow planning from a named toolchain is largely solved while inferring the appropriate toolchain from biological intent is uniformly difficult across all tiers (44-57%% pass-rate band), the dependency-structured plan plus completeness-reflection drive performance while same-context validator-driven retry actively makes structural quality worse, and fault recovery + data-grounded interpretation remain unsolved with models proposing fixes that force clean exits while leaving data invalid — the architectural lesson is to constrain the agent's choice space, not expand it. Then a bioRxiv preprint on a convergent metabolic-kinase signaling axis linking Parkinson's disease and multiple system atrophy — iPSC-derived midbrain spheroids from monogenic PD, idiopathic PD, MSA, and controls with integrated proteomics + metabolomics + phosphoproteomics reveal highly concordant molecular remodeling despite distinct etiologies, cellular metabolism is the dominant shared disturbance across central carbon, oxidative phosphorylation, branched-chain amino acid catabolism, pantothenate/CoA, and lipid remodeling, with MAPK + mTOR + AMPK + PKA/PKC kinase networks rewired, and the patterns are conserved in postmortem substantia nigra — AMPK + mTOR as candidate synucleinopathy-class drug targets. Then a bioRxiv preprint flagging microRNA-181-d as a regulator of both arms of Alzheimer's pathology simultaneously — miR-181 reduces neprilysin 3'-UTR activity, mRNA, protein, and enzymatic activity (impairing amyloid-β clearance) while increasing tau mRNA and protein, with higher miR-181-d levels associated with greater AD probability in temporal lobe + cerebellum and lower in posterior cingulate of males (region- and sex-specific), and SNPs near the MIR181 locus associated with altered entorhinal cortical thickness — single antagomir intervention point hitting both amyloid and tau arms. Then a bioRxiv preprint linking prenatal alcohol exposure to Alzheimer's risk through γ-secretase — pregnant wild-type B6129 and 3xTg-AD mice fed an ethanol-containing liquid diet during gestational days 13-15 show persistent elevation of APP C-terminal fragments and Notch intracellular domain in cortex across the lifespan, with hippocampal-dependent learning + memory impairment at 3 and 6 months in wild-type and exacerbated 3xTg-AD deficits at 4 months, providing the first mechanistic linkage from PAE to AD/ADRD vulnerability through γ-secretase + Notch and implicating the same machinery that brought down decades of γ-secretase inhibitor clinical programs. Then a bioRxiv preprint testing the newer-generation systemic butenolide insecticide flupyradifurone (FPF, the neonicotinoid replacement) versus imidacloprid in C. elegans — exposure causes significant dendritic blebbing exclusively in dopaminergic neurons (more from FPF than IMI), with impaired dopamine-mediated behaviors, altered mitochondrial morphology in DA neurons, and elevated ROS pathways — the supposedly safer next-generation insecticide is hitting DA neurons harder than the one it was designed to replace, making FPF a candidate next entry to the PD environmental risk literature alongside paraquat and rotenone. Then a bioRxiv preprint introducing CTACIT, the Cell Type-Aware Conservation Inference Toolkit, for fine-mapping neuropsychiatric regulatory variants — integrates sequence conservation scores with cell-type-specific open chromatin from a few mammalian species and imputes function for hundreds more, gives higher heritability enrichment and more fine-mapped variants than nucleotide conservation or human chromatin alone for neuropsychiatric loci, and in vivo reporter assays validate predicted enhancers carrying risk variants near the DRD2 schizophrenia risk locus — a candidate-variant prioritization tool for CRISPR follow-up on psychiatric GWAS hits that respects the cell-type-specificity that brain regulatory biology demands. Then a bioRxiv preprint on striatal direct and indirect pathway projection neurons (D1 and D2 medium spiny neurons) maintaining a short-term action-outcome associative memory during memory-guided decision-making in head-fixed mice — ventrolateral striatum dopamine modulated by reward receipt/omission and recent outcome history, with closed-loop optogenetic activation + inhibition of direct vs indirect pathway during the outcome window or the pre-choice delay bidirectionally biasing future actions away from reward-history predictions, so the striatal pathways hold the last action-outcome memory shaping the next choice — a delay-window pathway-specific intervention is a more precise target than current options for compulsive disorders, OCD, and stimulant addiction. Closes on an arXiv step-back synthesis "Measuring Biological Capabilities and Risks of AI Agents" — not a new benchmark but guidance for interpreting existing ones, arguing that what an evaluation result implies about real biological risk depends heavily on methodological choices (task definition, scaffolding, scoring, trajectory summary, capability threshold) that are usually undocumented, so a paper that reports a frontier agent passing or failing a biological capability evaluation is almost uninterpretable as a risk signal without published design choices — the practical instruction is to read the methods section of every agentic benchmark carefully and ask which design choices are most upward-biasing the headline pass rate, the standard the agentic-AI-in-biology field should hold itself to. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-19-receptor-and-reason Fri, 19 Jun 2026 12:00:00 +0000 958 Daily roundup for June 19, 2026: bioRxiv cell-type-specific causal dissociation of psilocybin's hallucinogenic and neuroplastic effects — 5-HT-2A knockout in layer 5 cortical pyramidal neurons abolishes psilocybin-induced synapse formation + maturation + elimination of pre-existing synapses while preserving head-twitch response (mouse hallucination proxy), anatomical wedge alongside biased-agonism programs in the non-hallucinogenic-psychedelic translational program; bioRxiv single-nucleus RNA sequencing of mouse medial frontal cortex at 1/2/4/24/72h after psilocybin or ketamine — baseline excitatory + GABAergic subtypes selectively express different 5-HT-R transcripts (substrate for cell-type-specific 5-HT-2A engagement), psilocybin bimodal at 1h + 72h with excitatory-neuron synaptic plasticity vs GABAergic mitochondrial/metabolic programs, ketamine peaks distinctly at 2-4h; bioRxiv trustworthy agentic genomics through versioned skill libraries — 9 frontier LLMs across 44,550 scored evaluations + 110 pharmacogenomic cases + 7,000+ ancestrally diverse diplotypes, raw model reasoning stochastic + RAG paradoxically increases lethal-class errors + versioned-skill execution makes pharmacogenomic mapping exact + auditable + identical across models while removing ancestry gradient — trustworthiness is property of pipeline architecture not model; bioRxiv FlowBench decomposes agentic bioinformatics into planning + fault recovery + interpretation + output fidelity across 23 models — planning from named toolchain largely solved + inferring toolchain from biological intent uniformly difficult (44-57%% band) + fault recovery + data-grounded interpretation remain unsolved with models forcing clean exits leaving data invalid, constrain agent choice space not expand it; bioRxiv convergent metabolic-kinase axis links PD and MSA across iPSC-derived midbrain spheroids via integrated proteomics + metabolomics + phosphoproteomics — cellular metabolism as dominant shared disturbance with MAPK + mTOR + AMPK + PKA/PKC rewired, conserved in postmortem substantia nigra, AMPK/mTOR as synucleinopathy-class targets; bioRxiv miR-181-d regulates both Alzheimer's arms — reduces neprilysin 3'-UTR activity + mRNA + protein + enzymatic activity (impairing amyloid clearance) while increasing tau mRNA + protein, region- and sex-specific human associations, SNPs near MIR181 affect entorhinal cortical thickness; bioRxiv prenatal alcohol exposure (GD13-15) disrupts γ-secretase activity persistently in WT and 3xTg-AD mice — elevated APP C-terminal fragments + NICD in cortex + hippocampal-dependent memory impairment, first mechanistic link from PAE to AD/ADRD vulnerability through γ-secretase + Notch; bioRxiv newer-generation insecticide flupyradifurone (neonicotinoid replacement) causes dendritic blebbing exclusively in C. elegans DA neurons (worse than imidacloprid) + impaired DA behaviors + altered mitochondria + elevated ROS, candidate next entry to PD environmental risk literature; bioRxiv CTACIT Cell Type-Aware Conservation Inference Toolkit integrates sequence conservation with cell-type-specific open chromatin to fine-map neuropsychiatric regulatory variants, in vivo reporter assays validate predicted enhancers at DRD2 schizophrenia risk locus; bioRxiv striatal D1 + D2 MSNs maintain short-term action-outcome associative memory during memory-guided decision-making — closed-loop optogenetics during outcome window or pre-choice delay bidirectionally biases future actions away from reward-history predictions, delay-window pathway-specific target for compulsive disorders + OCD + stimulant addiction; closes on arXiv step-back synthesis "Measuring Biological Capabilities and Risks of AI Agents" — what evaluation results imply about real biological risk depends heavily on undocumented methodological choices, read methods sections carefully and ask which choices upward-bias headline pass rates. Receptor & Reason for June 18, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a bioRxiv preprint that is the clearest demonstration this week of ensemble-targeted antipsychotic design — the group identifies a clozapine-responsive prefrontal ensemble in mouse mPFC enriched for schizophrenia risk genes that stabilizes value-guided choice, NMDA receptor hypofunction increases value-independent decision noise and weakens the ensemble's pre-choice transient while clozapine rescues both, chemogenetic inhibition of the ensemble increases decision noise confirming causality, and crucially the authors cross-reference the receptor expression profile of the ensemble against affinity profiles of available compounds to design a rational multi-receptor antagonist cocktail that reactivates the ensemble and rescues decision noise without the increased NREM sleep time and NREM delta power that clozapine drags along — designing for a circuit-level signature rather than for one receptor at a time, with clozapine's sedation/metabolic profile decoupled from the part of polypharmacology that lifts the ensemble. Then a bioRxiv preprint pinning a female-biased lesion site in major depressive disorder to committed oligodendrocyte precursor cells — integrated human postmortem MDD transcriptomics with mouse validation converges on a committed-OPC subpopulation expressing high levels of synaptic genes, the neurexin pathway emerges as the most perturbed signaling axis in MDD with larger disruption in females, weighted gene co-expression network analysis flags GABRG3 (the GABA-A γ3 subunit) as a female-MDD-specific hub gene in those committed OPCs, and conditional deletion of Gabrg3 in committed OPCs in mice recapitulates impaired neuron-OPC synaptic communication, abnormal myelination, and depression-like behaviors — entry point is a GABA-A subunit on a non-neuronal cell with female-specific signal strong enough in human data to anchor a mouse genetic experiment, suggesting a slice of female MDD where the actionable receptor sits on the OPC. Then an arXiv preprint testing open-weights LLMs (Mistral 3.1, DeepHermes, Qwen3) for predicting dementia and depression severity from speech samples collected in standardized history-taking interviews with 154 German-speaking geriatric subjects, with parallel global staging on an observer-based Global Depression Scale aligned with the standard Global Deterioration Scale for cognition — depression severity is predicted well zero-shot (best MAE 0.60, the model just listens and rates) while dementia assessment substantially benefits from using the LLM as a structured-feature pipeline feeding support vector regression (best MAE 0.78, up to 35%% error reduction over zero-shot), and pause-enriched transcripts achieve nearly the performance of human transcriptions making the pipeline viable from raw audio for remote monitoring of geriatric neuropsychiatric mixed presentations. Then an arXiv preprint introducing TxBench-PP, a verifiable benchmark for AI agents on small-molecule preclinical pharmacology decisions — 100 evaluations indexed by program stage / assay type / task structure spanning mechanism-of-action and pharmacodynamic reasoning, compound-target engagement, causal target validation, developability and safety, and translational efficacy, designed so answers cannot be recovered by retrieving facts from literature with agents inspecting real files in a coding environment and returning structured answers graded deterministically — across 16 model-harness configurations comprising 11 models and 4800 trajectories no system reliably recovered the preclinical pharmacology decisions, best Claude Opus 4.8 / Pi 59.3%% pass rate and GPT-5.5 / Pi 55.3%%, so the best frontier agent today is wrong on roughly four of every ten preclinical pharmacology calls when the answer isn't already in training data, with a failure taxonomy worth digging into and an open re-run cadence for the next four quarters as models update. Then a bioRxiv pain pharmacology preprint that retrofits the triptans onto a circuit story — Gi/o-coupled GPCRs screened in CGRP+ mouse and human dorsal root ganglion neurons, Htr1b/Htr1d agonists sumatriptan and zolmitriptan attenuate CGRP+ nociceptor excitability directly in vitro and produce analgesia across multiple pain models including neuropathic pain in vivo, conditional genetic deletion pins the analgesia specifically to Htr1b in A-fiber mechanonociceptors (the fast-conducting fibers), triptan cardiovascular adverse effects are partially Htr1b-independent suggesting an Htr1b-selective agonist could drop the cardiovascular target list for a clean analgesic, and the orphan Gi/o-coupled receptor Gpr19 is nominated as an additional tractable nociceptor pain target. Then a bioRxiv glial pharmacology preprint mapping G-protein-coupled signaling effects on glial TNF production with pharmacological + chemogenetic activation across rat astrocyte and microglia cultures and human iPSC-derived astrocytes — Gs activation produces a stark decrease in TNF, Gq activation also reduces TNF mRNA, while Gi activation in astrocytes and microglia increases TNF both in vitro and in vivo, the directional inversion vs the inhibitory neuron pattern is the key teaching point, and the effect is selective to TNF rather than generalizable to other pro-inflammatory cytokines — cell-type-specific signaling direction must be tested rather than assumed when designing GPCR-based interventions for neuroinflammation. Then a bioRxiv therapeutic-chemogenetics preprint showing that Gq pathway activation specifically in hippocampal CA1 astrocytes rescues memory deficits and synaptic plasticity impairments in a mouse model of ischemic stroke — Gq activation restores dendritic complexity, spine density, long-term potentiation, and behavioral memory while Gi pathway activation in the same cells induces pathological neuronal hyperactivity without cognitive improvement, and fiber photometry reveals that astrocytic Ca²⁺ signals precede neuronal activity by 600 ms during novel environment exploration so astrocytes prime memory encoding rather than respond to it, offering a glia-targeted strategy for post-stroke cognitive impairment that current neuron-centric therapies cannot match. Then a bioRxiv cannabinoid-pharmacology preprint characterizing chronic vapourized high-THC cannabis flower exposure and CB1-antagonist-precipitated withdrawal in adult male rats using a Volcano vaporizer — closer human analog than the standard injected-synthetic-cannabinoid paradigm — reproducing the cannabinoid tetrad effects in vivo and producing reliable rimonabant-precipitated withdrawal including increased total withdrawal scores and somatic signs plus reduced sucrose preference, with behavioural network analysis revealing that both chronic cannabis exposure and withdrawal reorganize the transition structure of behavior beyond what conventional scoring detects (collapsed repertoire and modularity during exposure, shift toward exploratory sniffing in withdrawal). Then a bioRxiv addiction-microbiome preprint identifying a cocaine-associated oral pathobiont — cocaine selectively enhances Streptococcus parasanguinis growth in vitro, antibiotic-pretreated mice orally inoculated with SP develop spatial memory impairment, elevated brain IL-1β, and non-region-specific microglial activation while controls inoculated with S. salivarius, N. flavescens, or vehicle do not, with no detectable bacterial translocation into the brain, and untargeted metabolomics identifies cysteine-S-sulfate and altered histamine metabolism as oral-to-brain metabolite signatures that themselves drive neuroinflammatory and amyloid-associated responses in cell culture — cleanest causality chain yet for an oral-microbiome-to-brain story in psychiatric pharmacology with actionable interventions sitting at the oral microbiome or upstream metabolite enzymes rather than in the brain. Closes on a bioRxiv conceptual paper providing the first cross-condition empirical test of the Entropic Brain Theory by applying a single sample-entropy-of-time-resolved-small-world-topology pipeline to fMRI datasets from psychedelics, modafinil, propofol anaesthesia, and schizophrenia — propofol anaesthesia sits at the low-entropy end, psychedelics and schizophrenia colocalize at the high-entropy end despite very different phenomenology and clinical valence (consistent with 5-HT-2A pharmacology cutting both ways), the result is not reducible to mean functional connectivity fluctuations and is supported by convergent reorganisation of higher-order association cortex under psychedelics + anaesthesia alongside distributed loss of network specificity in schizophrenia, positioning the entropic axis as a candidate translational biomarker for next-generation psychedelic-adjacent compounds where the goal is therapeutic entropy without the dysregulation. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-18-receptor-and-reason Thu, 18 Jun 2026 12:00:00 +0000 807 Daily roundup for June 18, 2026: bioRxiv mouse prefrontal ensemble enriched for schizophrenia risk genes is clozapine-responsive and stabilizes value-guided choice — NMDA hypofunction increases decision noise + weakens ensemble pre-choice transient, clozapine rescues both, chemogenetic inhibition confirms causality, ensemble receptor-expression profile crossed with affinity atlas yields rational multi-receptor antagonist cocktail that reactivates ensemble + rescues decision noise without the clozapine NREM-sleep penalty; bioRxiv female-biased MDD oligodendroglial mechanism — committed OPCs (synaptic-gene-high) are female-specific lesion site, neurexin pathway most perturbed signaling axis with larger female disruption, GABRG3 (GABA-A γ3 subunit) female-MDD-specific hub gene in cOPCs, Gabrg3 cOPC conditional KO recapitulates MDD-like phenotypes including impaired neuron-OPC synaptic communication + abnormal myelination + depression-like behaviors; arXiv open-weights LLMs predict dementia + depression severity from speech in 154 geriatric subjects, depression MAE 0.60 zero-shot, dementia MAE 0.78 via LLM-feature-extraction + SVR (up to 35%% error reduction), pause-enriched transcripts competitive with human transcripts; arXiv TxBench-PP small-molecule preclinical pharmacology benchmark across 100 evaluations + 16 model-harness configs + 11 models + 4800 trajectories — best Claude Opus 4.8/Pi 59.3%% pass, GPT-5.5/Pi 55.3%%, no system reliably recovers preclinical pharmacology decisions designed so answers cannot be retrieved from literature; bioRxiv triptans attenuate CGRP+ DRG nociceptor excitability via Htr1b with conditional deletion pinning analgesia specifically to Htr1b in A-fiber mechanonociceptors, cardiovascular side effects partially Htr1b-independent (clean-analgesic potential), orphan Gpr19 nominated as additional nociceptor target; bioRxiv G-protein-mediated TNF production inverts direction in glia — Gs↓TNF + Gq↓TNF + Gi↑TNF in astrocytes + microglia both in vitro + in vivo, opposite to neuron pattern, selective to TNF only; bioRxiv chemogenetic Gq activation in CA1 astrocytes rescues post-stroke memory + LTP + spine density + dendritic complexity while Gi induces hyperactivity without cognitive benefit, astrocytic Ca²⁺ leads neurons by 600 ms during novel environment exploration; bioRxiv chronic vapourized high-THC cannabis flower 7d + rimonabant-precipitated withdrawal in male rats — cannabinoid tetrad + withdrawal signs + reduced sucrose preference, behavioural network analysis reveals reorganization of behavioral transition structure beyond conventional scoring; bioRxiv cocaine selectively enhances Streptococcus parasanguinis growth in vitro, antibiotic-pretreated mice inoculated with SP develop spatial memory impairment + elevated brain IL-1β + microglial activation without bacterial translocation, cysteine-S-sulfate + altered histamine metabolism identified as causal oral-to-brain metabolite axis; closes on bioRxiv first cross-condition empirical test of Entropic Brain Theory with sample-entropy-of-small-world-topology pipeline across psychedelics + modafinil + propofol + schizophrenia fMRI — propofol low-entropy end, psychedelics + schizophrenia colocalize at high-entropy end, entropic axis as candidate translational biomarker for next-generation psychedelic-adjacent compounds aiming for therapeutic entropy without dysregulation. Receptor & Reason for June 17, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a bioRxiv enhanced-sampling molecular-dynamics paper that puts a thermodynamic framework around biased signaling at the mu-opioid receptor — extensive simulations of MOR bound to buprenorphine (FDA-approved OUD partial agonist) vs endomorphin-1 (endogenous unbiased full agonist) reveal that endomorphin-1 lets the receptor sample two distinct active-like ensembles, one G-protein-compatible and one β-arrestin-1-compatible, while buprenorphine restricts access to that second ensemble and holds the receptor in a narrower G-protein-preferring subset; allosteric pathways from ligand to intracellular conformations are mapped, with direct implication that ligand efficacy and signaling preference can be screened from receptor-only simulations in absence of effector, grounding the next generation of biased opioid analgesic design. Then a bioRxiv whole-brain dynamics paper on the COMP360 PsiConnect DB-RCT (NCT03429075) building patient-level effective-connectivity models pre- and post-treatment with psilocybin vs escitalopram in MDD and quantifying deviation from the fluctuation-dissipation theorem as a one-number summary of hierarchical non-equilibrium brain dynamics — psilocybin and escitalopram drive opposite reconfigurations of FDT-deviation with corresponding changes in segregation and asymmetry across the cortical hierarchy, baseline FDT-deviation discriminates responders from non-responders within each treatment arm offering a candidate pretreatment stratification biomarker, and the two interventions emerge as mechanistically distinct rather than interchangeable. Then a bioRxiv AI-driven Alzheimer's drug-discovery preprint following last week's ASMS LILRB4 paper from an independent group — ~500M-compound virtual screen against the LILRB4 interdomain pocket yields nanomolar competitive ApoE-blockers validated by orthogonal biophysical assays plus structural/mutational pocket confirmation, in human iPSC-derived microglia LILRB4 inhibition suppresses SHP1/2 signaling, attenuates NF-κB activation, and restores amyloid-β uptake, and the brain-penetrant oral lead improves cognition, reduces amyloid burden, and dampens neuroinflammation in 5xFAD mice — second distinct chemical scaffold against the same neuroimmune checkpoint within ~10 days, moving the Alzheimer's microglia-checkpoint thesis from druggability question to lead optimization. Then a bioRxiv genomics preprint introducing a "liquidity" metric to convert postmortem snapshot gene-expression into a developmental dynamics series — neurotypical individuals carrying elevated schizophrenia polygenic risk show generalized delay in developmental liquidity strongly converging on delayed GABA-A receptor functional maturation, with elevated NKCC1/KCC2 chloride-transporter ratio (the perinatal GABA switch) detectable in high-risk pre-clinical brains, anchoring the schizophrenia E/I-balance model to a specific perinatal window and strengthening rationale for risk-stratified NKCC1-inhibition trials in schizophrenia-high-risk youth. Then a bioRxiv population-level dendritic spine nanostructure analysis across multiple mouse models of neuropsychiatric disorder showing schizophrenia models cluster by an enriched small-volume spine subpopulation + reduced nascent spine growth while autism models cluster by an enriched large-volume spine subpopulation — Ecrg4 (secretory peptide gene) elevated in SCZ models, knockdown rescues small-spine phenotype and impaired spine dynamics; two convergent SCZ preprints in 4 days pinning the disorder to developmental synaptic biology at both molecular and morphological layers. Then a bioRxiv environmental neurotoxicology piece — chronic oral mancozeb (manganese ethylene-bis-dithiocarbamate fungicide) at 0.5 μg/kg/day reduces hippocampal + corpus callosum GFAP indicating astrocyte atrophy, at the human acceptable daily intake (30 μg/kg/day) produces hippocampal astrocyte atrophy + open-field hyperlocomotion in one week, with the mechanism nominated as Orai1/STIM1-mediated SOCE inhibition (abolished by Orai1 or STIM1 knockdown, not by TRPA1 knockdown) leading to depleted ER Ca²⁺ stores, lost P2Y1-evoked Ca²⁺ transients, and selectively reduced IPSC frequency consistent with reduced astrocyte-mediated GABA release — SOCE-targeted environmental-toxicology mechanism in a druggable Orai1/STIM1 framework. Then a bioRxiv glymphatic-transport methodological paper using matched-dose long-vs-short cisterna-magna tracer-infusion paradigms to confirm α2-adrenergic agonist (ketamine/dexmedetomidine) anesthesia produces substantially greater perivascular glymphatic CSF influx than isoflurane independent of injection duration, refuting the basal-cistern rapid-clearance critique and anchoring anesthetic-regime choice for intrathecal CNS drug-delivery programs. Then a bioRxiv computational pharmacology piece on ADMETron, an AI-driven SaaS platform combining RNN SMILES embeddings with physicochemical descriptors through gradient-boosting machines across 34 ADMET endpoints achieving 2nd-place TDC benchmark rankings on Ames mutagenicity (AUROC 0.870) and LD50 (MAE 0.573) plus an interactive multi-compound radar visualization for SAR — the commoditization of ADMET prediction as a layer in computational pharmacology. Then a bioRxiv agentic AI infrastructure paper introducing AutoZyme, an autonomous framework that builds benchmarks, identifies bottlenecks, and iteratively tests code changes against an equivalence-preservation harness — improves runtime in >95% of 45 tested functions with 8.5-fold median speed-up and up to 676-fold across 38 Seurat/Scanpy/single-cell-genomics functions, distributed as drop-in AutoZyme-Library replacements; the agents-as-software-optimizers pattern reframes where agentic capability creates value. Then an arXiv agentic generative drug-discovery paper on AMPGAN v3, a dual-discriminator conditional GAN for antimicrobial peptide design with native support for D-amino acids and C/N-terminal modifications, orchestrated end-to-end by PepCraft (multi-agent Planning Agent + specialized generation/filter/verify executors) — 5 candidates across 3 structural classes tested in vitro with 2 showing Gram-positive activity, best candidate MIC 8 μg/mL against B. subtilis; agentic generative drug discovery with non-canonical chemistry as the real-world peptide-drug composition pattern. Closes on a bioRxiv theoretical-neuroscience paper introducing a plasticity-switching RNN where each synapse uses a Hebbian-like rule or a backpropagation-like credit-assignment rule depending on current strength (inspired by EM-observed spine-apparatus calcium reservoirs in large spines) — learns working-memory and cognitive tasks in fewer trials than backpropagation-only networks despite fewer credit-assigning synapses, yields lower-rank more feedforward recurrent topology as a testable connectomic prediction, and offers a mechanistic reconciliation of local Hebbian plasticity with non-local credit assignment via spine-apparatus-switched plasticity rules. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-17-receptor-and-reason Wed, 17 Jun 2026 12:00:00 +0000 1130 Daily roundup for June 17, 2026: bioRxiv enhanced-sampling MD of MOR bound to buprenorphine (FDA-approved OUD partial agonist) vs endomorphin-1 (endogenous full agonist) shows buprenorphine restricts the receptor to a narrower G-protein-preferring active subset while endomorphin-1 samples both G-protein- and β-arrestin-1-compatible ensembles, thermodynamic framework for biased agonism at MOR with allosteric pathways mapped + in-silico bias-screening implication; bioRxiv COMP360 PsiConnect DB-RCT (NCT03429075) whole-brain effective-connectivity models pre/post psilocybin vs escitalopram in MDD show opposite reconfigurations of fluctuation-dissipation-theorem deviation in hierarchical non-equilibrium brain dynamics with baseline FDT-deviation discriminating responders within each arm — pretreatment stratification biomarker candidate + structural-dynamics evidence psilocybin and SSRI are mechanistically distinct; bioRxiv AI-guided ~500M-compound virtual screen vs LILRB4 yields nanomolar ApoE-competitive binders + brain-penetrant oral lead improves cognition + reduces amyloid in 5xFAD, second distinct chemical scaffold against LILRB4 in ~10 days; bioRxiv "liquidity" metric over postmortem gene co-expression shows schizophrenia-PRS-elevated neurotypical brains carry generalized developmental-liquidity delay converging on delayed GABA-A maturation + elevated NKCC1/KCC2 chloride-switch ratio, anchoring SCZ E/I model to perinatal window with NKCC1-inhibition trial rationale; bioRxiv population-level spine nanostructure clusters SCZ models by small-volume spine subpopulation + reduced nascent spine growth vs ASD models by large-volume spines, Ecrg4 elevated in SCZ and knockdown rescues; bioRxiv mancozeb fungicide at acceptable daily intake produces astrocyte atrophy + hyperlocomotion via Orai1/STIM1-mediated SOCE inhibition leading to ER Ca²⁺ depletion + reduced astrocyte-mediated GABA release; bioRxiv matched-dose long-vs-short cisterna-magna tracer-infusion confirms α2-adrenergic (ketamine/dexmedetomidine) anesthesia produces greater glymphatic influx than isoflurane independent of injection duration, anesthetic-regime anchoring for intrathecal CNS delivery; bioRxiv ADMETron AI-driven SaaS ADMET-prediction platform achieves 2nd-place TDC benchmarks on Ames mutagenicity + LD50, ADMET-layer commoditization in comp pharm; bioRxiv AutoZyme agentic framework optimizes Seurat/Scanpy bioinformatics software with 8.5x median speed-up + up to 676-fold across 38 functions via equivalence-preserving iteration, agents-as-software-optimizers pattern; arXiv AMPGAN v3 dual-discriminator GAN with D-AA + terminal-modification support + PepCraft multi-agent Planning Agent yields wet-lab-confirmed Gram-positive AMP (MIC 8 μg/mL vs B. subtilis), agentic generative drug discovery with non-canonical chemistry; closes on bioRxiv plasticity-switching RNN where synapses pick Hebbian vs backprop rules by strength learns cognitive tasks in fewer trials than BP-only nets + yields lower-rank feedforward recurrent topology as connectomic prediction, mechanistic reconciliation of local Hebbian + non-local credit assignment via spine-apparatus switch. Receptor & Reason for June 16, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a bioRxiv preprint that performs the direct mechanistic test the REBUS relaxed-priors hypothesis has needed for years — in humans, recent (<3 weeks) users of serotonergic psychedelics (psilocybin, LSD, 5-MeO-DMT) produced fewer fast saccades and less suppression of EEG delta and theta power to predictively presented stimuli in a saccadic prediction task vs age- and sex-matched non-users, with the effect correlating with time since dosing and replicating across a separate 5-MeO-DMT cohort; in mice, a single 1 mg/kg psilocybin dose drove the same loss of predictive suppression in V1 recordings 24 hours later, coinciding with weakened top-down anterior-cingulate→V1 modulation and visible spine growth on ACa neurons projecting to V1, anchoring the post-acute therapeutic window in plastic remodeling of cortico-cortical feedback circuits and giving a portable EEG biomarker of mechanistic engagement for next-round psychedelic trials. Then a bioRxiv C. elegans + Drosophila neurexin behavioral screen across 190 compounds testing isoform-specific NRXN1 (nrx-1) deletions shows monoamine-targeting compounds rescue allele-specifically, with FDA-approved atypical antipsychotic olanzapine the only compound rescuing both tested alleles plus the conserved human autism-associated L18Q missense variant in C. elegans and additionally rescuing activity deficits and extending survival in Drosophila Nrx-1 mutants; asenapine improves activity but worsens social-feeding showing pharmacological specificity across behavioral domains — monoamine modulation as conserved compensatory axis for NRXN1 loss with already-marketed drug as immediate pharmacogenomic stratification candidate within the NRXN1-deficiency population. Then a bioRxiv preprint identifying KCNQ2/3 potassium channels as the master synaptic brake on cocaine seeking — 14-days-post-withdrawal D1R-MSN hyperactivity in nucleus accumbens, CP-AMPAR surface insertion, and increased spine density all reversed by repeated KCNQ2/3 opener dosing or D1R-MSN-specific constitutively active KCNQ2 expression with concomitant suppression of cocaine seeking and normalized D1-MSN excitability; positions retigabine-class Kv7 openers (including next-generation azetukalner / BHV-7000) for repositioning into addiction-relapse prevention. Then a bioRxiv reward-pharmacology piece on the selective mGluR2/3 antagonist LY341495 — dose-dependent reduction of anticipatory and consummatory sucrose licking with preserved temporal anticipation (mice still know reward is coming but stop working for it), reduced food intake and social preference without impairing locomotion, orofacial motor function, or courtship ultrasonic vocalizations — mGluR2/3 antagonism affects ingestive and social reward through a common motivational mechanism, refining the reward-domain map for treatment-resistant-depression programs built on this mechanism (TS-161 prodrug class). Then a bioRxiv neuroimmune preprint placing microglial CSF1R signaling upstream of stress-induced MMP-2/9 activation in nucleus accumbens core — PLX3397 (the FDA-approved CSF1R inhibitor pexidartinib) pre-dosed ahead of acute restraint stress in rats blunts the MMP-2/9 increase, blunts the microglial CD68-associated phagocytic signature, and dampens the TNFα/Cnr2/Mmp16 transcriptional program; closes the upstream loop on stress-driven accumbens ECM remodeling and provides a peripherally accessible repurposable target. Then a bioRxiv two-hit rat model combining gestational LPS maternal immune activation with peripubertal unpredictable stress and a careful neurochemical map of the adult dorsal hippocampus — MIA raises GluN1, PUS shifts Grin2a/Grin2b ratio toward incomplete NMDA subunit maturation, GABA and GABA-γ2 fail to compensate, MIA and PUS converge on reduced 2-AG endocannabinoid tone from opposite routes (PUS↑Mgll degradation vs MIA↓Dagla synthesis), and taurine is elevated only in the double-hit while glycine is MIA-driven independent of stress; taurine and glycine emerge as candidate stratification markers for developmental vs stress-driven contributions to adult E/I imbalance. Then a bioRxiv structural-pharmacology piece on TRPM4 cryo-EM with the anthranilic-anilide inhibitor PBA in membrane vesicles and detergent — PBA occupies two entirely different binding pockets at 8°C vs 37°C, with the cold-state canonical S3/S4/S4-S5-linker/TRP-helix pocket relocating at physiological temperature into an S1-S4 voltage-sensing-domain site proximal to the Ca²⁺ regulatory region; methodological wake-up call for the TRP-channel SBDD pipelines running at 4°C in detergent, with the validation that membrane vesicles preserve native-like paralipid lipid binding similar to GDN-solubilized channel. Then a bioRxiv Alzheimer-tau mechanism piece using tau phosphomimetic and phospho-resistant mutants in vitro and in live neurons — tau hyperphosphorylation collapses cooperative envelope binding to microtubules, redistributes tau uniformly along the axon, accelerates microtubule detachment, weakens already-mild KIF5C inhibition, and strengthens KIF1A inhibition through reduced motor processivity and faster motor detachment, with downstream dysregulation of lysosome transport in neurons — concrete tau-cooperativity-plus-motor-protein mechanism for early-tauopathy organelle-trafficking failure to screen disease-modifying tauopathy strategies against. Then a bioRxiv ALS preprint showing that pre-myelin-compaction Cre-mediated excision of mutant SOD1 in the oligodendrocyte lineage slows disease onset, improves motor performance, and prolongs survival in SOD1-G37R familial-ALS mice, while post-compaction excision fails — EM localizes mutant-SOD1 aggregates to paranodal loops and the inner periaxonal tongue (the myelinic nanochannels carrying cytosolic transport to the underlying axon), and in a second model (SOD1-G93A) CNP-depletion-driven channel collapse accelerates disease — familial-ALS initiation and progression driven by oligodendrocyte myelinic-channel integrity loss within a developmentally constrained window, reframing motor-neuron-only therapeutics. Then a bioRxiv computational pharmacology piece on TITAN-BBB — a multi-modal deep-learning architecture combining tabular molecular descriptors, image representations, and text features with attention fusion, trained on the largest aggregated BBB-permeability dataset published to date — achieving 86.5% balanced accuracy on classification (a 3.1 pp improvement over state-of-the-art) and 0.436 MAE on regression (20% error reduction), with model, weights, and dataset openly released on GitHub and Hugging Face for adoption into CNS drug discovery campaigns. Closes on the arXiv LiteOdyssey rare-disease diagnostic agent — single reasoning model plus a thin layer of public biomedical tools plus a policy-iteration-with-human-feedback-built clinical-genetics workflow, no fine-tuning, no ensemble, no large case-retrieval database — achieves state-of-the-art 59.3% Recall@1 across LIRICAL plus PhenoPacket Store (1,243 cases with ~45-53% ultra-rare diseases under 1-in-a-million prevalence) and 60.7% on the harder PhenoPacket subset where the same baseline GPT-5.4 without tools gets 10.7%; a counterexample to the scaling-is-the-only-path assumption in medical AI with audit-transparency fit for the rare-disease deployment context. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-16-receptor-and-reason Tue, 16 Jun 2026 12:00:00 +0000 1082 Daily roundup for June 16, 2026: bioRxiv direct mechanistic test of REBUS relaxed-priors hypothesis in both humans (EEG saccadic prediction in <3-week recent psilocybin/LSD/5-MeO-DMT users vs non-users) and mice (1 mg/kg psilocybin V1 recordings 24 h post-dose) shows less predictive suppression coinciding with weakened top-down ACa→V1 modulation + visible spine growth on ACa→V1 projection neurons — post-acute psychedelic therapeutic window underwritten by feedback-circuit remodeling, EEG-predictive-suppression as portable biomarker; bioRxiv C. elegans + Drosophila neurexin allele-specific 190-compound screen identifies monoamine-targeting compounds as conserved NRXN1-loss compensatory axis with FDA-approved olanzapine rescuing both tested alleles + L18Q missense + Drosophila survival, asenapine dissociates activity vs social-feeding — already-marketed-drug pharmacogenomic stratification candidate; bioRxiv KCNQ2/3 master synaptic brake on cocaine-seeking — D1R-MSN hyperactivity + CP-AMPAR insertion + spine-density increase all reversed by KCNQ2/3 opener dosing or D1R-MSN-specific constitutively active KCNQ2, retigabine/azetukalner-class Kv7 openers as relapse-prevention target; bioRxiv mGluR2/3 antagonist LY341495 reduces anticipatory + consummatory licking while preserving temporal anticipation, reduces food + social reward without motor impairment — common motivational mechanism across ingestive + social reward, refines reward-domain map for TS-161-class TRD pipeline; bioRxiv CSF1R inhibitor PLX3397 (pexidartinib) blunts acute-stress-induced MMP-2/9 + CD68 + TNFα/Cnr2/Mmp16 in NAc-core — microglial CSF1R upstream of MMP-driven ECM remodeling underlying stress-induced accumbens synaptic plasticity with repurposable target; bioRxiv two-hit MIA + peripubertal stress rat model adult dorsal hippocampus shows MIA-driven GluN1↑ + PUS-driven Grin2a/Grin2b shift + opposite endocannabinoid disruption routes converging on reduced 2-AG tone + taurine/glycine as candidate stratification markers for developmental vs stress-driven adult E/I imbalance; bioRxiv TRPM4 cryo-EM with PBA inhibitor at 8°C vs 37°C reveals two distinct temperature-dependent binding pockets (canonical S3/S4/TRP-helix vs S1-S4 VSD-Ca²⁺ proximal) — methodological wake-up for TRP-channel SBDD running at 4°C in detergent; bioRxiv tau hyperphosphorylation collapses cooperative MT envelope binding, redistributes tau, weakens KIF5C inhibition + strengthens KIF1A inhibition via reduced processivity + faster detachment, dysregulates neuronal lysosome transport — concrete tau-cooperativity-plus-motor-protein mechanism for early-tauopathy organelle-trafficking failure; bioRxiv pre-compaction Cre excision of mutant SOD1 in oligodendrocytes (not post-compaction) slows ALS in SOD1-G37R + EM localizes aggregates to paranodal loops + inner periaxonal tongue myelinic nanochannels + CNP-depletion channel collapse accelerates SOD1-G93A — familial-ALS driven by oligodendrocyte myelinic-channel integrity loss in developmentally constrained window, reframes motor-neuron-only therapeutics; bioRxiv TITAN-BBB multi-modal tabular + image + text deep-learning attention-fusion architecture on largest BBB-permeability dataset hits 86.5% balanced accuracy (+3.1 pp) + 0.436 MAE (-20%) with model/weights/data openly released; closes on arXiv LiteOdyssey lightweight single-reasoning-model + public-biomedical-tools + policy-iteration-with-human-feedback rare-disease diagnostic agent hits SOTA 59.3% Recall@1 across LIRICAL + PhenoPacket Store and 60.7% on PhenoPacket subset (vs 10.7% same model without tools) — counterexample to scaling-is-the-only-path medical AI with audit-transparency fit. Receptor & Reason for June 15, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a Zurich/Harvard/MIT collaboration involving the Hannan, Hensch, and Tyagarajan groups in Molecular Psychiatry showing that the schizophrenia-risk-locus secreted glycoprotein Adamtsl3 acts cell-autonomously in parvalbumin cortical interneurons to regulate MMP9-dependent perineuronal net formation and maintenance — adult conditional Adamtsl3 deletion reopens juvenile-like ocular dominance plasticity in mouse visual cortex, the cleanest cell-autonomous mechanism for a schizophrenia risk gene the field has produced in years, putting Adamtsl3, perineuronal nets, MMP9, and critical-period reopening on the same drugable axis alongside chondroitinase ABC for PTSD/addiction/amblyopia plasticity-enabling strategies. Then a bioRxiv allelic series for Adnp — conditional knockout, germline deletion heterozygote, and germline knock-in of the patient L822fs frameshift — showing quantitatively identical cortical hypoplasia and remarkably similar sex-specific Morris water maze learning deficits between the two heterozygotes despite producing different truncated proteins, behavior more sensitive to Adnp dosage than neuroanatomy is, frameshifts behaving as loss-of-function for Helsmoortel-Van der Aa Syndrome behavioral readout giving HVDAS gene-therapy dosage-restoration programs a clear behavioral readout that should respond to partial restoration. Then a Neuropharmacology paper using selective optogenetic photostimulation of tyrosine-hydroxylase-positive locus coeruleus terminals in CA1 in freely behaving mice — robust in vivo Schaffer-collateral LTP that persists for hours, microdialysis confirms elevated noradrenaline plus dopamine plus glutamate in CA1, pharmacological dissection abolishes the LTP separately with a β-adrenergic antagonist, with SCH-23390 D1 antagonist, and with an NMDA antagonist; post-training oLTP strengthens weak object-location memory and prolongs Morris and Barnes maze spatial memory out to 10 days, a clean causal LC-catecholaminergic→hippocampal-LTP→durable spatial memory link giving D1 + β-adrenergic combination dosing for memory consolidation in Alzheimer's and post-anesthesia delirium concrete circuit and receptor rationale. Then a Biological Psychiatry piece dissecting how morphine-withdrawal-conditioned place aversion gets locked in at the cytoskeletal level in the dorsal hippocampus through sirtuin 2 NAD-dependent deacetylase — Sirt2 deacetylates tubulin, recruits calcineurin to microtubules, dephosphorylates end-binding protein 3, drives the drebrin/p140Cap/Src/cortactin/cofilin actin polymerization pathway at dendritic spines, pharmacological Sirt2 inhibition blocks both the cytoskeletal rearrangement and the aversive memory plus contextual fear, giving a non-opioidergic molecular handle on the relapse-driving aversive memory in opioid use disorder with broader PTSD read-through. Then a Neuropharmacology structural-pharmacology paper on α-conotoxin LvIC variant D1G-ΔQ14 achieving 19 nM potency at rat α6/α3β4 nicotinic acetylcholine receptor vs >10 μM at α3β4 — 67 site-directed α-subunit mutations identify E152, I157, and T195 as the selectivity-driving residues; opens dissection of dopaminergic-terminal α6-containing nAChRs in nicotine addiction, L-DOPA dyskinesia, and chronic pain models that have been stuck on non-selective compounds, and provides rational starting point for small-molecule mimetic design. Then a bioRxiv cell-biology paper using cryo-electron tomography and genetic perturbation across ALG-2, CASM, PITT, and VPS13C lysosomal damage-and-repair components to map how tau pre-formed fibrils seed templated misfolding via endolysosomal escape — PITT pathway and Parkinson's-risk lipid transporter VPS13C strongly restrict tau seeding in neurons and astrocytes, CASM matters only in astrocytes, ALG-2 plays a smaller role, and cryo-electron tomography shows tau PFFs do not visibly damage the lysosomal membrane but rather nucleate aggregate co-formation with reversible pores mediating cytosolic escape; small-molecule modulators of the PITT pathway or VPS13C as a candidate disease-modifying lever convergent across tauopathy and Parkinson's. Then a bioRxiv strain-biology paper on multiple system atrophy showing limited proteolysis, conformational stability assay, seed amplification, and M83 mouse propagation across MSA-C cerebellar-variant and MSA-P parkinsonian-variant brain extracts yield biochemically and functionally indistinguishable α-synuclein aggregates, with identical mouse disease kinetics and deposition patterns — MSA clinical subtypes do not arise from distinct α-synuclein strains, the same strain initiating in different brain regions drives clinical heterogeneity, single-conformer immunotherapies are likely equipotent across MSA subtypes which is good for trial design. Then a bioRxiv biophysics paper on the A2A adenosine receptor reconstituted in nanodiscs of varying size with varying anionic lipid content — POPS and POPG form lipid clusters within the nanodisc sequestered in part by positive residues on the membrane scaffold protein, the receptor sees only a fraction of nominal anionic lipid, threshold concentration for full active conformation is higher for POPS than for POPG and scales with nanodisc size, targeted scaffold-protein engineering reduces the threshold and tunes anionic lipid accessibility — explains GPCR cryo-EM reconstitution inconsistency and offers a structural-engineering fix, with broader read-through to native HDL particles and likely native membrane microdomains. Then a bioRxiv stroke neuroprotection paper on capsinoids, non-pungent peripheral TRPV1 agonists, inducing 2-4°C mild hypothermia in 18-20-month-old mice via targeted peripheral thermoeffector activation without triggering shivering — post-stroke IP capsinoid dosing in permanent dMCAO reduces PSD3 infarct volume 48% and PSD30 cortical atrophy 44% with concordant gait/grip/foot-fault gains, MCAO/R survival jumps from 33% to 80% through PSD3, aged-mouse data materially strengthens the translational case for capsinoid-induced awake-patient stroke hypothermia adjunct. Then a bioRxiv comp-neuro paper introducing DigiMus, a multi-region spiking neural network that adds three-node circuit-motif regularization from 38,481 reconstructed mouse neuronal morphologies across ~50 brain regions as a soft structural prior to a trainable sequence-modeling architecture — matches or modestly outperforms TCN/LSTM/Transformer baselines on 18 rule-based cognitive tasks with larger gains on hard decision-making, small consistent improvements over structure-free baselines on real auditory/lick-timing/visual decoding datasets while preserving motif priors in trained connectivity, the right way to use connectome statistics as inductive bias rather than decoration. Closes on bioRxiv M3A Multistep Multimodal Multiomic Agentic framework systematically benchmarking LLM-based agents at multi-omic single-cell analysis across 11 cancer types on autonomous cell-type annotation, hypothesis generation from gene programs, and human-AI copilot settings — agents are effective at broad exploratory data interrogation but domain experts remain critical for methodological guidance and biological synthesis, principled evaluation tooling open-sourced for the next generation of biology agents. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-15-receptor-and-reason Mon, 15 Jun 2026 12:00:00 +0000 1098 Daily roundup for June 15, 2026: Zurich/Harvard/MIT Hannan-Hensch-Tyagarajan Molecular Psychiatry schizophrenia-risk glycoprotein Adamtsl3 acts cell-autonomously in PV interneurons to regulate MMP9-dependent perineuronal nets, adult conditional deletion reopens juvenile-like ocular dominance plasticity in visual cortex — cleanest SCZ-genetics-to-critical-period mechanism in years putting Adamtsl3 + PNNs + MMP9 + plasticity-reopening on same drugable axis; bioRxiv Adnp allelic series shows quantitatively identical cortical hypoplasia + sex-specific Morris water maze deficits in deletion heterozygotes vs patient L822fs frameshift heterozygotes, behavior more dosage-sensitive than neuroanatomy is, HVDAS gene-therapy programs get clear behavioral readout; Neuropharmacology optogenetic LC-TH+→CA1 photostimulation drives robust in vivo Schaffer-collateral LTP with elevated NA + DA + glutamate, abolished separately by β-adrenergic + D1 + NMDA blockers; post-training oLTP strengthens weak object-location memory and prolongs Morris/Barnes spatial memory to 10 days, causal D1 + β-adrenergic combination dosing target logic for memory consolidation; Biological Psychiatry Sirtuin 2 NAD-dependent deacetylase in dorsal hippocampus drives morphine-withdrawal aversive memory via tubulin deacetylation → calcineurin/EB3/drebrin/p140Cap actin pathway, pharmacological Sirt2 inhibition blocks both cytoskeletal rearrangement + aversive memory + contextual fear, non-opioidergic molecular handle on relapse-driving aversive memory with PTSD read-through; Neuropharmacology α-conotoxin LvIC variant D1G-ΔQ14 achieves 19 nM potency at α6/α3β4 vs >10 μM at α3β4 with E152 + I157 + T195 as selectivity-driving residues, tool for dopaminergic-terminal α6-nAChR dissection in nicotine addiction + L-DOPA dyskinesia + chronic pain; bioRxiv cryo-ET + genetic screen shows PITT pathway + VPS13C strongly restrict tau seeding from endolysosomes in neurons + astrocytes while CASM matters only in astrocytes, tau PFFs nucleate aggregates without visibly damaging lysosome membrane, PITT/VPS13C modulators as disease-modifying lever convergent across tauopathy + Parkinson's; bioRxiv biochemical + propagation studies across MSA-C and MSA-P show indistinguishable α-synuclein strains + identical mouse kinetics — MSA clinical subtypes driven by anatomy not strain, single-conformer immunotherapies likely equipotent; bioRxiv A2A adenosine receptor nanodisc biophysics + MD shows anionic POPS/POPG cluster within nanodisc sequestered by positive scaffold-protein residues, scaffold engineering tunes lipid accessibility, explains GPCR reconstitution inconsistency with read-through to HDL/microdomains; bioRxiv capsinoid peripheral TRPV1 agonist induces 2-4°C hypothermia in 18-20-month-old mice without shivering, post-stroke IP dosing reduces dMCAO infarct 48% + 30-day atrophy 44% + MCAO/R survival 33%→80%, aged-mouse data strengthens awake-patient hypothermia adjunct case; bioRxiv DigiMus multi-region spiking network adds 3-node motif regularization from 38,481 reconstructed mouse neurons as soft structural prior, matches Transformer/LSTM/TCN on 18 cognitive tasks + small consistent gains on real neural decoding while preserving motif priors, connectome stats as inductive bias not decoration; closes on bioRxiv M3A Multistep Multimodal Multiomic Agentic framework benchmarks LLM agents across 11 cancer single-cell multi-omic datasets — agents broadly explore well but domain experts remain critical for methodological guidance + biological synthesis, principled evaluation tooling open-sourced. Receptor & Reason for June 14, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with Nick Tonks's group at Cold Spring Harbor Laboratory on bioRxiv showing that selective allosteric small-molecule inhibition of PTP1B, the phosphatase Tonks discovered in 1988, produces robust and sustained improvement of multiple Rett-syndrome symptoms in female mouse models — muscle weakness, motor and coordination deficits, and the cardiac and respiratory dysrhythmias that drive Rett mortality — with brain distribution localized to motor coordination and cardio-respiratory control nuclei and concordant year-long PTP1B genetic ablation phenocopy, validating PTP1B as a clinically actionable disease-modifying target reactivating TRKB and insulin/leptin signaling. Then Avner Schlessinger's group at Icahn Mount Sinai virtually docks 3.5 million ZINC20 compounds against a putative allosteric dimer-interface pocket on the citrate transporter NaCT (SLC13A5), tests 54 in a cell-based citrate-uptake assay, optimizes through 26 analogs to six low-double-digit micromolar inhibitors, and confirms Phe362 as the key pocket residue through deep-mutational-scanning agreement — first chemically distinct non-substrate-mimetic allosteric series for the SLC13A5 epilepsy target. Then Claude Brodski's group at Ben-Gurion University combines genetics (BMP polygenic risk score significantly associated with Parkinson's disease at OR 1.21 with proxy-case replication at OR 1.14), genetic and pharmacological BMP inhibition in mice producing PD-like motor deficits and dopaminergic neuropathology, and BMP5/7 administration in an alpha-synuclein preformed-fibril model — neuroprotective concurrent with PFFs and, critically, neurorestorative when delivered after motor symptom onset, a rare post-symptomatic rescue in synuclein models with both polygenic-genetic and functional validation. Then Vanessa Wheeler's group at Massachusetts General Hospital and Harvard Medical School uses somatic CRISPR-Cas9 to knock out Msh3 in HttQ111 HD knock-in mice at 6, 16, and 24 weeks of age bracketing increasing pre-existing CAG expansion — all three intervention timings slow further striatal CAG expansion, reduce nuclear huntingtin pathology, and suppress transcriptional dysregulation with earlier intervention greater impact, and as a bonus suppress the toxic exon-1 Htt1a transcript; provides the preclinical case that an MSH3 therapeutic targeting only a subset of brain cells (which is all any CNS delivery vehicle can manage) can still bend the HD trajectory. Then Andrew Kruse's group at Harvard Medical School presents three cryo-EM structures of the RXFP1 relaxin receptor — ligand-free, bound to the native peptide relaxin-2, and bound to the AstraZeneca small-molecule agonist AZD5462 currently in mid-stage trials — relaxin-2 engages the receptor ectodomain and reorganizes the linker domain into helical secondary structure, while the small molecule binds inside the seven-transmembrane bundle stabilizing a unique active conformation specifically capable of recruiting beta-arrestin, defining divergent activation mechanisms for protein and small-molecule agonists and reopening rational drug-design space for the relaxin receptor family. Then Andrea Hohmann's group at Indiana University on bioRxiv mechanistically dissects how cannabidiol attenuates paclitaxel-induced peripheral neuropathic pain — PPAR-α and PPAR-γ antagonists abolish the anti-allodynic effect while CB1 and CB2 antagonists do not, NAPE-PLD knockouts lose the effect entirely across both development and maintenance phases while GPR55 knockouts retain it, supporting a model in which CBD shifts endogenous N-acylethanolamine signaling through NAPE-PLD to generate in-situ PPAR agonists, with stratified-medicine implications for CIPN. Then Meredith Bauer and Stephen Boehm's group at Indiana University Indianapolis dissects the role of dorsomedial striatum AMPA receptors across the development of aversion-resistant alcohol drinking using quinine-adulterated alcohol — DMS GluA1 is negatively correlated with QuA drinking in aversion-sensitive mice but DMS and DLS GluA1/A2 ratios are positively correlated with drinking in aversion-resistant mice, DMS spiny projection neurons in ARD mice show elevated rectification index indicating recruited calcium-permeable AMPARs, and DMS CP-AMPAR antagonism increases QuA drinking — DMS calcium-permeable AMPARs are an anti-compulsion brake whose recruitment co-defines aversion-resistant drinking, giving the cleanest circuit rationale yet for IEM-1460-class CP-AMPAR antagonist strategies in treatment-resistant alcoholism. Then Rosemary Bagot's group at McGill in Neuropsychopharmacology with fiber photometry of D1 and D2 medium spiny neurons in NAc medial shell during a mixed-valence conditioning task (cues paired to food reward, foot-shock, or no outcome) — both D1 and D2 MSNs show bidirectional valence responses to conditioned cues only in female mice; both populations show salience-style magnitude-scaling activation to unconditioned reward and shock regardless of valence or sex; D2 but not D1 MSNs show enhanced responses to unsignaled outcomes consistent with prediction-error signaling — the strict D1-go-D2-no-go opponent model of accumbens function does not survive high-resolution photometry, sex differences are large enough to obscure NAc biology when pooled, and dissociation between salience and valence within a cell-type argues for circuit-specific rather than receptor-class targeting in anhedonia. Then Anand Pathak, Earl Miller, and Richard Granger's Dartmouth/MIT/Stony Brook collaboration introduces a multiscale mechanistic core-matrix thalamocortical brain model driven solely by GABA-A modulation without parameter fitting to anesthesia data — the simulation reproduces propofol effects on macaque auditory oddball, anesthetized-human functional connectivity, and selective attenuation of matrix-thalamocortical loops relative to core loops, and predicts a dose-dependent biomarker of elevated residual inter-stimulus cortical activity that was subsequently confirmed in held-out empirical macaque data; a generative framework for translating receptor-level pharmacology into circuit-scale biomarkers in predictive neuropharmacology. Then Jiankun Lyu's group at The Rockefeller University on bioRxiv with the most careful sober assessment to date of AlphaFold3 as a virtual screening engine — AF3 beats DOCK3 retrospectively on DUDE-Z but only via ligand-only decoy bias artifacts, DOCK3 achieves stronger overall enrichment in three large experimental screens (σ2 receptor, D4 dopamine receptor, AmpC β-lactamase, 2500+ molecules) with AF3 contributing mainly to early enrichment, out-of-sample pose reproduction on 8000+ post-cutoff complexes shows AF3 accuracy strongly depends on training-set similarity indicating atomic-pose memorization rather than learned molecular recognition, and in a prospective σ2 head-to-head AF3 hit at 13 percent and found a 13 nM binder but DOCK3 hit at twice the rate; AF3 is a complement to physics-based docking and a useful post-docking re-ranker, not a docking replacement. Closes on the agentic-AI computational pharmacology item — MDForge on arXiv treats molecular dynamics pipeline design as open-ended LLM code generation reshaped by verbal reward from multi-agent debate among synthetic physics experts, designs MD pipelines competitive with human experts on SAMPL host-guest binding benchmarks, and prospectively predicts a novel picomolar CB[7] binder subsequently confirmed by wet-lab NMR — the first agentic-AI MD demonstration with a confirmed wet-lab hit beyond a demo-stage benchmark. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-14-receptor-and-reason Sun, 14 Jun 2026 12:00:00 +0000 1032 Daily roundup for June 14, 2026: CSHL Tonks bioRxiv selective allosteric PTP1B inhibitor in female Rett mouse model produces robust sustained improvement in motor + cardiorespiratory phenotype with year-long genetic ablation phenocopy — strongest rare-disease small-molecule preprint of the week reactivating TRKB + insulin/leptin signaling; Mount Sinai Schlessinger bioRxiv 3.5M-compound ZINC20 virtual screen against NaCT (SLC13A5) allosteric dimer-interface pocket yields six 12-15 μM inhibitors with Phe362 confirmed pocket residue — first non-substrate allosteric series for SLC13A5 epilepsy; Ben-Gurion Brodski bioRxiv BMP polygenic risk score associates with Parkinson's (OR 1.21) and BMP5/7 in α-synuclein PFF mouse model is neuroprotective concurrently + neurorestorative post-symptom-onset — rare post-symptomatic rescue in synuclein model; MGH/Harvard Wheeler bioRxiv somatic CRISPR Msh3 knockout in HttQ111 HD mice at 6/16/24 weeks slows striatal CAG expansion, reduces nuclear huntingtin + exon-1 Htt1a, earlier intervention greater impact — case for partial-cell MSH3 targeting bending HD trajectory; Harvard Med Kruse bioRxiv three cryo-EM structures of RXFP1 (apo, relaxin-2-bound, AZD5462-bound) reveal peptide engages ectodomain via linker helical reorganization while AstraZeneca small molecule binds transmembrane bundle stabilizing β-arrestin-recruiting conformation — divergent activation mechanisms reopen drug-design space; Indiana Hohmann bioRxiv CBD anti-allodynic effect in paclitaxel CIPN blocked by PPAR-α/γ antagonists not CB1/CB2, abolished in NAPE-PLD KO but preserved in GPR55 KO across development + maintenance — endogenous lipid-mediated PPAR mechanism for CBD analgesia; Indiana Indianapolis Bauer/Boehm bioRxiv DMS GluA1/A2 ratio and rectification index identify recruited DMS CP-AMPARs as anti-compulsion brake in quinine-adulterated alcohol model — circuit rationale for IEM-1460-class CP-AMPAR therapeutics in treatment-resistant alcoholism; McGill Bagot Neuropsychopharmacology fiber photometry of D1/D2 NAc medial shell MSNs in mixed-valence conditioning shows bidirectional valence responses in both cell types only in females, salience encoding for unconditioned stimuli across sexes, D2-only prediction error — falsifies strict D1-go-D2-no-go model + argues for circuit-specific anhedonia targeting; Dartmouth-MIT Pathak/Granger/Miller bioRxiv multiscale core-matrix thalamocortical model driven solely by GABA-A reproduces propofol macaque oddball + human FC + matrix-loop attenuation and predicts dose-dependent residual cortical activity biomarker confirmed in held-out macaque data — generative framework for receptor-to-biomarker translation; Rockefeller Lyu bioRxiv AlphaFold3 vs DOCK3 head-to-head across DUDE-Z + σ2/D4/AmpC experimental screens + 8000-complex pose reproduction + prospective σ2 screen shows AF3 hits at 13% with 13 nM binder but DOCK3 hits at 2x rate, repositions AF3 as complement + post-docking rescorer not replacement; closes on arXiv MDForge LLM-agent MD pipeline designer with multi-agent physics-expert debate beats human experts on SAMPL benchmarks and prospectively identifies picomolar CB[7] binder confirmed by wet-lab NMR — first agentic-AI MD result with confirmed wet-lab hit beyond demo stage. Receptor & Reason for June 13, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with Shashank Dravid's group at Texas A&M College of Medicine in Molecular Psychiatry reporting that the orphan delta-family glutamate receptor GluD1 is localized at cholinergic synapses in mouse and primate striatum and behaves functionally as an acetylcholine receptor — with cholinergic, AMPA, and GABA-A receptors all pharmacologically blocked, bath acetylcholine still elicits a NASPM-sensitive current in medium spiny neurons, and GluD1 knockouts lose cholinergic terminals plus excitatory drive at those synapses. GluD1 was filed in the ionotropic glutamate receptor family but glutamate barely gates it; the Dravid paper reassigns it as a third cholinergic receptor class alongside nicotinic and muscarinic, with direct hooks into autism and addiction copy-number-variant hotspots. Then a Biological Psychiatry in vivo PET study using the M1-selective ligand 11C-LSN3172176 in 16 schizophrenia patients and 16 matched controls — reductions of 13 to 19 percent in M1 receptor availability across frontal, temporal, parietal, and occipital cortex plus caudate, putamen, hippocampus, and amygdala, with roughly 44 percent (by DVRCS) or 27 percent (by VT) of patients showing whole-brain deficits greater than 20 percent. The spatial pattern is exactly where the xanomeline-trospium KarXT hypothesis needs it, the subgroup heterogeneity opens a plausible stratification biomarker route for muscarinic agonist or PAM programs, and this is the first replication-quality imaging data tying M1 directly to schizophrenia pathophysiology. Then AbbVie's ABBV-2505 bretisilocin, a short-duration intramuscular psychedelic, starts ethnic-bridging Phase 1 single-dose PK in 26 healthy Japanese, Han Chinese, and non-Asian volunteers (NCT07604558) pairing with AJ Cannon's Psychiatric Times conversation on next-generation psychedelics — the bridging study tells you AbbVie is committed to global development, and the IM short-duration profile is a deliberate move away from the all-day-in-clinic ritual of psilocybin and LSD toward something more like rescue-medication dosing. Then Tomas Palenicek and Nikola Jajcay's Czech Academy group on bioRxiv with resting-state EEG microstates in 15 healthy volunteers in a double-blind placebo-controlled crossover with psilocybin at five time points — peak intoxication shows increased global field power peaks, reduced microstate lifespan, and higher transition frequency, with the repertoire of canonical microstates preserved, and the magnitude of microstate acceleration at peak correlated with both acute subjective intensity and self-reported persisting psychological changes at 28 days; a candidate scalable EEG biomarker for psychedelic-induced state change. Then a Neuropharmacology patch-clamp and behavioral study showing bilateral microinjection of the 5-HT2A agonist TCB-2 into the ventrobasal thalamus suppresses GABA reuptake, raises tonic extrasynaptic GABA-A current in thalamocortical neurons, and induces ethosuximide-reversible spike-and-wave absence seizures in Wistar rats — a counter-intuitive pro-epileptic mechanism of thalamic 5-HT2A relevant to rare seizure reports with high-dose psychedelics in vulnerable patients. Then Amynah Pradhan's group at Washington University in St. Louis on bioRxiv with PN6047, a piperidylidene benzamide selective delta-opioid receptor agonist that has already completed Phase 1 — a single injection blocks chronic cephalic allodynia in the nitroglycerin chronic migraine model and post-traumatic headache, reduces and delays KCl-evoked and optogenetic cortical spreading depression events, chronic dosing does not produce medication-overuse headache itself, and chronic PN6047 prevents sumatriptan-induced MOH; the cleanest preclinical package yet for a selective DOR migraine asset. Then Paul Seidler's group at USC on bioRxiv with PT-13, a coumarin-based tau disaggregator that wedges into AD-brain-derived tau fibrils via a stacking-driven co-assembly mechanism — reduces fibril and oligomer burden without generating soluble oligomeric intermediates (addressing the central worry that has haunted the disaggregator field for a decade), brain-penetrant, well-tolerated, and in a tauopathy mouse model preserves proteasome capacity, synaptic integrity, and behavior. Then Mariangela Scarduzio and David Standaert at UAB on bioRxiv using GRAB-ACh fiber photometry in the unilateral 6-OHDA mouse — dopamine depletion disrupts slow delta-band striatal acetylcholine rhythmicity and replaces it with irregular higher-frequency phasic activity, acute L-DOPA broadly suppresses ACh across frequencies without restoring slow rhythm, chronic dyskinetic L-DOPA further degrades delta-band coordination during the ON state, and amantadine restores low-frequency temporal structure; the temporal structure of cholinergic signaling (not mean level) is the variable that breaks in PD and LID, with implications for M4 PAM and other cholinergic-targeting LID strategies. Then Brendan Walker's group at the University of South Florida Morsani on bioRxiv with eight weeks of alcohol vapor exposure in Wistar rats increasing dynorphin-A immunoreactivity in the islands of Calleja alongside elevated alcohol self-administration and 22-kHz negative-affect ultrasonic vocalizations, with mean dynorphin-A neuron size correlating with USV count only in dependent animals — identifies the islands of Calleja as a previously underappreciated dynorphin-rich AUD-recruited region outside the canonical NAc-shell and amygdala dynorphin/KOR map. Then a Molecular Psychiatry paper on cannabinoid-driven generalization of stress responses — restraint-stress odor conditioning followed three weeks later by vaporized then intravenous THC plus CBD self-administration generalizes the stress response to a neutral odor in male rats only, driven by NAc-core astrocyte retraction from synapses, drop in synapsin-1 density, and matrix metalloproteinase 2 and 9 activation; a tractable molecular axis explaining the acute-calm versus chronic-sensitize discordance in the cannabinoid PTSD literature. Closes on Sangseon Lee's group at Inha University on bioRxiv with CAREPath, a knowledge-graph plus LLM drug repurposing framework that combines a DFS-style module constraining traversal to short disease-gene-drug paths and encoding each as a structured prompt for a biomedical language model with a BFS-style module building one-hop mechanism-context embeddings augmented via pharmacologically related drugs and gene-signature-similar diseases — achieves best overall AUPRC across 5 biomedical KGs and 18 baselines with up to 3.8% improvement, semantic short-path encoding does most of the work, BFS context augmentation adds sparse-evidence robustness, and case studies reproduce recent FDA-approved indication switches. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-13-receptor-and-reason Sat, 13 Jun 2026 12:00:00 +0000 805 Daily roundup for June 13, 2026: Texas A&M Dravid Molecular Psychiatry GluD1 reassigned as ACh receptor at cholinergic synapses in mouse + primate striatum — bath ACh produces NASPM-sensitive current in MSNs with all other ACh/AMPA/GABA receptors blocked, GluD1 KO loses cholinergic terminals + excitatory drive, orphan delta-glutamate-family receptor becomes third cholinergic class; Biological Psychiatry in vivo 11C-LSN3172176 PET shows 13-19% lower M1 muscarinic availability across cortex + caudate/putamen/hippocampus/amygdala in 16 SCZ vs 16 HC with ~44%/27% subgroup >20% deficits, first imaging biomarker hook for muscarinic psychiatric era + KarXT stratification; AbbVie ABBV-2505 bretisilocin Phase 1 ethnic-bridging IM single-dose PK in 26 healthy Asian + non-Asian volunteers signals global development of short-duration psychedelic; Czech Academy Palenicek-Jajcay psilocybin EEG microstate study in 15 volunteers double-blind placebo-controlled crossover shows accelerated microstate transitions + preserved repertoire at peak intoxication correlating with acute subjective intensity + 28-day persisting psychological change; Neuropharmacology bilateral thalamic 5-HT2A agonist TCB-2 reduces GABA reuptake, raises tonic GABA-A, induces ethosuximide-reversible absence seizures in Wistar rats; Washington University Pradhan PN6047 delta-opioid agonist Phase 1 complete blocks chronic migraine + PTH allodynia + KCl/optogenetic CSD without inducing MOH and prevents sumatriptan-induced MOH; USC Seidler tau disaggregator PT-13 wedges into AD-brain tau fibrils without generating soluble oligomeric intermediates, preserves behavior + proteasome + synaptic integrity in tauopathy model; UAB Scarduzio/Standaert GRAB-ACh fiber photometry in 6-OHDA mouse shows dopamine loss disrupts slow delta-band ACh rhythm replacing with phasic activity, chronic L-DOPA further degrades delta coordination during ON state, amantadine restores temporal structure — temporal structure not mean level of ACh is the LID variable; USF Walker 8-week alcohol vapor in Wistar rats elevates dynorphin-A in islands of Calleja correlating with 22-kHz negative-affect USVs in dependent animals only, new dynorphin/KOR-AUD region; Molecular Psychiatry vaporized + IV THC+CBD generalizes stress response from conditioned odor to neutral odor in male rats only via NAc-core astrocyte retraction + synapsin-1 drop + MMP-2/9 activation; Inha University Lee CAREPath KG-LLM drug repurposing combines DFS short-path encoding with BFS mechanism-context augmentation, beats 18 baselines across 5 biomedical KGs with up to 3.8% AUPRC improvement, semantic short-path encoding does most of the work. Receptor & Reason for June 11, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with Autobahn Therapeutics Phase 2 topline AMPLIFY-BD data on elunetirom in bipolar depression — a brain-penetrant CNS thyroid hormone receptor agonist designed to engage central thyroid receptors without driving the peripheral axis. Open-label, 21 adults with bipolar I or II in a moderate-to-severe depressive episode, six-week adjunctive dosing, 16.8-point HAMD-17 reduction at week 6 (p<0.001), 75% response, 50% remission, 9.7-point drop by week 2, mild-to-moderate adverse events with no serious treatment-related events, and resting-state fMRI connectivity changes in depression-relevant regions — mechanistically differentiated bipolar depression asset backed by FDA Fast Track with AMPLIFY-MDD reading out Q3. Then Daniel Karlin's Psychiatric Times interview on DT120 ODT, Definium's pharmaceutical-grade orally-disintegrating lysergide tartrate — Phase 2 Emerge in MDD reads out this quarter, Phase 3 Haven in PTSD initiating, and the functional-unblinding problem for 5-HT2A agonist trials is the pending FDA design-guidance question that will shape the next year of psychedelic clinical work. Then Bruna Panizzutti's group at Deakin University with whole-transcriptome sequencing of iPSC-derived cortical neuron-astrocyte networks from 12 bipolar disorder participants and 12 controls — 191 enriched pathways, with toll-like receptor signaling downregulated as the headline innate-immune signature in BD, consistent with broader IL-6 and inflammasome literature on mood-disorder immune dysregulation though the n-of-12-per-group caveat is real. Then Sergio Osorio's group at Mass General with resting-state MEG-derived functional E/I from critical brain dynamics across 500 cortical parcels in 172 participants spanning ages 6-32 — autism shows elevated fEI in childhood, decreased fEI in adolescence, no group difference in adulthood, an age-independent right dlPFC reduction, and a diverging left inferior parietal trajectory, meaning cortical E/I imbalance in autism is regionally and developmentally specific not static, with implications for age-stratified GABA/glutamate modulator trial design. Then Dhruv Grover's group at UC San Diego on multi-timescale learning signals in Drosophila PPM3 dopaminergic neurons — same neurons carry classical reward-prediction-error dynamics (CS-US shift, omission suppression, exceeded-expectation increase) and slower state-like tonic transitions that track behavioral acquisition, trace conditioning delays both signal types and produces anticipatory outcome-timing responses, single dopamine population multiplexing fast and slow learning signals reopens the multiplex question for mammalian midbrain dopamine and computational psychiatry models of addiction, schizophrenia, and Parkinson's. Then Jingqiong Kang's group at Vanderbilt on a de novo SLC6A1 A305V missense in a patient with myoclonic-atonic epilepsy producing trafficking-defective GAT-1 with ER colocalization rising from ~30% wildtype to ~80% mutant — 4-phenylbutyrate, a clinically available pharmacological chaperone, reduces ER retention to ~40% and restores GABA uptake, and combined PBA plus wildtype GAT-1 gene augmentation produces rescue greater than either alone — a two-axis precision-medicine framework for SLC6A1-related developmental and epileptic encephalopathies with trafficking-defective variants. Then Alexandre Mourot's group at ESPCI on photoswitchable nanobody MalAzoCh-C4 combining high-affinity α7 nicotinic selectivity from a single-domain nanobody with an azobenzene-acetylcholine warhead — wavelength-controlled activation of recombinant α7 in Xenopus oocytes plus robust photocontrol of endogenous α7 on hippocampal interneurons sufficient to modulate firing, a genetically independent subtype-selective photopharmacology platform addressing the α7 selectivity graveyard across schizophrenia cognitive symptoms and Alzheimer's. Then Michela Deleidi's group at the Imagine Institute Paris on a midbrain organoid protocol combining triphasic WNT modulation with dynamic bioreactor culture to produce SNpc-like dopamine neurons enriched in TH+/GIRK2+/ALDH1A1+ markers with SOX6+ identity, enhanced synaptic maturation, robust electrophysiology, and elevated DA release — and critically, α-synuclein preformed fibril exposure drives aggregate formation and DA neuron degeneration, recapitulating PD-relevant vulnerability that earlier VTA-like organoid protocols failed to model, with implications for whether existing iPSC-DA neuroprotection screens have been hitting the right population. Then Crownlands' Gateway platform from Kevin Zhu and colleagues — device-guided olfactory epithelium biopsy plus on-site fixation plus 10x Genomics FLEX RNA profiling across 202 donors including AD and PD patients, released as a 4-million-cell atlas on CELLxGENE, capturing more brain-enriched genes than other clinically accessible tissue sources, with exploratory analyses flagging dysregulation of neuroinflammation, endolysosomal biology, proteostasis, and synaptic maintenance in GWAS-implicated genes — a scalable, serially-accessible living-patient CNS tissue source for target discovery and biomarker development. Then Heather Mortiboys's group at Sheffield on patient fibroblasts from primary mitochondrial Leigh syndrome and Huntington's disease converging on an imbalance between functional and dysfunctional mitochondria despite distinct genetic causes — a high-content phenotypic screen identifies AMPK activator A769662 as rescuing both backgrounds through different pathways with identical net mitochondrial-quality outcome, useful convergent phenotypic screen design across genetically distinct CNS disorders plus a pharmacological class with existing clinical development experience for a stagnant Huntington's neuroprotection pipeline. Closes on arXiv OmniBioTwin, a system-of-twinned-systems framework that organizes health digital twins as modular computational entities coupled by explicit interaction operators across a seven-layer architecture spanning data integration, autonomous twin modeling, cross-scale coupling, temporal synchronization, and human-in-the-loop decision support — a multiscale GLP-1 signaling pathway twin in Alzheimer's disease composes molecular, cellular, and organ-level twins within a unified system, an architectural attempt at the cross-scale composition problem with a well-chosen GLP-1/AD demo and open portability question. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-11-receptor-and-reason Thu, 11 Jun 2026 12:00:00 +0000 978 Daily roundup for June 11, 2026: Autobahn elunetirom Phase 2 AMPLIFY-BD open-label in 21 adjunctive bipolar depression patients reports 16.8-point HAMD-17 reduction at week 6 + 75% response + 50% remission + 9.7-point drop by week 2 + resting-state fMRI connectivity changes — CNS-penetrant thyroid hormone receptor agonist mechanistically differentiated for bipolar depression, FDA Fast Track granted, AMPLIFY-MDD reading out Q3; Definium Karlin Psychiatric Times interview on DT120 ODT lysergide tartrate — Phase 2 Emerge MDD this quarter, Phase 3 Haven PTSD, functional-unblinding problem for 5-HT2A agonist trials as pending FDA guidance; Deakin Panizzutti iPSC-derived cortical networks from 12 BD vs 12 controls show 191 enriched pathways with toll-like receptor signaling downregulated as innate-immune BD signature; MGH Osorio resting-state MEG fEI across 500 cortical parcels in 172 participants spanning ages 6-32 — autism shows elevated childhood fEI, decreased adolescent fEI, no adult group difference, right dlPFC reduction, diverging left inferior parietal trajectory, regionally and developmentally specific E/I imbalance; UCSD Grover Drosophila PPM3 dopaminergic neurons multiplex classical reward-prediction-error with slower state-like learning-stabilization signals, trace conditioning delays both, reopens dopamine multiplex question for mammalian mesolimbic models; Vanderbilt Kang de novo SLC6A1 A305V trafficking-defective GAT-1 rescued by 4-phenylbutyrate pharmacochaperone plus wildtype gene augmentation combination, two-axis precision-medicine framework for SLC6A1 DEEs; ESPCI Mourot photoswitchable α7 nicotinic nanobody MalAzoCh-C4 enables genetically independent subtype-selective photocontrol of endogenous receptors; Imagine Paris Deleidi midbrain organoid protocol with triphasic WNT + bioreactor produces SNpc-like SOX6+ DA neurons that recapitulate α-synuclein PFF-driven Parkinson's vulnerability; Crownlands Gateway 4-million-cell olfactory epithelium atlas from 202 donors releases AD/PD discovery platform on CELLxGENE; Sheffield Mortiboys Leigh syndrome and Huntington's fibroblasts converge on functional/dysfunctional mitochondrial imbalance with AMPK activator A769662 rescuing both through different pathways; arXiv OmniBioTwin system-of-twinned-systems digital twin framework with GLP-1 signaling in Alzheimer's as multiscale demo. Receptor & Reason for June 10, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with Valentina Mosienko's group at the University of Bristol on how fluoxetine elevates astrocytic cyclic AMP — in primary rat astrocytes with FRET cAMP sensors, fluoxetine raises intracellular cAMP by 28% without changing calcium, blocked by either 5-HT2B or adenosine 2B antagonists, fluoxetine pushes astrocytes to release ATP in a 5-HT2B-dependent manner, and microglia depletion abolishes the cAMP effect — the antidepressant cascade requires glial crosstalk where astrocyte 5-HT2B drives ATP release, microglia convert it to adenosine, and astrocyte A2B closes the loop on cAMP, putting microglia inside the SSRI mechanism beyond synaptic serotonin. Then Loren Martin's lab at the University of Toronto Mississauga on placebo analgesia in nerve-injured mice — morphine-conditioning produces robust drug-free pain suppression, whole-brain c-Fos mapping shows distributed cortical/subcortical changes with functional connectivity reorganization, excitatory ACC activation blocks placebo while basomedial amygdala and paraventricular thalamus perturbations have no effect, and cell-type-specific access to ACC mu-opioid-receptor-expressing neurons proves their necessity — a cortical endogenous-opioid circuit gates placebo analgesia. Then Alexey Ostroumov's lab at Georgetown on acute stress potentiating operant reward learning through circuit-specific VTA-GABA changes — prior acute restraint stress enhances operant sucrose self-administration in both sexes with increased temporal coincidence of GABA release events and increased GABAergic input excitability selectively onto dopamine neurons projecting to NAc lateral shell and dorsomedial striatum, KCC2 enhancer CLP290 normalizes the inhibitory transmission and blocks the stress-induced learning potentiation — clean target validation for KCC2 enhancers in stress-driven reward learning. Then Alexander Chubykin's group at Purdue in Neuropsychopharmacology on ventrolateral periaqueductal gray GABAergic neurons mediating fentanyl withdrawal — whole-brain Fos flags PAG as a central withdrawal hub, vlPAG GABA neurons show increased Fos and inhibitory synaptic transmission during withdrawal, and chemogenetic inhibition alleviates withdrawal-induced hyperalgesia — translational hook to the hyperalgesia problem in medication-assisted opioid use disorder. Then Bin Gu's group at Ohio State in Neuropsychopharmacology on cannabidiol in Angelman syndrome model mice — chronic CBD increases REM sleep during the dark cycle, restores the siesta phase, improves sleep homeostasis, and reduces spontaneous seizures after flurothyl kindling in Ube3a-deficient mice — chronic CBD signal on both sleep and seizure phenotypes in a genetic neurodevelopmental model is informative for trial design. Then Hyejung Won and Eva Anton at the University of North Carolina with in vivo CRISPR droplet sequencing across 12 fine-mapped schizophrenia risk genes in postnatal mouse neocortex — 3,031 DEGs recapitulate postmortem SCZ signatures, DEG clustering plus factor analysis plus gene regulatory network inference all converge on ciliary transcriptional programs, and perturbation of key cilia contributors alters ciliary structure — synapse-independent signaling through primary cilia as convergent schizophrenia mechanism outside the standard glutamatergic/synaptic drug-discovery surface. Then Wolfdieter Springer and Fabienne Fiesel at the Mayo Clinic on PINK1 loss in astrocytes driving non-cell-autonomous neuronal death in Parkinson's — human astrocytes have robust PINK1 activity, bulk transcriptomics of PINK1-mutant astrocytes shows homeostatic collapse, co-culture demonstrates glia-driven neuronal damage, and pharmacological autophagy enhancement suppresses the inflammatory secretome — reframes whether mitophagy enhancers in development are engaging astrocytes adequately. Then Moustafa Gabr's group at Weill Cornell using affinity-selection mass spectrometry against an Enamine fragment library to identify a first-in-class small-molecule ILT3/LILRB4 inhibitor LT12 with nanomolar binding by MST and SPR — disrupts the ILT3-ApoE interaction in human iPSC-derived microglia attenuating SHP1/SHP2/NF-κB signaling and restoring amyloid uptake, improves cognition and reduces amyloid burden and neuroinflammation in 5xFAD mice — neuroimmune-checkpoint drug discovery adjacent to the TREM2 program. Then Alik Widge's group at the University of Minnesota on personalized pathway-activation modeling across 27 VCVS DBS stimulation settings in 4 patients running the multi-source interference task — right-sided VCVS stimulation outperforms left-sided for cognitive control, right dorsal pathways connected to dorsolateral and dorsomedial PFC most strongly predict faster reaction times, no left-side pathways significantly predict improvement — imaging-defined target pathways for optimization of psychiatric DBS programming. Then Kuan-lin Huang's group at Icahn Mount Sinai with LinkD, an agentic drug-repurposing platform unifying diffusion-based binding affinity prediction (14,981 drugs by 20,385 targets, best on 8 of 9 BindingDB/Davis/KIBA benchmarks especially in cold-start), proteome-wide selectivity, phenotypic validation across 960 cancer cell lines, and population-scale clinical evidence from 11.5 million Mount Sinai and UK Biobank individuals — headline result is LinkD-prioritized beta-blockers propranolol and carvedilol reducing 5-year prostate cancer incidence vs metoprolol (HR 0.82 and 0.92) corroborated by beta-2 docking and LNCaP growth inhibition — agentic cross-scale architecture beats single-line-of-evidence repurposing predecessors with retrospective-confounding caveat acknowledged. Closes on arXiv Dep-LLM, a training-free depression-diagnosis framework that decomposes long clinical interviews into five clinically aligned themes with chain-of-thought rationale, quantifies epistemic reliability from token-level entropy and modulates within-label and across-theme contributions, then aggregates by confidence — beats zero-shot baseline on all 21 foundation models across 9 metrics, outperforms supervised domain-specific and latest closed-source commercial LLMs without training, treating the LLM less as a diagnostic agent and more as a structured reasoning engine forced to surface per-theme evidence and weight its own contributions by entropy. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-10-receptor-and-reason Wed, 10 Jun 2026 12:00:00 +0000 910 Daily roundup for June 10, 2026: Bristol Mosienko fluoxetine astrocyte cAMP cascade via glial crosstalk — SSRI engages astrocyte 5-HT2B → ATP release → microglia convert ATP to adenosine → astrocyte A2B → cAMP up 28% in primary rat astrocytes with FRET sensors, microglia depletion abolishes the cAMP effect, antidepressant mechanism beyond synaptic serotonin; Toronto Mississauga Martin conditioning-based placebo analgesia in nerve-injured mice maps to ACC mu-opioid-receptor-expressing neurons — ACC excitatory activation blocks placebo, basomedial amygdala/PVT perturbations have no effect, ACC mu-opioid cell-type-specific targeting proves necessity; Georgetown Ostroumov acute stress potentiates operant reward learning via circuit-specific KCC2 downregulation in VTA GABA neurons, KCC2 enhancer CLP290 rescues; Purdue Chubykin NPP vlPAG GABAergic neurons mediate fentanyl withdrawal — chemogenetic inhibition alleviates withdrawal-induced hyperalgesia; Ohio State Gu NPP cannabidiol increases REM sleep, restores siesta, improves sleep homeostasis, reduces flurothyl-kindled seizures in Ube3a-deficient Angelman syndrome model; UNC Won/Anton in vivo CRISPR droplet sequencing across 12 fine-mapped schizophrenia risk genes identifies convergent ciliary transcriptional programs, synapse-independent signaling through primary cilia as new SCZ mechanism; Mayo Springer/Fiesel PINK1 loss in astrocytes drives non-cell-autonomous neuronal death in Parkinson's, pharmacological autophagy enhancement rescues; Weill Cornell Gabr first-in-class small-molecule ILT3/LILRB4 inhibitor LT12 with nanomolar affinity disrupts ILT3-ApoE in iPSC microglia and improves cognition + reduces amyloid in 5xFAD mice; Minnesota Widge personalized pathway activation modeling across 27 VCVS DBS settings in 4 patients identifies right dorsal PFC pathways as key for cognitive control; Mount Sinai Huang LinkD agentic drug-repurposing platform across 14,981 drugs × 20,385 targets unified with 11.5M-individual clinical evidence flags propranolol/carvedilol for prostate cancer (HR 0.82/0.92); arXiv Dep-LLM training-free depression diagnosis via chain-of-thought five-theme decomposition + token-entropy confidence modulation beats supervised domain models on long clinical interviews. Receptor & Reason for June 9, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with Mark von Zastrow's group at UCSF on temporal coding expanding the bandwidth of GPCR-mediated neuromodulation: four Gs-coupled receptors natively co-expressed in essentially all hippocampal pyramidal neurons produce similar cyclic-AMP elevations on acute activation but diverge dramatically downstream — they differ in transcriptional response and in functional responsiveness, with neuropeptide receptors remaining responsive for hours while monoamine receptors rapidly desensitize. The effective chemical bandwidth of cellular neuromodulation is not limited by transducer count, additional temporal coding lets individual neurons distinguish a richer neuromodulatory input, and this reframes how to think about polypharmacology and biased agonism because two ligands producing identical acute second-messenger profiles can still drive different cellular outcomes through different desensitization kinetics. Then Kasia Radwanska's lab at the Nencki Institute on estrogen receptor alpha in the basolateral amygdala regulating alcohol craving in females — convergent mouse IntelliCage AUD-prone stratification, amygdala transcriptomics identifying ESR1 as top transcription factor, local Esr1 knockdown reducing excitatory synaptic transmission and cue-induced seeking, ovariectomy phenocopy, plus three ESR1 polymorphisms tracking binge drinking and craving in a longitudinal human cohort with elevated blood ESR1 transcript in women with AUD diagnosis — female-specific circuit-defined estrogen-receptor target as a tractable handle drug discovery has historically missed. Then Liping Wang's group at the Shenzhen Institutes of Advanced Technology on a neuroimmune feedforward circuit linking REM sleep to stress susceptibility — chronic stress increases splenic nerve activity, monocytes recruited to the brain via choroid plexus, pro-inflammatory microglial remodeling in the external globus pallidus, IL-6/gp130 signaling driving sustained PV+ neuron hyperexcitability, GPe PV+ activity during REM sleep but not wakefulness bidirectionally regulates anxiety-like and defensive behaviors — positions REM sleep as a predictive physiological entry point for stress-related emotional dysregulation. Then Fereshteh Nugent's lab at the Uniformed Services University on a previously unrecognized intrinsic population of CRF-expressing lateral habenula neurons mostly VGLUT2+ with a rostral GAD2 co-expressing subpopulation — chemogenetic activation does not alter place preference or anxiety but biases defensive strategies toward passive action-locking in the visual looming shadow test, with sex-dimorphic cellular and connectivity differences. Then Yohei Kondo's group at Nagoya University with a differentiable reaction-diffusion framework that infers spatially resolved tau amplification rates from tau PET data using MRI-informed forward simulation plus error backpropagation — reconstructs voxel-wise propagation kinetics across the human brain revealing structured non-uniform amplification distinct from static tau burden, transcriptomic integration identifies gene-expression programs tracking regional variation — method-level step up from regression-style tau burden analyses toward per-voxel kinetics as the biomarker substrate anti-tau trials have been wanting. Then Fumiaki Ito's group at the Institute of Prophylactic Pharmacology on a five-residue catalytic peptide SKGQA cleaving amyloid-beta at multiple sites — HADDOCK docking plus MD simulations show the substrate itself acts as a scaffold stabilizing serine-protease-like active geometries from a flexible conformational ensemble, distinct stable-binding and stochastic-attack modes explain multi-site cleavage, the five-mer size should diffuse freely into dense amyloid environments inaccessible to larger proteases — substrate-induced anchoring as a reusable concept for amyloid-degrading catalytic peptides. Then Richard Lee's group at Harvard on recombinant GDF11 improving neurological recovery in mouse models of intracerebral hemorrhage and closed-head TBI out to 28 days with improved neurobehavioral scores, enhanced vascular area, neuronal density, and reduced microglia/macrophage density — a single circulating TGF-beta-family protein hitting two anatomically and mechanistically distinct brain injury models with consistent functional readout. Then Jelena Radulovic's lab at Albert Einstein on hippocampal CA2 contributing to trace fear conditioning through circuit-specific control of CA1 with the non-obvious twist that chronic dCA2 inhibition decreases cue-associated freezing while acute dCA2 to dCA1 projection inhibition increases freezing — global and projection-specific manipulations give opposing answers, which should inform hippocampal subfield drug targeting. Then a Noah AI arXiv paper "AI Scientists Are Only as Good as Their Evidence" running a three-arm ablation of a production drug-asset valuation agent — plain web-only LLM versus structured tools plus 14-dimension playbook plus verifier plus objectivity policy plus red-team scaffold versus proprietary Noah AI corpus, scaffolding lifts tier-in-range accuracy 0.80 to 0.89 and objectivity 3.16 to 3.30 but recovers only 0.38 of gold competitive record vs the proprietary corpus at 0.96, completeness-aware decision utility reaches 7.43 for the proprietary arm versus 1.76/2.57 for the other two because a perfect non-proprietary report is capped at 3.83 by coverage — reasoning scaffolds improve calibration but proprietary evidence sets the upper bound of what the AI scientist can decide, useful counterweight to the current pure-scaffolding wave. Then an arXiv paper on closing the prior-posterior loop in self-reflective LLM molecular design — replaces scalar score feedback with full physicochemical rationale from first-principles calculations (orbital energies, atomic charges, electron densities) fed into a retrieval-augmented self-reflection module, achieves deviation as low as 0.0003 eV on HOMO-LUMO gap targets with 100% success rate on moderate tasks generalizing to dipole moment design across five LLM backbones — mechanistic rationale transforms LLM molecular design from stochastic sampler into causal reasoner, with the open question being whether this extends from electronic-structure targets to empirical/noisy bioactivity targets. Closes lighter on Natale Sciolino's group at the University of Connecticut showing phasic locus coeruleus activation transforms cortical taste representations in primary gustatory cortex — miniscope calcium imaging in awake mice plus optogenetic LC activation shows phasic LC expands dynamic range of population representation along a palatability-relevant axis driven primarily by an aversive shift in representation of every tastant except sucrose, mixture-ratio and concentration axes both stretch and rotate, tonic LC activation engages fewer neurons and does not produce the same expansion — phasic norepinephrine reshapes the geometry of affect-relevant sensory coding rather than just gain-modulating it. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-09-receptor-and-reason Tue, 09 Jun 2026 12:00:00 +0000 854 Daily roundup for June 9, 2026: UCSF von Zastrow on temporal coding expanding the bandwidth of GPCR-mediated neuromodulation — four Gs-coupled receptors co-expressed in hippocampal pyramidal neurons produce similar acute cAMP elevations but diverge downstream in transcriptional response and desensitization kinetics (neuropeptide receptors sustaining for hours vs monoamine receptors rapidly desensitizing), effective chemical bandwidth not limited by transducer count, reframes polypharmacology/biased agonism; Nencki Radwanska estrogen receptor alpha in basolateral amygdala regulates alcohol craving in females — convergent mouse IntelliCage AUD-prone stratification + amygdala transcriptomics ESR1 top TF + local knockdown reducing cue-induced seeking + ovariectomy phenocopy + three ESR1 polymorphisms tracking binge drinking/craving in human longitudinal cohort + elevated blood ESR1 in women with AUD diagnosis; Shenzhen Wang neuroimmune feedforward circuit linking REM sleep to stress susceptibility — splenic nerve → choroid plexus monocyte recruitment → pro-inflammatory GPe microglia → IL-6/gp130 → PV+ neuron hyperexcitability with REM-sleep-gated behavioral effect; Nugent USUHS intrinsic CRF-expressing lateral habenula neurons biasing defensive strategy toward passive action-locking with sex-dimorphic cellular and connectivity differences; Nagoya Kondo differentiable reaction-diffusion inference of spatially resolved tau amplification rates from tau PET — voxel-wise propagation kinetics distinct from static tau burden with transcriptomic integration; Institute of Prophylactic Pharmacology Ito pentapeptide SKGQA "catalytide" cleaving amyloid-beta at multiple sites via substrate-induced serine-protease-like active geometry (HADDOCK + MD); Harvard Lee recombinant GDF11 improves neurological recovery in mouse ICH and closed-head TBI out to 28 days; Einstein Radulovic hippocampal CA2 contributes to trace fear conditioning through circuit-specific control of CA1 with global-vs-projection inhibition giving opposing answers; Noah AI arXiv "AI Scientists Are Only as Good as Their Evidence" three-arm ablation showing proprietary evidence corpus (not reasoning scaffolds) sets the upper bound on drug-asset-valuation agent decision quality, completeness-aware decision utility 7.43 vs 1.76/2.57; arXiv closing the prior-posterior loop in self-reflective LLM molecular design replacing scalar score feedback with first-principles physicochemical rationale (0.0003 eV deviation on HOMO-LUMO targets across five LLM backbones); Sciolino UConn phasic LC activation reshapes geometry of palatability coding in primary gustatory cortex with aversive shift for all tastants except sucrose. Receptor & Reason for June 8, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with the Ramaswamy/Litvak group at UCL Queen Square with the first direct intracranial electrophysiology of reward processing in the human ventral tegmental area, piggybacked on deep brain stimulation surgery for chronic cluster headache in 13 patients doing an instrumental learning task. VTA local field potentials selectively encoded rewarding outcomes over losses and neutrals independent of whether the patient learned the contingency; in the eight participants who did learn, VTA responses to unexpected rewards exceeded expected-reward responses (a reward-prediction-error signal in human VTA recorded directly); and in two sufficient-explorers, VTA activity encoded the expected value of the chosen option during decision-making — as direct a translational confirmation as you currently get between rodent/monkey dopamine reinforcement learning and the human VTA. Then a Wuhan University Molecular Psychiatry paper on NgR-dependent plasticity in a dCA3-CaMKIIα → VTA-DA → NAc circuit for stress-induced depression — chronic unpredictable mild stress raises Nogo receptor expression in hippocampal dCA3 glutamatergic neurons that become less excitable, dCA3 NgR knockdown or glutamatergic excitability boost alleviates depressive-like behavior, the dCA3 glutamatergic neurons project specifically to VTA dopamine neurons with downstream NAc D1/D2 inhibition required — myelin-associated growth-inhibitor signaling now upstream of the canonical hippocampal-VTA-NAc reward circuit as candidate antidepressant target, with male-only CUMS-only caveat. Then a Yale Neuropsychopharmacology paper on the submedius thalamus and its reciprocal OFC connection in incubation of oxycodone craving in male rats — bilateral submedius inactivation reduces oxycodone seeking on abstinence day 15 but not day 1, retrograde tracing plus Fos shows both directions of submedius-OFC activated during incubated seeking, both contralateral and ipsilateral submedius-OFC disconnection block incubated seeking — new thalamocortical node for opioid relapse circuits. Then a Rochester RNAscope paper across A10 VTA, A9 SNc, and A8 RRF in five young macaques showing multiplexed dopamine neurons predominate over TH-only — TH+VGluT2 22%, TH+VGluT2+GAD1 23% vs TH-only 19% across all three subregions — fast synaptic co-transmission alongside dopaminergic neuromodulation is the rule not the exception in primate midbrain, with direct consequences for reading VTA local field potentials including the UCL result. Then a Nagoya City University preprint on PDZD8-deficient mice localizing brain cholesteryl ester accumulation specifically to SNr parvalbumin GABAergic neurons with lipofuscin-like lipid accumulation, reduced SNr projections to thalamus and midbrain dopamine targets, and increased striatonigral input from caudate-putamen — cholesterol metabolism as a basal-ganglia vulnerability and candidate ADHD-like phenotype mechanism outside the standard catecholamine framing. Then a Neuropsychopharmacology paper on adolescent cannabis use quantity, frequency, and problems associated with lower subcortical tissue iron (an MRI T2-star proxy for dopaminergic activity) across 81 adolescents aged 14 to 17 — lower 1/nT2-star associated with more daily concentrate hits, cannabis hours high, use frequency, and CUD severity with VTA as the key region in post-hoc analysis — scalable radiation-free imaging strategy for adolescent substance use research. Then a Yale ABCD regularized canonical correlation analysis on 5,408 adolescents aged 13 to 14 identifying two distinct modes of multimodal brain-psychopathology covariation that predict school impairment a year later — a broad psychopathology mode (higher diffusivity, lower reward activation across networks, lower corticostriatal surface area) and a dimension-dissociating mode separating anxiety from externalizing with inverse impairment relationship — methodologically right direction for adolescent computational psychiatry. Then a Neuropsychopharmacology paper on smartphone ecological momentary assessment plus adaptive design optimization Bayesian delay-discounting and risk-ambiguity decision tasks in 79 smoking cessation participants — reliable parameter estimates from 20 to 30 trials per task, elevated craving plus depression plus ambiguity tolerance predict next-day smoking, first-week models reach AUC 0.76 for predicting cessation success approaching the 0.83 full-study AUC. Then a Kanazawa University multi-omics ATAC-seq plus RNA-seq plus single-cell reference mapping in an adolescent social isolation mouse hippocampal CA1 — calcium-transport enhancers open and synaptic-organization promoters close, AP-1 motif loses accessibility, 106 DEGs including upregulated Fosl2 and Hdac9 with myelination pathway enrichment and increased oligodendrocyte representation — coordinated chromatin and transcriptional remodeling driving adolescent-stress long-term psychiatric vulnerability. Closes on an Astera Institute and UC Davis hippocampal slice test of the temporal-derivative synaptic plasticity rule — positive temporal derivative (low-to-high activity across a 200ms theta cycle) gives LTP, negative gives LTD, both no-change conditions give no net synaptic change even with high overall activity — direct empirical support for a backpropagation-approximating temporal-derivative learning rule over pure Hebbian/rate-based interpretations, cleanest biological constraint yet on biologically-plausible gradient-based credit assignment. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-08-receptor-and-reason Mon, 08 Jun 2026 12:00:00 +0000 894 Daily roundup for June 8, 2026: UCL Queen Square first direct intracranial electrophysiology of reward processing in the human ventral tegmental area during instrumental learning in 13 cluster headache DBS patients — VTA LFPs selectively encode rewarding outcomes over losses/neutrals independent of behavioral learning, unexpected-reward response exceeds expected-reward in learners, VTA encodes expected value of chosen option during decision-making, as direct a translational confirmation of RPE/value coding between rodent/monkey dopamine literature and human VTA as currently possible; Wuhan University Molecular Psychiatry NgR-dependent dCA3-CaMKIIα → VTA-DA → NAc circuit for stress-induced depression — CUMS raises NgR in dCA3 glutamatergic neurons, dCA3 NgR knockdown or glutamatergic excitability boost alleviates depressive-like behavior, dCA3 glutamatergic neurons specifically project to VTA dopamine with downstream NAc D1/D2 inhibition required, myelin-associated growth-inhibitor signaling upstream of canonical hippocampal-VTA-NAc reward circuit as candidate antidepressant target; Yale NPP submedius thalamus + reciprocal OFC connection causal mechanism for incubation of oxycodone craving — bilateral Sub inactivation reduces oxycodone seeking on abstinence day 15 not day 1, Sub-OFC disconnection blocks incubated seeking, new thalamocortical node for opioid relapse circuits; Rochester RNAscope across A10/A9/A8 in five young macaques shows multiplexed dopamine neurons predominate over TH-only (TH+VGluT2 22%, TH+VGluT2+GAD1 23% vs TH-only 19%) — fast synaptic co-transmission alongside dopaminergic neuromodulation is the primate-midbrain rule with consequences for reading VTA LFPs; Nagoya City University PDZD8-deficient mice localize cholesteryl ester accumulation specifically to SNr parvalbumin GABAergic neurons with reduced SNr projections and ADHD-like behavior — cholesterol metabolism as basal-ganglia vulnerability; NPP adolescent cannabis use associated with lower subcortical tissue iron (MRI T2-star dopamine proxy) in n=81 with VTA as key region — scalable radiation-free imaging strategy for adolescent substance use research; Yale ABCD rCCA on n=5408 adolescents identifies two distinct brain-psychopathology modes predicting school impairment one year later — broad psychopathology mode + dimension-dissociating anxiety-vs-externalizing mode, right direction for adolescent computational psychiatry; NPP smartphone EMA + adaptive design optimization Bayesian decision tasks in n=79 smoking cessation participants — first-week models reach AUC 0.76 approaching 0.83 full-study AUC for cessation success prediction; Kanazawa University multi-omics ATAC-seq + RNA-seq in adolescent social isolation hippocampal CA1 — AP-1 motif accessibility reduced, myelination pathway enriched, oligodendrocyte representation increased; Astera Institute + UC Davis hippocampal slice test of temporal-derivative synaptic plasticity rule — positive temporal derivative gives LTP, negative gives LTD, both no-change conditions give no net change even with high overall activity, direct empirical support for backpropagation-approximating learning rule over pure Hebbian/rate-based interpretations. Receptor & Reason for June 7, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a Middlesex University PRISMA systematic review of beta-blocker use in ED cocaine presentations across four retrospective cohorts (N=1140, 503 receiving beta-blockers) — no increase in in-hospital mortality, malignant arrhythmias, MI, or hypertensive surge, with modest systolic BP reductions, GRADE certainty very low to low — quietly refutes the unopposed-alpha doctrine inherited from 1980s case reports that has driven ED beta-blocker avoidance for three decades. Then the Picciotto and Jiang labs at Yale on Shank3 nucleus accumbens encoding of social information using genetic capture, chemogenetics, optogenetics, and one-photon calcium imaging — WT NAc social ensembles encode appetitive social cues while Shank3-Δe4-22 NAc social ensembles actively encode avoidance with NAc hyperactivity and hypermodulatory response to social novelty, suppressing social-ensemble activity during social novelty prevents future avoidance and restores social investigation — reframes ASD social deficit as active aversion not lost motivation. Then a Northeastern genetic-algorithm face-synthesis methodology paper separating cognitive process of building internal emotional representations from endpoint product — at-risk-for-depression group converges on embarrassment representations roughly twice as fast with greater anger-representation evolutionary instability but identical endpoint face intensities. Then a University of Chicago paper with simultaneous scalp-EEG plus thalamic stereo-EEG from 24 full-night recordings (six epilepsy patients, 11 thalamic nuclei) — medial pulvinar has strongest phase locking across subjects, Frontal slow oscillations synchronize pulvinar while Global slow oscillations synchronize centromedian at cortical down-state trough, pre-onset pulvinar activity predicts propagation extent before the SO begins — contradicts uniform-thalamus NREM picture and gives real target for closed-loop neurostimulation. Then the Ilmoniemi group at Aalto with adaptive Bayesian TMS localization for cortical motor representations — halves required stimuli for stable localization vs randomized single-coil reference, converging in 60 stimuli on average with 95% highest-density regions below 10 mm² by 150 stimuli. Then the Schulman lab at Max Planck Biochemistry with the Harper lab at Harvard on cryo-EM structures of FBXO7 + PI31 regulating the proteasome 20S core particle — FBXO7 C-terminal domain engages multiple subunits and blocks beta-5 peptidase, PI31 engages all three catalytic sites revealing previously unknown structural basis for beta-1 inhibition, disease-associated variants disrupt proteasome inhibition and SKP1-FBXO7-PI31 complex assembly. Then the Stahlberg group at EPFL with correlative light and electron microscopy of phospho-serine-129 alpha-synuclein inclusions in postmortem PD substantia nigra — somatic inclusions are consistently fibrillar while membranous-type inclusions are restricted to neuritic processes — resolves the compartmental basis of the Lashuel membranous-Lewy-body observation. Then Auburn's PAG-Agent pathway-analysis virtual research assistant integrating pathway-level statistics + context-aware interpretation + literature-supported reasoning, validated on three Alzheimer's transcriptomic datasets and outperforming six competing LLMs across five common literature-support scenarios in citation-retrieval benchmark. Then BioManus on arXiv, an MCP-native graph-planning biomedical agent built on a BioinfoMCP Compiler converting heterogeneous bioinformatics software into standardized MCP servers plus a typed heterogeneous MCP graph over tools/operations/datatypes/workflow stages — decouples planning complexity from raw tool inventory size at compression ratio Theta(N/(h·m_bar)), beats advanced biomedical agent baselines on BioAgentBench and LAB-Bench. Closes on a curiosity item from Wuhan Sports University on exercise-induced sweat suppressing depression-like behaviors in an LPS mouse model approaching or exceeding fluoxetine on sucrose preference, tail suspension, and elevated plus maze — gene-signature anti-correlation is a legitimate CMap-style discovery pattern but the active component is unidentified and LPS depression-model specificity issues mean anti-inflammatory interventions of almost any kind tend to produce signal here, file as hypothesis-generating not therapeutic claim. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-07-receptor-and-reason Sun, 07 Jun 2026 12:00:00 +0000 1023 Daily roundup for June 7, 2026: Middlesex University PRISMA systematic review of beta-blockers in ED cocaine presentations (four retrospective cohorts N=1140) finds no increase in in-hospital mortality, malignant arrhythmias, or hypertensive surge — quietly refutes the unopposed-alpha doctrine inherited from 1980s case reports; Picciotto + Jiang Yale Shank3 nucleus accumbens social-ensemble encoding shows mutant NAc social ensembles actively encode avoidance not lost motivation, suppressing social-ensemble activity during social novelty prevents future avoidance — inverts pharmacological default toward NAc-dampening over reward-promoting programs; Bex Northeastern genetic-algorithm face-synthesis task separates cognitive process of building emotional representations from endpoint product, at-risk-for-depression group converges on embarrassment ~2× faster with unstable anger templates but identical endpoint intensities — depression-related cognitive axis missed by endpoint-recognition paradigms; University of Chicago simultaneous scalp-EEG + thalamic stereo-EEG (24 nights, 11 nuclei, 6 epilepsy patients) shows nucleus-specific thalamocortical coupling for slow oscillations, medial pulvinar pre-onset activity predicts Frontal vs Global SO propagation — target for closed-loop neurostimulation; Ilmoniemi Aalto adaptive Bayesian TMS localization halves required stimuli for stable cortical motor mapping vs randomized reference; Schulman Max Planck + Harper Harvard cryo-EM structures of FBXO7 + PI31 regulating proteasome 20S core, FBXO7 C-terminal engages multiple subunits blocking beta-5 peptidase, PI31 engages all three catalytic sites with structural basis for beta-1 inhibition, disease-associated variants disrupt proteasome inhibition + SKP1-FBXO7-PI31 complex assembly; Stahlberg EPFL CLEM of phospho-serine-129 alpha-synuclein inclusions in postmortem PD substantia nigra shows somatic inclusions consistently fibrillar while membranous-type restricted to neuritic processes — resolves compartmental basis of the Lashuel membranous-Lewy-body observation; Auburn PAG-Agent pathway-analysis virtual research assistant validated on three AD transcriptomic datasets, outperforms six competing LLMs on citation-retrieval benchmark — domain-specific agent with measurable citation-fidelity advantage; BioManus arXiv MCP-native graph-planning biomedical agent with BioinfoMCP Compiler standardizing bioinformatics tools and typed heterogeneous MCP graph for compact subgraph retrieval, beats advanced biomedical agent baselines on BioAgentBench + LAB-Bench — architectural argument for structured executable capability graphs over flat prompt-level tool retrieval; Wuhan Sports University exercise-induced sweat suppresses depression-like behaviors in LPS mouse model approaching or exceeding fluoxetine — gene-signature anti-correlation is a legitimate CMap-style discovery pattern but LPS depression-model specificity issues + unidentified active component, file as hypothesis-generating not therapeutic claim. Receptor & Reason for June 6, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with the Zurzolo group at Institut Pasteur identifying N-acylphosphatidylethanolamines (NAPEs) as endogenous lipid regulators of LRRK2 kinase activity in Parkinson's disease — boosting NAPE synthesis or inhibiting NAPE-phospholipase D suppresses LRRK2 kinase, restores lysosomal function, and accelerates clearance of alpha-synuclein aggregates, with the result holding in iPSC-derived dopaminergic neurons carrying the LRRK2-G2019S variant. NAPEs had been filed primarily under endocannabinoid-precursor biology; their direct signaling role upstream of LRRK2 reframes the pathway and offers a potentially wider therapeutic window than direct kinase inhibitors that have struggled with peripheral lung liabilities at CNS-engaging doses. Then the Sulzer and Bartolini labs at Columbia on a microtubule-nucleation mechanism for how alpha-synuclein regulates synaptic dopamine release — activity-evoked gamma-tubulin-dependent microtubule nucleation at presynaptic boutons is required for interbouton synaptic vesicle transport and sustained dopamine release during high activity, alpha-synuclein binds directly to gamma-tubulin and the alpha-beta tubulin heterodimer and positively regulates nucleation, and activity-evoked phosphorylation of alpha-synuclein at serine 129 sits in the tubulin-binding region and is both necessary and sufficient for initiation — recasts the standard synucleinopathy pathology marker as an activity-coupled regulator of dopamine release and gives a biochemical handle on dopaminergic synaptic dysfunction in early PD. Then the Selkoe lab at Brigham and Women's on direct binding between amyloid-beta aggregates and the U1 spliceosomal ribonucleoprotein in AD brain — Aβ immunoprecipitation plus deep sequencing identifies U1 snRNA specifically, double immunoelectron microscopy shows Aβ fibrils decorated with U1-70k protein more than tau fibrils, and immunofluorescence plus in situ hybridization confirm U1-70k and U1 snRNA labeling at a subset of amyloid plaques — resolves the long-standing proteomics discrepancy that insoluble U1 RNA-binding proteins track Aβ better than tau and unites Aβ with the TDP-43, FUS, and C9orf72 dipeptide-repeat family of RNA-binding aggregating proteins. Then the Polley group at Mass Eye and Ear and Harvard with optic-fiber GRAB-acetylcholine sensor recordings in auditory, visual, and prefrontal cortex of awake head-fixed mice — multivariate modeling cleanly separates stimulus-locked from arousal-locked ACh and identifies a regional dissociation where sound-evoked prefrontal ACh is largely explained by arousal and movement while auditory cortical ACh retains a stronger stimulus-feature relationship — cortical cholinergic system multiplexes area-specific sensory processing onto shared arousal, donepezil and xanomeline-trospium and varenicline behavioral readouts should be reinterpreted in that light. Then the Huda lab at Rutgers on dorsolateral striatal astrocyte modulation of alcohol-induced stimulation in mice — acute ethanol dose-dependently decreases astrocyte calcium without affecting direct or indirect pathway neuronal activity, astrocyte-specific CalEX calcium-extruding pump mimics the reduction and facilitates ethanol stimulation establishing causality, and GRAB-ACh shows ethanol also inhibits striatal ACh but chemogenetic restoration of ACh does not rescue astrocyte calcium — astrocyte calcium is now a tractable non-neuronal target for AUD pharmacology. Then the Joffe lab at Pittsburgh validating an oral xylazine self-administration model in mice — four-hour drinking sessions with xylazine-adulterated water at 0.01–1 mg/mL achieve pharmacologically meaningful brain concentrations, three-bottle choice reveals mice consume less xylazine than fentanyl or plain water — supports the clinical narrative that xylazine in illicit opioid supply is a contaminant rather than sought-after, and the model is built specifically for polysubstance studies. Then the West lab at Rowan on behavioral flexibility and gut microbiome as predictors of oral oxycodone self-administration in Long-Evans rats — distinct high vs low intake phenotypes emerge, contingency-degradation behavioral flexibility correlates with oxycodone intake, and oral oxycodone exposure alters gut microbiome with microbiome features correlated with both behavioral flexibility and drug-taking — multifactorial vulnerability framework rather than single biomarker. Then the Siegel lab at UC Davis on computational design of two new 5-HT1B serotonin receptor agonists derived from naratriptan for migraine — improved docking scores at the 5-HT1B binding pocket vs the parent compound with explicit acknowledgment that next steps are synthesis feasibility and laboratory characterization — clean early-stage comp pharm example anchored on a clinically used scaffold and a structurally characterized GPCR with honest assessment of what docking scores alone do and do not predict. Then QuantisMol's evaluation of a pretrained Graph Isomorphism Network with attribute masking on the Pgp_Broccatelli Therapeutics Data Commons benchmark — GIN pretrained on ~2M molecules plus an MLP classification head achieves AUROC 0.937 ranking second on TDC leaderboard as of May 2026, beating an XGBoost-on-Morgan-fingerprints baseline at 0.912 — confirmatory data point that graph-based representations plus transfer learning beat traditional fingerprint baselines on small-dataset ADMET tasks. Closes on Mayo Clinic's ChatSpatial schema-enforced agentic orchestration for reproducible cross-platform spatial transcriptomics — LLM selects from pre-validated tool schemas rather than generating free-form code, with domain expertise embedded in schema descriptions for context-aware parameter inference, built on Model Context Protocol unifying 60+ methods across 15 categories spanning Python and R, with replication of two published studies and cross-platform validation across seven LLM backbones demonstrating near-deterministic reproducibility — schema-enforcement-not-code-generation is the credible architectural answer to the LLM-driven bioinformatics reproducibility problem that has blocked real-world trust in agentic analytical pipelines. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-06-receptor-and-reason Sat, 06 Jun 2026 12:00:00 +0000 1034 Daily roundup for June 6, 2026: Zurzolo Institut Pasteur identifies NAPEs as endogenous lipid regulators of LRRK2 kinase — boosting NAPE synthesis or inhibiting NAPE-PLD suppresses LRRK2, restores lysosomal function, clears alpha-synuclein in LRRK2-G2019S iPSC dopaminergic neurons — upstream-of-LRRK2 lipid axis with potentially wider therapeutic window than direct kinase inhibitors; Sulzer + Bartolini Columbia microtubule-nucleation mechanism for alpha-synuclein regulation of dopamine release — gamma-tubulin-dependent presynaptic MT nucleation required for interbouton SV transport, alpha-syn binds tubulin and positively regulates nucleation, serine-129 phosphorylation in tubulin-binding region is necessary and sufficient for initiation — recasts standard synucleinopathy pathology marker as activity-coupled DA release regulator; Selkoe Brigham + Women's amyloid-beta aggregates directly bind U1 spliceosomal ribonucleoprotein in AD brain — IP + deep sequencing identifies U1 snRNA specifically, immunoEM shows Aβ fibrils decorated with U1-70k more than tau fibrils, IF + ISH confirm U1 labeling at amyloid plaques — resolves proteomics discrepancy that insoluble U1 RNA-binding proteins track Aβ better than tau, unites Aβ with TDP-43/FUS/C9orf72-DPR RNA-binding aggregating-protein family; Polley Mass Eye and Ear GRAB-ACh recordings across auditory + visual + prefrontal cortex separate stimulus-locked from arousal-locked ACh — sound-evoked prefrontal ACh is largely arousal/movement-explained, auditory cortical ACh retains stimulus-feature relationship — cholinergic drug behavioral readouts should be reinterpreted; Huda Rutgers dorsolateral striatal astrocyte calcium suppression is causally required for ethanol stimulation, ACh inhibition by ethanol is parallel-not-upstream — astrocyte calcium tractable non-neuronal AUD pharmacology target; Joffe Pittsburgh validates oral xylazine self-administration model in mice — pharmacologically meaningful brain concentrations achieved, three-bottle choice shows mice consume less xylazine than fentanyl or plain water — supports clinical narrative that xylazine in illicit opioid supply is contaminant not sought-after, polysubstance study model; West Rowan behavioral flexibility + gut microbiome predict oral oxycodone self-administration in rats — contingency-degradation flexibility correlates with intake, oxycodone alters microbiome with microbiome features correlated with both behavior and drug-taking — multifactorial vulnerability framework; Siegel UC Davis structure-based design of two new 5-HT1B agonists derived from naratriptan with improved docking scores at the binding pocket — early-stage comp pharm anchored on clinically used scaffold and characterized GPCR; QuantisMol pretrained Graph Isomorphism Network with attribute masking achieves AUROC 0.937 on Pgp_Broccatelli TDC benchmark ranking second on leaderboard May 2026 — graph-based + transfer learning beats fingerprint baselines for small-dataset ADMET; Mayo Clinic ChatSpatial schema-enforced agentic orchestration for spatial transcriptomics — LLM selects from pre-validated tool schemas not free-form code, MCP-based unifies 60+ methods across Python + R, replication + cross-platform validation across seven LLM backbones shows near-deterministic reproducibility — schema-enforcement-not-code-generation is the architectural answer to LLM-bioinformatics reproducibility blocker. Receptor & Reason for June 5, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with the Fenno lab at UT Austin building a Cre-dependent AAV-SaCas9 strategy to knock down Clock selectively in VTA dopamine neurons in mouse — a method-and-model paper confirming robust titer-dependent reduction across targeted sequencing, in situ, and IHC, with changes across behavioral battery + EEG-EMG sleep architecture + slice electrophysiology. Cleanest template I've seen for going after any psychiatric risk gene cell-type-specifically without breeding a flox line, and isolates the mesolimbic dopaminergic arm of Clock biology from suprachiasmatic, peripheral, or non-dopaminergic VTA contributions for bipolar-relevant phenotypes. Then the Karlsgodt lab at UCLA on a transdiagnostic early-psychosis HCP-EP sample (N=124) using partial least squares correlation to map twenty-one PET-derived receptor and transporter distributions across nine neurotransmitter systems onto whole-brain resting-state functional connectivity signatures of five symptom dimensions — the negative-symptom RSFC signature correlates with norepinephrine transporter and vesicular acetylcholine transporter maps, the cognition signature also correlates with VAChT, anatomical signatures confounded by antipsychotic dose and substance use — direct argument for prioritizing noradrenergic and cholinergic mechanisms in the refractory symptom domains where D2-centric antipsychotics fail. Then the Chakravarty group at McGill running whole-brain structural MRI plus connectomics in a phenotype-matched chronic variable stress mouse model (female 7d vs male 21-28d stress duration) — both sexes show neuroanatomical changes in nucleus accumbens and hippocampus but in opposite directions in some regions, the female signature couples depression-like and anxiety-like behavior together while the male signature has two orthogonal anxiety vs social-preference dimensions, cortical-hub-dominant in females vs subcortical-hub-dominant in males, with spatial gene expression pointing to cholinergic signalling genes in females — cleanest mechanistic argument for why mixed-sex preclinical antidepressant screens have been noisy. Then the Barbier lab at Linköping reporting reduced PFC PRDM2 (histone methyltransferase) in postmortem human AUD across both sexes, with viral knockdown in rat dmPFC increasing stress-induced alcohol reinstatement in both sexes without estrous-cycle confound and projection-specific dmPFC-NAc knockdown sufficient and shock-intensity-dependent — sex-equivalent circuit mechanism for stress-driven alcohol relapse, anchors behavioral/chemogenetic dmPFC-NAc interventions that bypass the epigenetic upstream. Then the Unterwald lab at Temple on chronic cocaine + IV SARS-CoV-2 spike protein at two months in rat — spike-plus-cocaine produces persistent weight-gain reduction, lower withdrawal thresholds suggesting increased pain sensitivity, and hippocampal IL-6/IL-10 ratio shift toward pro-inflammatory, with paw withdrawal positively correlated with IL-10 in hippocampus and PFC and an impulsivity-like maze phenotype — specific clinical prediction that cocaine-use-disorder plus Long-COVID comorbidity carries excess pain/inflammatory burden not currently stratified for. Then the Harding lab at Toronto using Random nonstandard Peptide Integrated Discovery (RaPID) to recover nanomolar-affinity macrocyclic peptides against distinct HAP40 epitopes — binding confirmed by SPR, fluorescence polarization, and HDX-MS, with engagement of endogenous HAP40 in cellular lysates and selective isolation of HAP40-containing complexes by macrocycle precipitation — tool paper filling the endogenous-engagement gap for HTT-HAP40 biology in Huntington's, with Suga-platform RaPID macrocycles increasingly the routine answer when selective endogenous PPI ligand is the bottleneck. Then the Lewis group at EPFL on correlative light + electron microscopy of Lewy pathology across cortical regions and substantia nigra in confirmed PD and DLB donors — Lewy body ultrastructure alone does not distinguish PD from DLB (shared diverse maturation from low-density fibrils to compact fibrillar inclusions, plus a previously-unreported electron-dense degenerating αSyn+ cortical neuron population), but quantitative 10000+ mitochondria analysis shows PD has increased SN mitochondrial density and enlargement while DLB shows increased density only in entorhinal cortex with mitochondrial enlargement exclusive to PD — disease identity at ultrastructure is more about regional mitochondrial homeostasis than aggregate morphology. Then the Kucinski lab at Michigan adding transient optogenetic inhibition of basal-forebrain cortical cholinergic projections to rats with dorsomedial striatal dopamine lesions, producing substantially more falls than either disruption alone on the Michigan Complex Movement Control Task with the largest effect on the zig-zag beam — causal preclinical confirmation that PD falls are a dual-system failure not pure dopamine loss, anchoring donepezil/varenicline PD-fall-reduction programs with a bounded upper-limit expectation. Then Fan Fan at Medical College of Georgia showing chronic TPPU (soluble epoxide hydrolase inhibitor, 1 mg/kg/day × 9wk) in a diabetic-related-dementia rat model rescues cortical + hippocampal recognition memory, improves cerebral perfusion and normalizes whisker-evoked functional hyperemia via Kir2.1 upregulation in cerebral capillaries, restores BBB tight junctions ZO-1/OCLN, reduces capillary rarefaction and astrocyte/microglial activation, restores synaptic PSD-95 and SY38, and suppresses pro-inflammatory chemokines — sEH GWAS signal in AD/ADRD makes diabetes-ADRD intersection a natural enrichment population for an early sEH-inhibitor trial. Pivots to a Morphological Shortcuts by LLMs in Pharmacology paper on drug-name affix heuristics — fictitious drug names built from real WHO INN affixes ("wugcillin") elicit class-level pharmacological responses, an attribution framework across 653 real drugs shows models induce drug semantics primarily from the affix without disclosing reliance and sometimes conflate properties across affix-sharing drugs, activation patching localizes the behavior to early-mid layers — non-trivial clinical-safety implication for clinical-decision-support LLMs. Closes light on a Ten Headache Specialists versus AI head-to-head evaluation of a RAG-based agentic framework (Claude Sonnet, GPT-4o, Llama 3.1) on clinical literature summarization — expert summaries preferred but specialists frequently can't reliably distinguish human from LLM, with expert-valued features beyond standard correctness/completeness/conciseness underweighted by LLMs — the win condition for RAG clinical-literature agents isn't beating median specialist on the rubric but matching specialists on decision-shaping features the rubric misses. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-05-receptor-and-reason Fri, 05 Jun 2026 12:00:00 +0000 980 Daily roundup for June 5, 2026: Fenno UT Austin Cre-dependent AAV-SaCas9 cell-type-specific knockdown of Clock in VTA dopamine neurons — method-and-model paper for bipolar-relevant mesolimbic-dopaminergic Clock biology, template for psychiatric risk-gene causality without flox breeding; Karlsgodt UCLA transdiagnostic early-psychosis HCP-EP (N=124) maps PET receptor/transporter distributions to RSFC symptom-dimension signatures — negative-symptom + cognition signatures correlate with NE-transporter + VAChT, direct argument for noradrenergic + cholinergic prioritization in refractory symptom domains; Chakravarty McGill phenotype-matched chronic variable stress in mice shows opposite-direction neuroanatomical changes between sexes in NAc + hippocampus with cortical-hub-dominant females + subcortical-hub-dominant males — cleanest case for why mixed-sex preclinical antidepressant work is noisy; Barbier Linköping PRDM2 reduced in PFC of human AUD, dmPFC viral knockdown increases stress-induced alcohol reinstatement in both sexes, projection-specific dmPFC-NAc knockdown sufficient — circuit anchor for stress-relapse interventions bypassing epigenetic upstream; Unterwald Temple chronic cocaine + SARS-CoV-2 spike rat model shows synergistic weight loss + pain sensitization + hippocampal IL-6/IL-10 shift — specific prediction for cocaine-use-disorder + Long-COVID comorbidity not currently stratified; Harding Toronto RaPID macrocyclic peptide tools against HAP40 — endogenous-engagement gap filled for HTT-HAP40 biology in Huntington's; Lewis EPFL CLEM of Lewy pathology in PD vs DLB — fibrillar ultrastructure shared, region-specific mitochondrial signatures distinguish diseases; Kucinski Michigan dual-disruption rat optogenetic confirmation that PD falls are dual-system cholinergic + dopaminergic failure not pure dopamine loss; Fan Fan MCG chronic TPPU sEH-inhibition rescues cognition in diabetic-related dementia via cerebrovascular + synaptic + neuroinflammatory multi-arm mechanism — sEH GWAS makes diabetes-ADRD natural enrichment for early trial; Morphological Shortcuts by LLMs in Pharmacology — affix-only fictitious drugs elicit class-level pharmacological responses, attribution framework on 653 real drugs shows models reason from suffix without disclosing reliance — clinical-safety implication for CDS LLMs; closes light on ten headache specialists vs RAG-based agentic AI on clinical literature summarization — expert summaries preferred but indistinguishable by some specialists, win condition is matching expert-valued decision-shaping features beyond the standard rubric. Receptor & Reason for June 4, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with a paired set of preprints arguing that the ketamine metabolite (2R,6R)-hydroxynorketamine produces rapid antidepressant effects by selectively engaging astrocytic mu-delta opioid receptor heterodimers in the hippocampus — pharmacological and genetic disruption of either MOR or DOR abolishes the behavioural rescue, PAINT-MINFLUX super-resolution nanoscopy directly visualizes increased heterodimerization after a single dose, molecular dynamics simulations pinpoint selective binding at mu-receptor Asp147/Tyr148 (mutating these residues abolishes both heterodimer formation and antidepressant-like effects), and the downstream signaling is unusually Gs-coupled rather than Gi (raising cAMP and CREB phosphorylation in astrocytes and shifting calcium dynamics) — new drug-discovery target distinct from the NMDAR-centric rapid-acting antidepressant pipeline, with the next experiment being a small-molecule screen for selective mu-delta heterodimer assembly enhancers. Then high-resolution hippocampal-amygdala fMRI during emotional mnemonic discrimination in N=117 humans (single-action SSRI vs multi-action SNRI/NDRI/polypharmacy vs unmedicated controls) finds that non-responders to single-action vs multi-action antidepressants show opposite balances of dentate gyrus/CA3 vs CA1 activity, while responders regardless of class show stronger DG/CA3-BLA coactivation during negative emotional memory — the circuit signature of treatment failure is mechanism-class specific and persists beyond symptom severity, candidate mechanistically-specific class-selection biomarker for matching antidepressants to patients. Then a spatiotemporal-dynamics analysis applying complex principal component analysis to resting-state fMRI across NREM sleep, propofol anesthesia, and five psychedelic-or-dissociative states (LSD, DMT, psilocybin, N2O, ketamine) identifies four dominant brain-wave propagation motifs — the global synchronized wave is most state-dependent, slowing/fragmenting/spatially concentrating in diminished states and accelerating/distributing/homogenizing in psychedelic states along the same axis in opposite directions — receptor-profile-agnostic target-engagement biomarker framework for psychedelic and dissociative therapeutics. Then multi-omic profiling of dorsal forebrain organoids from 17 idiopathic schizophrenia individuals and 17 matched controls finds transcriptomic differences modest (strongest cell-type-specific changes in Cajal-Retzius neurons) while proteomic and especially PTM-level analyses reveal widespread disruption in neuronal migration, neurite development, synaptic function, protein kinase signalling, ECM, and lipid metabolism — phosphorylation as the earliest disease-associated change, sex-specific dysregulation of protein state as a major molecular feature, most alterations occurring independently of transcript or protein abundance — argues that transcriptomic-only studies of schizophrenia genetics are systematically missing most of the signal. Then cell-type-specific longitudinal fiber photometry of prelimbic intratelencephalic (IT) neurons in rat intermittent-access fentanyl self-administration with sex-stratified vulnerability phenotypes finds female rats more prone to high-risk classification (robust cue responsiveness and seeking during drug unavailability), with IT neurons dynamically encoding fentanyl-associated stimuli across acquisition/escalation/extinction/reinstatement — genetically-defined PFC projection population now mapped well enough for cell-type-specific interventions in opioid use disorder. Then melanin-concentrating-hormone (MCH) neurons in lateral hypothalamus identified as a tractable sleep target in tauopathy — longitudinal EEG/EMG in PS19 mice shows progressive sleep architecture impairment paralleling tau accumulation, fiber photometry reveals selective functional impairment of MCH neurons (not neighbouring hypocretin population) characterized by reduced REM-sleep activity, cell-autonomous mutant-tau expression in MCH neurons alone recapitulates the phenotype, and optogenetic/chemogenetic activation restores sleep in aged mice (neurons rescuable despite damage). Then the PFKFB3-fructose-2,6-bisphosphate metabolic-proteostatic axis established as a central driver of AD pathogenesis — pathological tau aggregates sequester PFKFB3 depleting F2,6BP, which normally activates PNKP for DNA strand-break repair, upregulates PP2A to limit tau phosphorylation, stabilizes free PFKFB3, and directly inhibits tau aggregation, with exogenous F2,6BP supplementation rescuing pathology across iNeurons, primary neurons, organotypic slices, and Drosophila — metabolite with multiple convergent mechanisms as a drug-discovery lead. Then cryo-EM + biochemical reconstitution of LRRK2 resolves the ROC GTPase as master switch between autoinhibited and active states (GTP activates, GDP enforces autoinhibition), with G2019S and R1441 mutations activating through structurally distinct mechanisms — stabilizing the GDP-bound conformation as alternative to active-site kinase inhibition, GTP-bound state promotion potentially beneficial in lung where LRRK2 may be protective, opens tissue-selective conformational modulator strategy for programs hit by peripheral toxicity. Then NULISA multiplex proteomic analysis of 852 CSF samples from the Parkinson's Disease Biomarker Program identifies 54 of 124 markers altered in PDD/DLB vs controls and 15 proteins tracking cognitive progression, with stratification by synuclein (seed amplification) vs amyloid (NULISA p-Tau217) pathology revealing that select synaptic markers move in opposite directions across co-pathology cases — replicated in 917 independent ADNI cases, enables trial enrichment by pathological substrate rather than treating LBD as homogeneous. Closes on Autopipe — an MCP-compatible platform that converts natural-language LLM-driven bioinformatics conversations into containerized re-executable pipelines runnable on remote HPC, with desktop app + embedded MCP server + online pipeline/plugin registry + web result viewer + plugin CLI — solves the reproducibility and sharing problem (not the agent quality problem) for prompt-driven bioinformatics analyses, making agent benchmark scores operationally meaningful for end users. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-04-receptor-and-reason Thu, 04 Jun 2026 12:00:00 +0000 1136 Daily roundup for June 4, 2026: paired preprints argue ketamine metabolite (2R,6R)-hydroxynorketamine works through astrocytic mu-delta opioid receptor heterodimers in hippocampus (PAINT-MINFLUX nanoscopy + MD sims pinpoint Asp147/Tyr148, mutation abolishes heterodimer formation and antidepressant effect, Gs-coupled signaling raises cAMP/p-CREB) — new target distinct from NMDAR-centric rapid-acting antidepressant pipeline; hippocampal-amygdala fMRI (N=117) finds DG/CA3 vs CA1 balance differentiates single-action vs multi-action antidepressant non-responders, DG/CA3-BLA coactivation marks responders — mechanism-class-specific selection biomarker; spatiotemporal brain-wave propagation analysis across NREM sleep + propofol + LSD/DMT/psilocybin/N2O/ketamine reveals global synchronized wave slows/fragments in diminished states and accelerates/distributes in psychedelic states along same axis in opposite directions — receptor-profile-agnostic biomarker framework; multi-omic schizophrenia organoid profiling (17 vs 17) finds transcriptomic differences modest but proteomic/PTM disruption widespread with phosphorylation as earliest disease-associated change — transcriptomic-only studies miss most of the signal; cell-type-specific fiber photometry of prelimbic IT neurons in rat fentanyl self-administration with sex-stratified vulnerability — dynamic encoding of fentanyl-associated stimuli, female rats more prone to high-risk classification; melanin-concentrating-hormone neurons in lateral hypothalamus identified as tractable sleep target in tauopathy (PS19 mice) — cell-autonomous mutant tau in MCH neurons alone recapitulates sleep phenotype, optogenetic/chemogenetic activation restores sleep despite damage; PFKFB3-F2,6BP metabolic-proteostatic axis as AD driver — F2,6BP activates PNKP DNA repair, upregulates PP2A, stabilizes PFKFB3, inhibits tau aggregation, supplementation rescues across models; cryo-EM resolves LRRK2 ROC GTPase as autoinhibited/active master switch with G2019S vs R1441 activating through distinct mechanisms — stabilizing GDP-bound conformation as alternative inhibition strategy; NULISA multiplex CSF proteomics across 852 PDBP samples identifies 54 of 124 markers altered in PDD/DLB and 15 cognitive-progression markers, stratification by synuclein vs amyloid pathology enables trial enrichment; closing on Autopipe MCP-compatible platform converting NL LLM-driven bioinformatics conversations into containerized reproducible pipelines on remote HPC — solves reproducibility problem (not agent quality) for prompt-driven bioinformatics. Receptor & Reason for June 3, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with the Newell group at Wollongong delivering the first transdiagnostic bulk RNA-seq of human habenula from Netherlands Brain Bank tissue (6 control, 6 MDD, 6 BPD) — single-diagnosis comparisons modest (~60 DEGs each) but the affective-illness pooled contrast yields 378 differentially expressed genes enriched for potassium-channel activity, calcium homeostasis, and Wnt signalling — robust shared molecular signature consistent with the preclinical hyperexcitability story, implying potassium-channel modulators and Wnt-pathway tools as a logical second-pass after the ketamine signal. Then the Cocchi lab at QIMRB fitting a hidden-Markov-style dynamic network model to resting-state EEG across 70 patients undergoing left dlPFC TMS for depression — higher baseline severity tracks reduced anterior default mode state occupancy and fewer transitions plus elevated posterior DMN activity, and per-session decreases in anterior DMN engagement during treatment predict second-half symptom improvement — dynamic transition-rate biomarker suitable for closed-loop TMS titration. Then the Moayedi group at Toronto with Atlas at NIH on the endocannabinoid system as the non-opioid component of placebo analgesia in N=48 — individual placebo pain reduction tracks the FAAH-substrate composite (anandamide, palmitoylethanolamide, oleoylethanolamide) but not 2-AG or beta-endorphin alone, with beta-endorphin moderating the relationship (FAAH substrates predict pain reduction only when beta-endorphin is low) — endocannabinoid and opioid systems substitute for each other state-dependently, opening FAAH inhibitors as candidate placebo-amplifiers in chronic pain. Then the Wang lab at Augusta University with single-nucleus multiome sequencing of hippocampus from mice exposed to chronic unpredictable mild stress — the oligodendrocyte lineage shows the heaviest vulnerability, with CUMS preferentially disrupting immature OPC/intermediate states and impairing OPC migration, OPC-to-ODC progression, and myelination, integrated multiomics converging on SOXD transcription factors (particularly SOX5 and SOX6) as key regulators, SOX6 ChIP-seq confirming direct binding at OPC morphogenesis/migration/glutamatergic-signalling regulatory elements, and enhancer-driven SOX6 restoration in OPCs rescuing stress-induced migration and myelination defects — multiple-sclerosis remyelination pharmacology infrastructure becomes potentially relevant for depression. Then the Delevich lab at Washington State on orchiectomy in male mice either before puberty or in adulthood — both timings increase sensitivity to effort costs without altering food intake (energy-efficient responding), but adult ORX raises NAc DA content and reduces turnover with broad SPN excitability changes while prepubertal ORX selectively reduces D1R+ SPN excitability in NAc core — same behavioural phenotype emerges from developmental-timing-dependent circuit adaptations, argues for age-of-onset stratification in clinical trials of dopamine-targeting agents for androgen-deficiency motivation deficits. Then the Olsen lab at Pittsburgh + Broad metabolomics on a Drosophila alpha-synuclein glial kinome screen converging on adenosine metabolism (AK1, AK6, ADK1, ADK2, awd) — glial knockdown raises brain adenosine, rescues locomotion, reduces alpha-synuclein oligomers, and requires neuronal adenosine receptor signalling, with human astrocyte AK1 knockdown decreasing reactivity markers and increasing extracellular adenosine — argues for adenosine-kinase inhibition (raising glial adenosine release) as disease-modifying therapeutic logic opposite in sign to the approved A2A antagonist istradefylline. Then the Mailman lab at Penn State showing levodopa drives substantia nigra iron accumulation independent of dopamine neurons — 15 days of allometric mid-stage PD doses raise SN iron on both lesioned and unlesioned sides of 6-OHDA rats, and 4-month daily dosing in unlesioned rats shows progressive iron increase via R2* MRI — reframes a substantial fraction of the SN iron literature from etiology to iatrogenesis, reviving iron chelation as adjunct therapy and iron MRI as a dose-finding safety endpoint. Then the Parish lab at the Florey using single-nuclei RNA-seq of hPSC-derived ventral midbrain DA grafts to identify Semaphorin, Netrin, and Wnt axon-guidance pathways preferentially active in graft DA neurons — host striatal overexpression of SEMA3A, NTN1, or WNT5A selectively increases A9 DA specification and striatal innervation without GDNF's off-target extrastriatal sprouting, NTN1 in Parkinsonian rats drives motor recovery with cell-type-specific plasticity confirmed by snRNA-seq — more biology-aware trophic strategy than GDNF for stem-cell-derived DA cell therapy. Then the Mujica-Parodi lab at Stony Brook with a pseudo-malate-aspartate shuttle (PMAS) computational model of brain glutamate-GABA cycling that predicts MRS response to administered ketones — metabolic control analysis identifies partial displacement of glucose metabolism through PMAS as the dominant lever with enzyme-specific knobs selectively modulating each neurotransmitter — QSP-style patient-specific ketone dosing framework that plugs into individual MRS measurements for epilepsy, mood, and brain-aging interventions. Closes on PromptBio-Bench (PromptBio Inc) — a structured-file-comparison benchmark of LLM-based bioinformatics agents across 244 expert-curated tasks at three difficulty levels, finding comparable performance between Biomni and ToolsGenie with the expected sharp drop-off at the hardest tier, structured output scoring sidestepping LLM-as-judge blind spots, and versioned reference-file infrastructure as the shared evaluation dependency the Biomni-versus-competitors landscape needed. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-03-receptor-and-reason Wed, 03 Jun 2026 12:00:00 +0000 890 Daily roundup for June 3, 2026: Newell (Wollongong) first transdiagnostic human habenula bulk RNA-seq (6 control + 6 MDD + 6 BPD from Netherlands Brain Bank) finds 378 affective-illness DEGs enriched for potassium channels, calcium homeostasis, Wnt — implies potassium-channel and Wnt tools as logical second-pass after ketamine; Cocchi (QIMRB) hidden-Markov EEG model across 70 TMS depression patients finds anterior DMN state occupancy as on-treatment dynamic biomarker for closed-loop titration; Moayedi (Toronto) + Atlas (NIH) endocannabinoid system as non-opioid component of placebo analgesia (N=48) — FAAH-substrate composite predicts placebo pain reduction only when beta-endorphin is low, opening FAAH inhibitors as placebo-amplifiers; Wang (Augusta) snRNA-seq + ATAC of hippocampus after chronic unpredictable mild stress converges on SOXD/SOX6 as key regulator of OPC dysfunction, enhancer-driven SOX6 restoration rescues migration/myelination defects — MS remyelination pharmacology infrastructure becomes relevant for depression; Delevich (WSU) pre/post-pubertal orchiectomy in male mice increases effort-cost sensitivity through developmental-timing-dependent mesoaccumbal adaptations — argues for age-of-onset stratification in clinical trials; Olsen (Pittsburgh) Drosophila alpha-syn glial kinome screen converges on adenosine metabolism (AK1, AK6, ADK1, ADK2) — adenosine-kinase inhibition as disease-modifying logic opposite in sign to approved A2A antagonist istradefylline; Mailman (Penn State) levodopa drives SN iron accumulation independent of DA neurons in 6-OHDA and unlesioned rats — reframes SN iron from etiology to iatrogenesis, revives iron chelation; Parish (Florey) snRNA-seq identifies Netrin-1 for selective A9 DA graft plasticity without GDNF's off-target sprouting — more biology-aware trophic strategy for stem-cell DA cell therapy; Mujica-Parodi (Stony Brook) PMAS computational model predicts MRS glutamate/GABA response to ketones — QSP-style patient-specific ketone dosing for epilepsy/mood/brain-aging; closing on PromptBio-Bench (PromptBio Inc) structured-file-comparison benchmark of LLM bioinformatics agents (244 tasks) finds comparable Biomni and ToolsGenie performance, sidesteps LLM-as-judge blind spots, versioned reference files as the shared evaluation dependency the bioinformatics-agent landscape needed. Receptor & Reason for June 2, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with the Isom lab at the University of Miami arguing all three opioid receptors — mu, delta, kappa — behave as proton-gated coincidence detectors: agonist efficacy falls steadily as extracellular pH drops from physiological levels toward inflamed-tissue values in the DCyFIR humanized yeast platform, and structure-based modeling pinpoints conserved aspartate D149 (D3.32) as the pH sensor whose protonation disrupts the salt bridge anchoring opioid agonists — clean mechanistic explanation for reduced opioid analgesia in inflamed tissue and an imperative to re-measure tool-compound efficacy across physiologically relevant pH. Then the Hohmann lab at Indiana University on a substantially new mechanism for acetaminophen — analgesia in mouse post-surgical and inflammatory pain models is blocked by DAGL and MAGL inhibitors and by global (not peripherally restricted) CB1 antagonists, with corticosterone rising >2-fold, prostaglandins dropping >5-fold, and ~40% of detected signaling lipids suppressed across brain and inflamed paw — the textbook weak-COX-inhibitor story is incomplete, opening a different design space for next-generation non-opioid analgesics and raising co-administration questions with cannabinoid drugs. Then Li (Sichuan) HPD1 — an esterase-responsive hydrogen persulfide (H2S2) donor — reduces mechanical and cold allodynia in mouse chronic constriction injury and paclitaxel neuropathic pain models at 14 mg/kg with negligible toxicity at 28 mg/kg, potentiating dorsal root ganglion A-type potassium currents via a novel LIMK1→cofilin inactivation→F-actin stabilization→Filamin A/Kv4.2 coupling axis (LIMK1 inhibition with BMS-5 dampens both the IA restoration and the analgesia) — cytoskeletal-scaffold lane for restoring A-type currents distinct from direct channel agonism. Then the Psychiatric Times May 2026 pipeline review — first-in-class SIRT6 inhibitor (the first explicitly epigenetic-mechanism antidepressant) with a sex-specific female-only efficacy signal presented at ASCP, Synendos SYT-510 selective endocannabinoid reuptake inhibitor entering phase 2 in GAD (pairs mechanistically with the acetaminophen endocannabinoid result), and a sponsor-funded network meta-analysis ranking lumateperone as the top adjunct atypical antipsychotic for MDD across four efficacy metrics with relatively low weight gain. Then the Ramanathan lab at UC San Diego isolating the inter-train interval as the key parameter governing whether intermittent theta-burst stimulation drives plasticity — short 4-10s ITIs strongly recruit GABAergic/parvalbumin interneurons that clamp cortical excitability and spine plasticity, while extended 20s ITI (eTBS) releases the GABA brake, drives much larger spine/synaptic plasticity, and produces rapid durable antidepressant-like effects after a single session, with optogenetic GABA activation during eTBS abolishing both effects (clean causal link) — inter-train interval as underused lever for the plateaued iTBS response rates clinic sees today. Then Uygun (Warwick) on thalamocortical GABA-A receptor function as a circuit node for PTSD — localized CRISPR-Cas9 knock-down of GABA-A receptors specifically in thalamocortical neurons shortens and abnormally shapes sleep spindles (losing characteristic waxing-and-waning envelope) and blocks habituation in contextual fear conditioning, recapitulating both the PTSD sleep-spindle phenotype and the failure of trauma-memory extinction — nudges interest toward thalamic-targeted subtype-selective GABA-A PAMs. Then Flagel (Michigan) dual-vector chemogenetic inhibition of the lateral hypothalamus→paraventricular thalamus pathway selectively attenuates goal-tracking without affecting sign-tracking in Pavlovian conditioned approach, with the effect driven primarily by intermediate-responder rats (the most behaviorally flexible subgroup) — argues for individual-difference subgrouping when designing circuit-targeted addiction therapies. Then Ohba (Jichi Medical) showing transient inflammatory cytokine+complement stimulation locks human astrocytes into a persistent C3-high post-inflammatory state after stimulus washout — most inflammatory response genes (CXCL10, NFKBIA, TNFAIP3, RELB) return toward baseline but C3 remains elevated, and pharmacological JAK inhibition returns C3 to near-baseline — the approved JAK-inhibitor class is a plausible repurposing handle for persistent astrocyte reactivity that bridges synaptic pruning, neurodegeneration, and psychiatric inflammation hypotheses. Then Ghosh (Maulana Azad) Curli Carrier Burden — a mathematically explicit, evidence-weighted microbial trait index quantifying curated curli-producing bacterial taxa abundance, evaluated across five major PD gut metagenomic cohorts (Wallen 2022, Integrated-US, Mao Central China, Romano non-Wallen, DuruIC 2024) — transparent trait-level operationalization of bacterial amyloid exposure as a falsifiable arm of the gut-to-brain alpha-synuclein hypothesis, shaped for comparative meta-analysis. Closes on DrugClaw + DrugAudit — multi-agent retrieval-augmented system with reflection-driven state-machine workflow over a drug/pharmacovigilance skill registry hits >91% primary-source rate and ~89% faithfulness (+6-10pp over competitors), companion benchmark scores primary-source match/snippet overlap/citation faithfulness with dual-judge LLM-as-judge kappa 0.88 — the authority-aware framing is the durable contribution, distinguishing answer correctness from correct-source-grounding for high-stakes pharmacology QA. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-06-02-receptor-and-reason Tue, 02 Jun 2026 12:00:00 +0000 779 Daily roundup for June 2, 2026: Isom (Miami) opioid receptors as proton-gated coincidence detectors with conserved D149 (D3.32) as intrinsic pH sensor — agonist efficacy at mu/delta/kappa falls with decreasing pH in DCyFIR humanized yeast, mechanistic explanation for reduced opioid analgesia in inflamed tissue; Hohmann (Indiana) acetaminophen analgesia requires DAGL + MAGL + central (not peripheral) CB1 with >2-fold corticosterone rise, >5-fold prostaglandin drop, ~40% signaling-lipid suppression — endocannabinoid-tone mechanism rewrites textbook weak-COX-inhibitor story; Li (Sichuan) HPD1 hydrogen persulfide (H2S2) donor reduces neuropathic pain via novel LIMK1→cofilin→F-actin→Filamin A/Kv4.2 axis restoring DRG A-type K+ currents — cytoskeletal-scaffold lane for IA restoration; Psychiatric Times May pipeline roundup — first-in-class SIRT6 inhibitor with female-only efficacy signal at ASCP, Synendos SYT-510 endocannabinoid reuptake inhibitor enters phase 2 in GAD, lumateperone tops network meta-analysis for adjunct MDD; Ramanathan (UCSD) extended-interval theta burst stimulation (eTBS, 20s ITI) releases GABA/parvalbumin brake — larger plasticity and rapid durable antidepressant effects after single session, optogenetic GABA activation abolishes effect; Uygun (Warwick) thalamocortical GABA-A CRISPR knock-down shortens/distorts sleep spindles and blocks contextual-fear habituation — recapitulates PTSD spindle phenotype + extinction failure, argues for thalamic-targeted subtype-selective GABA-A PAMs; Flagel (Michigan) chemogenetic inhibition of LH→PVT pathway selectively attenuates goal-tracking driven by intermediate-responder rats (most flexible subgroup) — individual-difference subgrouping for circuit-targeted addiction therapy; Ohba (Jichi Med) transient inflammatory stimulus locks human astrocytes into persistent C3-high state after washout, JAK inhibition returns C3 to baseline — approved JAK-inhibitor class as repurposing handle; Ghosh (Maulana Azad) Curli Carrier Burden index quantifies bacterial amyloid exposure across 5 PD gut metagenomic cohorts — transparent trait-level operationalization of gut-to-brain alpha-synuclein hypothesis; closing on DrugClaw + DrugAudit authority-aware drug-information agent — multi-agent RAG with reflection-driven state-machine workflow hits >91% primary-source rate, ~89% faithfulness (+6-10pp over competitors), authority-aware benchmark framing is the durable contribution. Receptor & Reason for May 31, 2026 — daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with the Breen lab at Mount Sinai dropping a long-read isoform atlas of purified human prefrontal and orbitofrontal cortical cell types — about 220,000 full-length isoforms cataloged with 35-56% previously unannotated, glia (oligodendrocytes and microglia) emerging as the most isoform-diverse populations and inverting the neuron-centric view of cortical complexity, 59-62% of differentially regulated transcripts absent from current annotations, and pathogenic variants enriched more than two-fold at novel splice boundaries within POGZ, TARDBP, PLP1 — establishes isoform selection as a primary axis of cortical identity and exposes a layer of pathogenic variation invisible to canonical gene annotations, the new baseline for transcriptomic work on brain disorders. Then Suryadevara (Stanford) plasma proteomic profiling on 408 deeply phenotyped AD/LB/PD participants reveals sex-divergent bone-senescence signatures — inflammatory and bone-related pathways differ by sex within disease, shared mitochondrial/metabolic dysfunction sits underneath, candidate proteins correlate with pTau181 — systemic comorbidities are part of the disease phenotype and segregate by sex, arguing for sex-stratified rather than unified biomarker panels. Then Picotti (ETH Zurich) structural-proteomics-fingerprint approach quantifies alpha-synuclein conformational mixture in situ — measures structure-specific proteolytic fingerprints from in vitro conformations and back-calculates conformational composition of complex samples, benchmarked tracking disordered monomer falling, beta-sheet oligomers rising, delayed fibril pool in the same sample, then taken into yeast and complex biological samples — closes the translational loop for PD therapeutics targeting specific conformational transitions; technique generalizes to Tau and TDP-43. Then Conant (Georgetown) 4-methylumbelliferone (4-MU, already approved in Europe for biliary spasm) inhibits hyaluronic acid synthesis and reduces perineuronal nets — APP/PS1 mice from 3 months for 70 days or 52 weeks show reduced soluble A-beta42/40 ratio with the long arm reducing insoluble plaque, PNN intensity, and spatial memory deficits — ECM-targeted repurposing candidate for AD with known human safety profile. Then Vlachos (Freiburg) on rTMS microglia-dependent synaptic remodeling — 10 sessions of iTBS900 over 2 weeks in mouse organotypic cultures reduces AMPA mEPSC frequency and spine density in pyramidal neurons; microglial depletion inverts the response to increased synaptic strength and spine density — microglia as required mediators of homeostatic structural remodeling from multi-session rTMS, with neuroinflammatory state at baseline likely predicting patient response. Then Muntoni (UCL Queen Square) AAV micro-dystrophin brain-targeted gene therapy in mdx52 Duchenne model — IV and ICV delivery in neonatal mdx52 males both reduce anxiety-like phenotypes, IV superior across broader behavioral battery rescuing OCD-like behavior, motor coordination, and grip strength — single systemic gene therapy addresses both somatic and neuropsychiatric comorbidities (ASD/ADHD/OCD/ID affect ~40% of DMD patients via loss of brain dystrophin isoforms Dp427 and Dp140); generalizable logic for monogenic CNS-relevant disorders. Then Zamponi (Calgary) patch-clamp biophysics of CACNA1E L228P developmental epileptic encephalopathy mutation — Cav2.3 R-type channel activation shifts toward hyperpolarized potentials without changing current density, inactivation kinetics, or recovery — gain-of-function lowering excitability threshold, mutation-by-mutation biophysics needed for precision-medicine stratification toward channel blockers vs openers in refractory pediatric epilepsy. Then Cakir (Merck) fully automated robotic high-density extracellular recording from interior of intact human cortical organoids — LSTM classifier on physiologically grounded spike features distinguishes healthy iPSC organoids from familial AD APP-mutant organoids at 100% accuracy (small-cohort caveat); critically, AD organoids treated with a drug candidate that reliably lowers amyloid-beta are still called AD-like electrophysiologically — functional electrophysiological phenotype dissociates from molecular biomarker, uncomfortable implication for amyloid lowering as surrogate endpoint. Then Rose (UC Irvine) CARIBOU multi-agent system for autonomous bioinformatics analysis on institutional HPC — researcher-editable blueprints encoding analytical roles, workflow guidance, and domain-specific reasoning grounded in persistent executable Singularity/Apptainer environments; shared evolving analytical state enables iterative execute-observe-correct through QC, batch integration, clustering, cell-type annotation; benchmarked on unit tasks, metadata reconstruction, and full single-cell RNA-seq analyses with Allen Brain Atlas hippocampus and Tabula Sapiens — HPC-native design closes the gap that has stopped prior agentic-bioinformatics systems from running where the data actually lives. Closing on Jerbi (Montreal) source-resolved MEG of LSD vs placebo (with/without music) — distinguishes peak-frequency shifts from power attenuation, finds upward alpha/beta peak shifts alongside genuine oscillatory attenuation in partly dissociable cortical patterns, flattened aperiodic 1/f slope, increased signal complexity/fractality preferentially in sensory/language/emotion/imagery networks sparing motor cortex; music did not robustly amplify the neural signatures — methodological imperative to parameterize psychedelic EEG/MEG distinguishing peak shift from power attenuation, plus a small set-and-setting result worth holding alongside the clinical psychedelic literature; threads off yesterday's LSD-aging-signature-reversal piece. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-31-receptor-and-reason Sun, 31 May 2026 12:00:00 +0000 740 Daily roundup for May 31, 2026: Breen (Mount Sinai) long-read isoform atlas of purified human cortical cell types — ~220K full-length isoforms with 35-56% previously unannotated, glia (oligodendrocytes, microglia) most isoform-diverse populations inverting the neuron-centric view, 59-62% of differentially regulated transcripts absent from current annotations, pathogenic variants enriched >2-fold at novel splice boundaries in POGZ/TARDBP/PLP1 — isoform selection as primary axis of cortical identity exposing pathogenic variation invisible to canonical gene annotations; Suryadevara (Stanford) plasma proteomics on 408 AD/LB/PD participants reveals sex-divergent bone-senescence signatures with inflammatory/bone pathways differing by sex within disease, shared mitochondrial dysfunction underneath, candidates correlated with pTau181 — sex-stratified biomarker panels imperative; Picotti (ETH Zurich) structural-proteomics-fingerprint quantifies alpha-synuclein conformational mixture in situ tracking monomer/oligomer/fibril proportions in same sample — closes translational loop for PD conformational-state-targeted therapeutics; Conant (Georgetown) 4-methylumbelliferone (already approved in Europe for biliary spasm) attenuates amyloid pathology and spatial memory deficits via perineuronal-net reduction in APP/PS1 — ECM-targeted AD repurposing with known human safety profile; Vlachos (Freiburg) rTMS-induced homeostatic synaptic remodeling is microglia-dependent — microglial depletion inverts the iTBS900 response, microglia as required mediators of multi-session rTMS plasticity with neuroinflammatory state predicting response; Muntoni (UCL) AAV micro-dystrophin gene therapy in mdx52 Duchenne model rescues anxiety/OCD/motor phenotypes with IV superior to ICV — single systemic gene therapy addresses both somatic and neuropsychiatric comorbidities; Zamponi (Calgary) Cav2.3 L228P DEE mutation shifts activation toward hyperpolarized potentials (gain-of-function lowering excitability threshold) — mutation-by-mutation biophysics for precision-medicine stratification in refractory pediatric epilepsy; Cakir (Merck) robotic high-density recording from intact human cortical organoids — LSTM classifier distinguishes healthy vs familial-AD organoids, amyloid-lowering drug still called AD-like electrophysiologically (functional phenotype dissociates from molecular biomarker); Rose (UCI) CARIBOU multi-agent autonomous bioinformatics on institutional HPC with Singularity/Apptainer persistence and execute-observe-correct iteration; closing Jerbi (Montreal) source-resolved MEG LSD vs placebo distinguishes peak-frequency shifts from power attenuation, flattened aperiodic 1/f slope, increased complexity/fractality in sensory/language/emotion/imagery networks sparing motor cortex, music does not robustly amplify — methodological imperative for psychedelic oscillatory analysis. Receptor & Reason for May 30, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Opens with the Iorio group at Human Technopole applying the transcriptional-signature-reversal principle (LINCS/CMap lineage) to brain aging — chronic LSD-induced gene expression in rodent brain is strongly anti-correlated with aging- and dementia-associated transcriptional programs in human prefrontal cortex, more specific than other pharmacological perturbations tested, reproducible across datasets and species, and LSD attenuates amyloid-beta-induced structural/molecular damage in primary cortical neurons — a mechanistic anchor for the broader case for 5-HT2A psychedelics in age-associated cognitive decline. Then Carpenter (Broad) + McLean on a 168-patient fibroblast imaging study — mitochondria sit farther from the cell border in psychosis (bipolar, schizophrenia, schizoaffective subsets) with the opposite trend in major depression without psychosis; the phenotype reduces to a single metric enabling existing-database compound and gene ranking, with internal validation from known psychiatric hits — transdiagnostic imaging-based patient phenotype as entry point for novel-mechanism psychiatric drug discovery. Then Daskalakis (McLean) — PsychENCODE 1,022-sample postmortem cortex circRNA quantification identifies 23 SCZ and 3 BIP differentially expressed circRNAs at FDR<0.05; circRNA GReX models built in controls and applied to SCZ and BIP GWAS yield 22 and 4 trait-associated loci respectively; UK Biobank imaging associations anti-correlated with SCZ/BIP but positively correlated with Alzheimer's; circKLHL24 isoform standout — circRNAs as a non-redundant regulatory layer linking genetic risk to synaptic biology and brain structure. Then Desrochers (Brown) on OCD circuit configuration — task-based functional and effective connectivity in the abstract-sequencing circuit (RLPFC, ACC/DLPFC, SMA, MTG, TOJ) shows that OCD and healthy participants share underlying architecture but differ in functional coordination, with OCD showing increased direct RLPFC-to-MTG coupling that bypasses the usual relay; ACC/DLPFC and MTG act as hubs with task input entering via TOJ and propagating forward — circuit-targeted intervention design implication for repetitive transcranial magnetic stimulation and deep-brain stimulation programs in OCD. Then Pritzker (Stanford) on opioid withdrawal — an evolutionarily conserved Tshz1+ MOR+ NAc neuron population gates withdrawal aversion specifically; MOR deletion in Tshz1 cells abolishes the withdrawal-induced dopamine drop and affective aversion but spares physical symptoms; mGluR8 is preferentially expressed in Tshz1 cells, and its activation reduces withdrawal aversion — a non-opioid group-III mGluR substrate for the affective arm of opioid withdrawal as adjunct to MAT. Then Schwendt (Florida) — mGlu2 PAM LY-487379 fully reverses persistent object-in-place recognition deficits after extended-access methamphetamine self-administration plus 30-day abstinence; surface mGlu2/3 expression rises in prelimbic and perirhinal cortices (compensatory trafficking insufficient without pharmacological boost) — group-II mGluR pharmacology for stimulant-induced cognitive dysfunction. Then Wu (Georgetown) — TBOA-modeled EAAT inhibition (early-AD glutamatergic dysregulation) reduces hippocampal sharp-wave ripple amplitude and disrupts ripple-associated population calcium synchrony in mouse slices; KCNQ-channel openers ML213 and ICA-27243 partially restore ripple amplitude while AMPA/NMDA blockade, HCN-Ih blockade, BK or GIRK activation do not — KCNQ openers occupy a unique position for stabilizing amyloid-driven hippocampal network dysfunction. Then CuraGen AI APOE→ACAT1 pipeline — scaffold-constrained generative chemistry + multi-objective ADMET optimization across 30+ criteria yields ~7,300 CNS-optimized candidates from ~6M generated molecules with three leads at BBB probability >0.93 and stable ACAT1 binding poses, anchored on the mechanistic chain from APOE4 cholesterol-transport impairment to microglial cholesteryl-ester accumulation and lost amyloid clearance plus blunted tau autophagy — in-silico-stage target-validation + design exercise. Then Narayan (NIH/NIDDK) on isogenic iPSC-derived astrocytes — the APOE4 variant impairs clathrin-mediated membrane curvature, reduces clathrin-mediated endocytosis, and alters plasma-membrane lipid saturation and tension (flat clathrin structures, reduced coated-pit maturation, reduced early endosomes); INPP5D overexpression restores early endocytosis through clathrin-curvature promotion distinct from membrane-tension regulation, also reduces lipid droplets and inflammatory signaling — APOE4 reframes as a membrane-mechanics defect upstream of trafficking, with astrocyte endocytosis as a node multiple AD risk genes converge on. Closing on Karami (Inria/Lorraine) AlloDyn — integrates static pocket descriptors with dynamic features from all-atom MD and AlphaFlow-generated conformational ensembles, plus a methodological critique of standard allosteric-pocket benchmarks that apply fpocket to holo structures without removing the bound allosteric modulator (data leakage inflating reported numbers); AllyDyn achieves best precision/recall balance and highest F1 and MCC on properly preprocessed benchmark, AlphaFlow ensembles match MD-derived features at a fraction of compute — conformational dynamics belong inside the predictor, not bolted on after; pairs with yesterday's attention-enabled PocketMiner work in the cryptic-pocket-plus-coupling-prediction space. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-30-receptor-and-reason Sat, 30 May 2026 12:00:00 +0000 1170 Daily roundup for May 30, 2026: Iorio (Human Technopole) applies transcriptional-signature-reversal (LINCS/CMap lineage) to brain aging — chronic LSD-induced expression in rodent brain anti-correlated with human prefrontal-cortex aging/dementia signatures across datasets and species, more specific than other perturbations, plus amyloid-beta damage rescue in primary cortical neurons; Carpenter (Broad) + McLean fibroblast imaging shows mitochondria farther from cell border in psychosis (BIP/SCZ/SCZA subsets), enabling existing-database compound/gene ranking — transdiagnostic imaging phenotype as entry to novel-mechanism psychiatric drug discovery; Daskalakis (McLean) PsychENCODE circRNA TWAS yields 23 SCZ + 3 BIP differentially expressed circRNAs and 22 SCZ + 4 BIP trait-associated GReX loci with synaptic-pathway enrichment and divergent imaging-association direction between psychiatric and AD circGTAs, circKLHL24 standout; Desrochers (Brown) OCD abstract-sequence task-based functional/effective connectivity reveals shared circuit architecture (RLPFC, ACC/DLPFC, SMA, MTG, TOJ) but increased direct RLPFC-to-MTG coupling in OCD bypassing the usual relay — circuit-targeted intervention implications for OCD rTMS/DBS programs; Pritzker (Stanford) Tshz1+ MOR+ NAc neurons gate opioid withdrawal aversion specifically — MOR deletion in Tshz1 cells abolishes withdrawal dopamine drop and affective aversion but spares physical symptoms, mGluR8 activation reduces withdrawal aversion (non-opioid group-III mGluR adjunct to MAT); Schwendt (Florida) mGlu2 PAM LY-487379 reverses methamphetamine-induced object-in-place recognition deficit after 30d abstinence with compensatory surface mGlu2/3 upregulation in prelimbic and perirhinal cortices; Wu (Georgetown) KCNQ openers ML213/ICA-27243 partially restore hippocampal sharp-wave ripple amplitude under TBOA EAAT inhibition while AMPA/NMDA/HCN/BK/GIRK manipulations do not; CuraGen AI APOE→ACAT1 pipeline yields ~7,300 CNS-optimized candidates from ~6M molecules with three BBB-probability->0.93 leads; Narayan (NIH/NIDDK) isogenic iPSC astrocytes show APOE4 impairs clathrin-mediated membrane curvature/endocytosis with flat clathrin structures and reduced coated-pit maturation, INPP5D rescues endocytosis through clathrin-curvature promotion plus reducing lipid droplets and inflammatory signaling — APOE4 reframes as membrane-mechanics defect upstream of trafficking, multiple AD risk genes converge on astrocyte endocytosis; closing with Karami (Inria/Lorraine) AlloDyn integrating static + MD/AlphaFlow dynamic features with methodological critique of holo-with-modulator data leakage in allosteric-pocket benchmarks, best F1/MCC on preprocessed benchmark. Receptor & Reason for May 29, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate opens with Reunion Neuroscience's full Phase 2 RECONNECT readout for luvesilocin (formerly RE-104) presented this week at ASCP and APA — luvesilocin is a serotonin 2A agonist, a short-acting prodrug of 4-hydroxy-di-isopropyl-tryptamine with a ~3-4h psychoactive window designed to fit a single-day clinic workflow rather than the eight-hour psilocybin window. The single-dose 30 mg arm met the primary endpoint with a ~23-point MADRS reduction from baseline at day 7 (p=0.0094), 77% response and 71% remission, with effects emerging by day 1 and durable through day 28 and safety consistent with the psychedelic class profile. FDA breakthrough therapy designation is already in hand from February on the topline; this is the full trial. The path forward is a single pivotal Phase 3 trial which if positive would complete the registrational data package. Two reasons it matters: postpartum depression's current pharmacotherapy (brexanolone, zuranolone) is GABA-A allosteric modulation, so a 5-HT2A agonist is the first mechanistically distinct rapid-acting option in the indication; and the short psychoactive window is what will determine real-world adoption more than the day-7 response rate. Then Kentucky on lapatinib (dual EGFR/HER2 tyrosine kinase inhibitor approved for HER2+ breast cancer) — suppresses TNF, IL-1β, IL-6, COX-2, MMP, and PGE2 in LPS-stimulated macrophages and reduces both mechanical allodynia and thermal hyperalgesia in a mouse plantar-incision postoperative pain model with parallel decreases in pro-inflammatory mediators in spinal cord and dorsal root ganglia. Crucially, lapatinib does not engage central reward pathways the way opioids do; the mechanistic logic of a kinase inhibitor flattening the inflammatory cytokine surge that drives acute-to-chronic pain transitions is the repurposing pull, with the caveat that lapatinib's hepatic/cardiac toxicity tolerated in oncology is a different conversation in postoperative pain. Then Pasca (Stanford) on 22q11.2 deletion syndrome (strongest known genetic risk factor for schizophrenia at ~25-fold relative risk) — iPSC subpallial organoids and forebrain assembloids (cortical + subpallial organoids fused so migrating interneurons can be tracked into a developing cortical field) reveal that patient-derived interneurons show impaired migration, mitochondrial fragmentation, reduced oxidative phosphorylation, dysregulated calcium signaling, and a glycolytic shift consistent with the known position of mitochondrial genes in the deletion interval. Adrenomedullin — a 52-amino-acid peptide hormone signaling through CLR/RAMP2 — rescues essentially the full phenotype: mitochondrial morphology, membrane potential, respiration, calcium handling, actin retrograde flow, and interneuron migration. Mechanism runs through PKA-mediated DRP1-Ser637 phosphorylation (promotes mitochondrial fusion) and PLC-mediated calcium signaling (normalizes actin dynamics), recapitulated in primary developing human cortical tissue carrying the deletion. A GPCR agonist already studied in cardiopulmonary indications could in principle be repositioned as a developmental intervention in the population at highest genetic risk for schizophrenia, with the obvious caveat that brain-penetrance and chronic-CNS-therapeutic tolerability of adrenomedullin is unsolved. Then Ulusoy (DZNE) — 6G6 is a monoclonal antibody selective for Tyr39-nitrated alpha-synuclein. Tested in three mouse models: paraquat-induced oxidative stress on vagal alpha-synuclein overexpression, GBA1-L444P heterozygous background on the same vagal overexpression, and striatal pre-formed fibril injection into alpha-synuclein-overexpressing mice. 6G6 markedly counteracts caudal-to-rostral spreading in the first two and protects against fibril-induced aggregate pathology plus nigral dopaminergic neuron loss in the third. A small pilot human study finds Tyr39-nitrated alpha-synuclein elevated in Parkinson CSF relative to controls. Pairs naturally with last week's Wade-Martins humanized GBA-L444P transgene paper: that work argued the L444P enzyme drives a release-machinery dopamine deficit independent of alpha-synuclein accumulation, this work argues the alpha-synuclein arm of the same genetic background can be neutralized by selectively clearing a PTM-defined species. For active-immunotherapy and passive-antibody programs in PD, selectivity for pathogenic species rather than total alpha-synuclein is plausibly where the therapeutic window lives. Then Stern (Haifa) on iPSC-derived dopaminergic neurons from A53T SNCA patients — biphasic trajectory from an early hyperexcitable phase with elevated spontaneous firing into progressive hypoexcitability as cells mature, accompanied by network dysfunction and alpha-synuclein accumulation. Transcriptomic profiling at the transition shows simultaneous p53/JAK-STAT upregulation and metabolic/synaptic-maintenance downregulation — a signature that precedes functional collapse. Healthy iPSC-derived extracellular vesicles (small membrane-bound particles carrying mRNA/miRNA/protein cargo) rescue both electrophysiological deficits and cellular pathology in the human neurons and produce corresponding improvements in a mouse model. The reframe: early dopaminergic dysfunction in PD may be a hyperactivity-driven metabolic-stress problem before it is an alpha-synuclein problem, and intercellular signaling cargo from healthy cells can rebalance it — EV therapeutics as a cell-free alternative to cell-replacement strategies with a different regulatory/manufacturing profile than transplanted dopaminergic precursors. Then Siwani (Uppsala) — OLM-α2 cells (oriens-lacunosum-moleculare interneurons expressing the α2 subunit of the nicotinic acetylcholine receptor) along the longitudinal axis of hippocampus dissociate functionally: intermediate-hippocampus OLM-α2 manipulation selectively influences object exploration and novelty processing without affecting anxiety-like behavior, ventral-hippocampus OLM-α2 manipulation modulates arousal and emotionally driven avoidance without contributing to object memory. Circuit tracing reveals the dissociation is structural — intermediate hippocampus embedded in sensory cortical networks, ventral hippocampus wired to limbic structures including basolateral amygdala and nucleus accumbens — and OLM-α2 cells themselves receive only sparse direct amygdala input, meaning emotional modulation happens through their downstream targets rather than at the cell. For nicotinic-receptor or interneuron-targeted CNS interventions, the same molecular handle has different behavioral readouts depending on hippocampal subdivision. Then Morrison (Pittsburgh) — sign tracking (approach to reward-predictive cues, conventionally interpreted as incentive-salience transfer from reward to predictor) is paradoxically less common in adolescents than in adults despite adolescents carrying heavier impulsivity burden. Single-unit recordings in NAc and GRAB-DA dopamine release in male and female rats during Pavlovian conditioning at both ages reveal cue-evoked NAc activity increases with training in adolescents and develops further into adulthood; the majority of adolescents are goal-trackers or intermediates with reward-evoked activity peaking in adolescence and declining into adulthood, correlating with eventual emergence of sign tracking; cue-evoked dopamine release is markedly higher in sign-trackers than goal-trackers at every age. Interpretation: the adolescent-to-adult transition is a shift in which NAc signals dominate behavior — adolescents track outcomes through reward-evoked activity, adults shift toward cue-evoked dopamine signaling. For pediatric/adolescent ADHD and substance-use-risk pharmacology, developmentally distinct dopamine substrates that should not be treated as a single adult-like system. Then Bowman (UPenn) — attention-enabled PocketMiner uses graph attention to predict, in a single forward pass, both cryptic-pocket locations (binding sites present only transiently in a protein's conformational ensemble) and which residues elsewhere in the protein are allosterically coupled to pocket opening. Outperforms prior cryptic-pocket predictors, recapitulates known allosteric interactions, and the group experimentally confirms newly predicted cryptic pockets plus predicted mutations that allosterically gate pocket opening. For receptor-targeted drug discovery — GPCRs and ion channels whose pharmacophores depend on transient conformations or allosteric sites that conventional pocket-finding misses — a single-shot predictor returning both the site and its coupling map is the right input shape for downstream design. Then Lu/Desikan (CPT:PSP May 2026) perspective on quantitative systems pharmacology amid agentic AI — QSP has long lived under a tension between fidelity (trustworthiness) and timeliness (decision-readiness). The argument is that agentic AI collapses that tradeoff in both directions: autonomous agents absorb routine modeling tasks (parameter calibration, sensitivity analysis, scenario simulation, report generation), releasing modeler time toward problem framing and validation; QSP models themselves become first-class tools multi-agent systems call into for mechanistic reasoning, anchoring LLM-agent outputs to interpretable biology rather than free-form chain-of-thought. Editorial rather than empirical, but the practical implication is real — the right team structure for next-generation clinical-pharmacology programs is one where mechanistic models and LLM agents are operationally fused, not run in parallel silos. Closing with SpatialClaw (Shang lab, Northwestern Polytechnical University) — spatial-omics analysis pipelines have proliferated to the point where each lab writes a bespoke chain of incompatible tools and reproducibility has degraded. SpatialClaw bundles 30 specialized analysis skills under a single natural-language interface and adds a graph-based persistent memory architecture with three hierarchical layers (Session/Episodic/Semantic) where dataset metadata, analysis lineage, biological insights, and user preferences are stored as versioned nodes and edges with a deterministic promotion policy moving high-confidence findings up the hierarchy. A memory-augmented reasoning operator sits between memory store and planning agent, synthesizing retrieved experiences into task-specific guidance for each new query. Benchmarks across ten spatial-omics skills and three memory-sensitive scenarios show consistent improvement over conventional conversational agents and over a non-memory ablation. Same architectural pattern as yesterday's AutoScientists paper at the multi-agent level — decentralized memory and explicit lineage as what converts one-shot reasoning into accumulating institutional knowledge — applied here at the single-analyst level. For neuroinformatics work involving spatial transcriptomics of brain tissue, this is the kind of substrate worth watching. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-29-receptor-and-reason Fri, 29 May 2026 12:00:00 +0000 1063 Daily roundup for May 29, 2026: Reunion Neuroscience full Phase 2 RECONNECT readout for luvesilocin (RE-104, 4-OH-DiPT prodrug short-acting 5-HT2A agonist) presented at ASCP/APA — single 30 mg dose met primary endpoint with ~23-point MADRS reduction at day 7 (p=0.0094), 77% response and 71% remission, durable through day 28, FDA Breakthrough Therapy in hand, single pivotal Phase 3 path — first mechanistically distinct rapid-acting option in postpartum depression beyond GABA-A allosteric modulators (brexanolone, zuranolone); Kentucky lapatinib (EGFR/HER2 TKI) suppresses TNF/IL-1β/IL-6/COX-2/MMP/PGE2 in LPS macrophages and reduces mechanical allodynia/thermal hyperalgesia in mouse plantar-incision postop pain without engaging central reward pathways — non-opioid non-reward repurposing lane against the inflammatory acute-to-chronic pain transition; Pasca (Stanford) 22q11.2 deletion syndrome iPSC subpallial organoids + forebrain assembloids show interneuron migration defect with mitochondrial fragmentation/glycolytic shift, rescued in full by adrenomedullin via CLR/RAMP2 GPCR (PKA/DRP1-Ser637 + PLC/Ca dual pathway), recapitulated in primary human cortical tissue — GPCR-agonist repositioning lane for highest-genetic-risk schizophrenia population; Ulusoy (DZNE) 6G6 antibody selective for Tyr39-nitrated alpha-synuclein counteracts spreading in three PD mouse models including GBA1-L444P background and PFF injection with elevated nitrated alpha-synuclein in PD CSF — PTM-defined pathogenic-species selectivity as therapeutic window, interlocks with last week's Wade-Martins humanized GBA-L444P GoF/LoF dissection; Stern (Haifa) iPSC-derived A53T-SNCA dopaminergic neurons show biphasic hyper-to-hypoexcitable trajectory with p53/JAK-STAT/metabolic transcriptional signature, healthy-iPSC-derived extracellular vesicles rescue physiology and pathology — cell-free EV therapeutic as alternative to dopaminergic-precursor cell replacement; Siwani (Uppsala) OLM-α2 hippocampal interneurons (α2-nicotinic-acetylcholine-receptor-expressing) show regional functional dissociation — intermediate hippocampus drives object/novelty without anxiety, ventral hippocampus drives arousal/avoidance without object memory, paralleled by sensory-cortical vs limbic wiring — hippocampal-compartment-aware nicotinic pharmacology required; Morrison (Pittsburgh) NAc single-unit + GRAB-DA dopamine recordings during Pavlovian conditioning reveal adolescents are goal-trackers with reward-evoked activity peaking in adolescence, adult sign-trackers show markedly higher cue-evoked dopamine — developmental NAc-signal-dominance shift with implications for adolescent ADHD and substance-use-risk pharmacology; Bowman (UPenn) AE-PocketMiner graph-attention model simultaneously predicts cryptic pocket locations and allosteric coupling from a single input structure with experimental validation — single-shot druggability plus coupling map for receptor/channel allosteric pharmacophores; Lu/Desikan (CPT:PSP May 2026) perspective argues agentic AI collapses QSP fidelity-vs-timeliness tradeoff in both directions, mechanistic models and LLM agents must be operationally fused not run in parallel silos; closing with SpatialClaw (Shang, Northwestern Polytechnical) memory-augmented autonomous ecosystem for spatial omics bundling 30 skills behind natural-language interface with three-layer persistent graph memory (Session/Episodic/Semantic) + deterministic promotion + memory-augmented reasoning operator — decentralized memory and explicit lineage pattern at single-analyst level, same logic as yesterday's AutoScientists at multi-agent level. Receptor & Reason for May 28, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate opens on Kiyonaka (Nagoya) — an off-target-free chemogenetic platform that engineers a clinical-grade A2A antagonist-insensitive A2A receptor mutant that retains adenosine responsiveness, pairing the antagonist with the mutant as a chemogenetic pair to interrogate endogenous A2A function without committing to a tool drug; A2A signaling is critical for neurite outgrowth under hypoxia in neuron-like cells, and the general principle generalizes across other class-A GPCRs, giving the field a way to ask the in-vivo question of what the endogenous receptor actually does without the off-target liability that contaminates classical pharmacological dissection. Then Goldschen-Ohm (UT Austin) with collaborators at Karolinska, Iowa, and UT Arlington — a tyrosine residue in the M2-M3 linker of inhibitory pentameric ligand-gated ion channels (GABA-A and glycine receptor family) has been positionally conserved for over 600 million years, and the group used noncanonical amino acid incorporation to ablate the hydrogen-bond donor or acceptor at the backbone level (not just side-chain substitution) and show that breaking this hydrogen bond markedly reduces both agonist sensitivity and maximal channel activation in GABA-A and glycine receptors with subunit-specific effects, with molecular dynamics confirming the gating role. Disease-associated variants at this tyrosine cause epilepsy, neurodevelopmental disability, and hyperekplexia — the paper provides a mechanistic explanation. For benzodiazepines, neurosteroid analogs like brexanolone and zuranolone, and many anesthetics that act on inhibitory channels, gating linkage being a single hydrogen bond is the kind of structural anchor allosteric modulator design can lean on. Then Freude (Copenhagen) — 120-day dorsal forebrain organoids from iPSCs of 6 schizophrenia patients and 5 matched controls reveal 19 differentially expressed miRNAs (9 progenitor-linked down, 10 differentiation/synaptic-maturation-linked up) and premature GABAergic lineage specification with secondary convergence on glutamatergic and dopaminergic development. Compressed developmental timeline rather than a global block, with miRNA-mediated post-transcriptional regulation as a primary upstream substrate linking parvalbumin-interneuron deficit and excitation-inhibition imbalance threads — upstream of where current antipsychotics act, but a future-intervention point for microRNA-based modulation during the vulnerable developmental window. Then Chaudhry (Oslo) — Slc38a1 glutamine-carrier knockout mice show learning/memory deficits in Morris water maze, increased despair in forced swim, and anxiety signatures in open field while preserving sociability and social novelty; a separately identified pathogenic Slc38a1 variant in a family with depression and suicidal behavior provides the human anchor. The reframe is that GABAergic deficit in MDD/anxiety may be a substrate-supply problem upstream of receptor pharmacology — explains why direct GABA-receptor modulators have a noisy track record in mood disorders, and points toward enzyme/transporter-level interventions as a complementary lane. Then Moschak (UT El Paso) — in-vivo endoscopic calcium imaging of prelimbic cortex in 32 male/female Sprague-Dawley rats across cocaine/water self-administration plus extinction, progressive ratio, and punished self-administration; the shared-latent-variable model fits water-reward behavior but not cocaine — a meaningful caveat for the shared-latent line; distinct cost-sensitive (engaged in PR) and reward-sensitive (engaged in extinction) prelimbic ensembles predict behavior across both reward types. Prelimbic motivation is not one signal but multiple ensembles tracking distinct decision variables, and prelimbic-targeted addiction pharmacology should expect dissociable cost-tolerance versus persistence effects rather than a single motivation knob. Then Nobis (Vanderbilt) — chronic restraint stress in Dravet SCN1A mouse model produces elevated corticosterone, female-specific mortality, and bed nucleus of the stria terminalis CRF-neuron genotype-specific glutamatergic remodeling with increased spontaneous excitatory frequency in both genotypes and selectively increased sEPSC/sIPSC amplitude in Dravet. BNST CRF circuit as a maladaptive node where chronic stress and epileptic genotype interact to amplify both seizure vulnerability and psychiatric comorbidity — argues for thinking about the same circuit when treating the comorbidity load that dominates the pediatric epilepsy quality-of-life burden. Then Wade-Martins (Oxford Parkinson's Disease Centre) with Aerts (Leiden) — BAC-recombineered humanized GBA-L444P transgene on mouse Gba+/- background dissociates the gain-of-function and loss-of-function arms of GBA1-PD: L444P mice show reduced GCase activity, ER-retention-consistent high-MW enzyme accumulation, and an early/persistent reduction in dorsal striatal dopamine release on fast-scan cyclic voltammetry with no dopaminergic cell loss and no synthesis/reuptake defect (a release-machinery deficit). Both L444P transgenic mice and the simple heterozygous knockout accumulate oligomeric alpha-synuclein, but only L444P mice develop behavioral changes — dopamine release and behavioral deficits from gain-of-function, alpha-synuclein pathology from loss-of-function. Matters for GCase-targeted therapeutics (chaperones, substrate-reduction therapy, gene therapy) because the same intervention will not address both arms equally. Then Liu (Peking University Third Hospital) — Pharmacokinetic Foundation Model (PKFM) is a grey-box Transformer pre-trained across 32 drugs that ingests sparse concentration observations, dosing events, molecular descriptors, and physiological covariates and reconstructs full concentration-time profiles while preserving interpretable output. Three sparse points recover the principal absorption-elimination trajectory with R² ≈ 0.99 for midazolam and verapamil; reconstructed curves improve NONMEM covariance stability and individual prediction accuracy; contrastive-learning embeddings retrieve top-10 PBPK candidate models with 75.6% of observations within the 2-fold range. The wrapped PM Agent outperforms general-purpose programming tools on a standardized pharmacometric benchmark under explicit human pharmacometrician confirmation — right shape of an agentic system: domain-pre-trained foundation substrate for sparse-data inference plus an expert-gated workflow agent on top. Then MolLingo — a multi-agent system with Literature Agent, Chemist Agent, and Orchestrator coordinating through shared memory plus a synthesis-aware BRICS-based Fragment Enumeration scheme that decomposes molecules into chemically meaningful building blocks as block-level SMILES paired with common chemical names, bridging molecular structure and LLM semantic space for block-level reasoning and editing. Grounded in docking-derived binding-site geometry and residue-level protein context, delivers 4× docking-score improvement over the base frontier LLM with consistent drug-property optimization gains and SOTA on TOMG-Bench. Architectural lesson for receptor-targeted drug discovery is that the agent's value comes from the chemically meaningful representation, not from the language model alone. Closing with CaMBRAIN — the first causal Mamba-based state-space model for real-time continuous EEG inference, replacing the attention backbone (which incurs quadratic cost as sequences lengthen and forces sliding-window context loss) with a linear-time unidirectional SSM explicitly designed for streaming. Multi-stage self-supervised training pipeline aimed at long-range memory retention hits SOTA on 3 EEG benchmarks with >10× higher throughput than existing models and produces the first model that runs long-horizon continuous inference on variable-length EEG. Architectural substrate for next-wave clinical EEG use: antiepileptic dose titration, anesthesia depth control, and closed-loop neuromodulation. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-28-receptor-and-reason Thu, 28 May 2026 12:00:00 +0000 1095 Daily roundup for May 28, 2026: Kiyonaka (Nagoya) off-target-free chemogenetic platform engineering antagonist-insensitive adenosine A2A receptor mutant paired with clinical A2AR antagonist as chemogenetic pair — A2AR critical for neurite outgrowth under hypoxia with generalizability across class A GPCRs, way to ask the in-vivo question without off-target tool-drug liability; Goldschen-Ohm (UT Austin) tyrosine in M2-M3 linker of inhibitory pLGICs conserved over 600 million years forms H-bond to Cys-loop backbone essential for ECD-TMD gating, noncanonical amino acid backbone ablation in GABA-A and glycine receptors markedly reduces agonist sensitivity and maximal activation — mechanistic explanation for epilepsy/NDD/hyperekplexia variants and structural anchor for benzodiazepine/neurosteroid allosteric design; Freude (Copenhagen) 120-day dorsal forebrain organoids from 6 SCZ/5 control iPSCs reveal 19 differentially expressed miRNAs (9 progenitor-linked down, 10 differentiation-linked up) and premature GABAergic lineage specification — compressed developmental timeline as upstream schizophrenia substrate; Chaudhry (Oslo) Slc38a1 glutamine-carrier knockout mice show learning/memory deficit, increased despair, anxiety without sociability loss, with human Slc38a1 variant in a depression/suicide family — GABAergic substrate-supply defect upstream of receptor pharmacology as complementary MDD/anxiety lane; Moschak (UT El Paso) in-vivo endoscopic calcium imaging of prelimbic cortex in 32 rats across cocaine/water tasks reveals shared-latent-variable model fits water but not cocaine, distinct cost-sensitive (PR) and reward-sensitive (extinction) ensembles predict behavior — prelimbic tracks multiple decision variables not one motivation knob; Nobis (Vanderbilt) chronic stress in Dravet SCN1A mouse model produces female-specific mortality and BNST CRF-neuron genotype-specific glutamatergic remodeling with selectively increased sEPSC/sIPSC amplitude in Dravet — stress-genotype interaction amplifying both seizure vulnerability and psychiatric comorbidity in pediatric epilepsy; Wade-Martins (Oxford) humanized GBA-L444P transgene on Gba+/- background dissociates gain-of-function (dopamine release/behavior) from loss-of-function (alpha-synuclein pathology) — implications for GCase-chaperone, substrate-reduction, and gene-therapy programs in GBA1-PD; Liu (Peking University Third Hospital) PKFM grey-box Transformer pre-trained across 32 drugs reconstructs full PK curves from 3 sparse points (R² ≈ 0.99) with PM Agent under human pharmacometrician confirmation outperforming general-purpose programming tools — domain-pre-trained foundation substrate plus expert-gated workflow agent as right shape for computational-pharmacology AI; MolLingo multi-agent system (Literature/Chemist/Orchestrator + BRICS-based block-level SMILES) grounded in docking geometry delivers 4× docking-score improvement and TOMG-Bench SOTA — chemically meaningful representation is the lever not model scale; closing with CaMBRAIN first causal Mamba-based SSM for real-time continuous EEG inference hitting SOTA with >10× throughput and linear-time variable-length streaming — architectural substrate for closed-loop clinical EEG. Receptor & Reason for May 27, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate opens with Pircs (Semmelweis University, Budapest) integrating Hungarian nationwide real-world prescription data on cariprazine, aripiprazole, and haloperidol monotherapy in schizophrenia — haloperidol associated with worse survival and higher cumulative incidence of first registered suicide attempt than either partial agonist — with primary mouse hippocampal multi-electrode-array recordings showing acute cariprazine moderately and dose-dependently reduces spontaneous firing and prolongs burst intervals while leaving network synchrony intact, the firing changes are cell-type dependent, and computational modeling ties the network signature to selective effects on a subset of pyramidal-like units. A drug shifting burst structure rather than overall firing rate is the kind of cellular phenotype that could rationalize cariprazine's profile in negative-symptom and cognitive-symptom domains where pure D2 antagonism has historically underperformed, and it gives the field a substrate to argue over. Then Castro (Washington University in Saint Louis) on pancreatic mu opioid receptors regulating metabolism and ingestive behaviors — Oprm1 global knockout males but not females show increased body mass and increased ad libitum food intake on continuous-access feeding devices, alongside enhanced glucose tolerance and insulin sensitivity on intraperitoneal challenge; conditional deletion of mu opioid receptors in glucagon-promoter alpha cells produces no metabolic phenotype, ruling out alpha-cell mu opioid signaling and pointing to beta-cell or non-alpha-cell pancreatic mu opioid receptor populations. Opens a lane for peripherally restricted mu opioid ligands to manage the metabolic and weight effects of methadone/buprenorphine OUD treatment and chronic-pain opioid analgesia historically ascribed to hypothalamic/mesolimbic mu opioid signaling. Then Connor (Macquarie) — (+)-trans-cannabidiol is a CB2 receptor agonist, rapid concentration-dependent hyperpolarization in AtT20 cells stably expressing human CB2 with pEC50 6.63 (~230 nM) and maximal effect ~90% of the reference full agonist CP55940, pertussis-toxin-blocked Gi-coupled and competitively inhibited by AM630 with Schild slope 1.1; behaves as low-efficacy low-potency CB1 agonist; in silico modeling identifies a (+)-enantiomer-specific residue contact in CB2 absent in the natural minus-trans enantiomer. Cleanest CB2 agonist in the cannabidiol-scaffold family, a starting point for CB2-selective medicinal chemistry that does not collapse into the polypharmacology of natural minus-trans-CBD. Then Sofuoglu (Yale) with Compton (Penn) on an opioid-withholding human laboratory paradigm during methadone OAT in 23 patients with co-occurring OUD and chronic pain (mean methadone dose ~85 mg/day) — randomized double-blind crossover of acute oral cannabidiol at 400/800/1200 mg across two phases (pre-OAT with methadone delayed, post-OAT after methadone administration) measuring conditioned pain modulation (descending inhibition) and temporal summation of pain (central sensitization) plus craving and cognition. Phase- and dose-dependent effects of CBD on pain modulation — methodological frame of dosing relative to opioid trough versus peak as the transferable advance for chronic-pain adjuncts in OUD, where most CBD work treats dose as monotonically additive and ignores where in the opioid-agonist cycle the patient sits. Then Jakubik (Czech Academy of Sciences) — KH-5 muscarinic antagonist functional potency in IP accumulation and GTPγS exceeds measured orthosteric binding affinity beyond what classical competitive antagonism explains. At human M1 in CHO cells antagonism is saturable and probe-dependent (allosteric hallmarks), at M2 it shows dualsteric behavior — simultaneously engaging the orthosteric pocket and an extracellular allosteric vestibule confirmed by molecular dynamics and site-directed mutagenesis at vestibule-lining residues. Dualsteric/dual-acting modes are pharmacologically distinct from the more common bitopic muscarinic ligands and predict different selectivity behavior across M1-M5. For programs trying to recapitulate xanomeline's M1 preference without broad-muscarinic GI side effects, dualsteric antagonist scaffolds are an under-explored chemical lane. Then Shadani (Monash) — single 1.5 mg/kg psilocybin in C57BL/6J mice produces acutely enhanced huddling and hypothermia selectively in female mice (5-HT2A-mediated effects with sex-specific magnitude), with post-acute novelty-seeking and social investigation enhanced at 4 hours. The post-acute window is where psilocybin's clinical mechanism in depression and anxiety has been pointing distinct from the acute psychedelic experience; differential sex profile across acute hypothermia and pro-social signatures means trials in indications such as depression-with-social-withdrawal phenotypes and anorexia nervosa (where female prevalence dominates) need stratified analyses for both window and sex. Then Jang (Korea University Guro Hospital, Seoul) — 70 medication-naive adolescents age 12-17 with major depressive disorder on 8-week open-label escitalopram with Children's Depression Rating Scale-Revised as outcome; baseline resting-state functional MRI uses seed-based default mode network connectivity from ventromedial prefrontal cortex, dorsomedial prefrontal cortex, and posterior cingulate cortex. Baseline DMN connectivity patterns prospectively predict escitalopram response, with effect sizes and seed specificity differing from the adult literature — a biomarker that survives in a real adolescent cohort with a real clinical scale, in a population where pediatric SSRI prescribing is contested and benefits from rational stratification. Then Gozes (Tel Aviv) — davunetide/NAP, the 8-amino-acid neuroprotective peptide derived from activity-dependent neuroprotective protein (a master regulator of cognition with prior tauopathy clinical exposure), tested as alternative to or augmentation of conventional hormone replacement therapy in ovariectomized 3-month-old female ICR mice receiving daily intranasal NAP, estradiol, or vehicle. ADNP-estrogen steroidogenic coupling rationalizes the design; the davunetide-plus-low-dose-estrogen combination outperforms either monotherapy on cognitive and motor preservation. Repositioning argument for the Alzheimer-and-menopause overlap population where estrogen withdrawal is epidemiologically linked to accelerated cognitive decline and conventional HRT is constrained by cardiovascular and oncologic concerns. Then Wanunu (Northeastern) NanoCortex — multi-agent framework on Google Gemini development kit with separate agents for task parsing, code generation, iterative code-level self-correction, and biological interpretation plus retrieval against public sequence and annotation databases. End-to-end user submits a natural-language question and raw nanopore signal file, the system produces both an analysis pipeline (code runnable offline) and a biological interpretation. Substantially higher usability on complex analytical tasks than a general-purpose LLM alone, with offline-runnable code addressing audit and reproducibility concerns that closed-API agentic systems raise. Representative architecture for agentic AI in biomedical workflows where the underlying tool ecosystem is mature but fragmented — agentic value is orchestration plus interpretation, not generation from first principles. Closing with ETIS (Cergy Paris University) spiking neural model of dopamine-driven reinforcement learning comparing disinhibition versus direct control — midbrain dopaminergic neurons receive convergent direct excitatory/inhibitory projections and disinhibitory motifs through GABAergic intermediaries (rostromedial tegmental nucleus and elsewhere), and the respective computational contributions have not been cleanly separated. Two fully spiking neural models (one direct, one disinhibition-only) put into a three-armed bandit value-learning task. Both reach asynchronous-irregular firing, but the direct-control model is more sensitive to incoming-signal disruption and learns less reliably; disinhibition is a functional advantage not just a wiring feature. For computational psychopharmacology, disinhibition motifs are exactly where benzodiazepines, GABAergic anesthetics, and several next-generation depression pharmacologies act — their effects on dopamine-mediated value learning may be larger than direct-pathway models predict. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-27-receptor-and-reason Wed, 27 May 2026 12:00:00 +0000 937 Daily roundup for May 27, 2026: Pircs (Semmelweis) cariprazine on hippocampal multi-electrode-array recordings — cell-type-dependent reduction of spontaneous firing and prolonged burst intervals while preserving network synchrony, tied to Hungarian nationwide real-world schizophrenia outcomes where haloperidol was associated with worse survival and higher first suicide-attempt incidence than cariprazine or aripiprazole — cellular phenotype as rationale for D2/D3 partial-agonist performance in negative- and cognitive-symptom domains where pure D2 antagonism has underperformed; Castro (Washington University in Saint Louis) pancreatic mu opioid receptor regulation of metabolism and ingestive behaviors — Oprm1 global knockout males (not females) show increased mass and food intake with enhanced glucose tolerance and insulin sensitivity, glucagon-Cre alpha-cell-specific deletion has no effect implicating beta-cell or non-alpha-cell pancreatic mu opioid receptor populations — peripherally restricted mu opioid lane for the metabolic effects of methadone/buprenorphine OAT historically ascribed to hypothalamic/mesolimbic mu opioid signaling; Connor (Macquarie) (+)-trans-cannabidiol is a CB2 receptor agonist (pEC50 6.63, ~90% max vs CP55940, PTX-blocked, AM630-antagonized Schild slope 1.1), low-efficacy low-potency CB1 agonist; in silico modeling identifies (+)-enantiomer-specific residue contact — cleanest CB2 agonist in the CBD scaffold family for CB2-selective medicinal chemistry outside the polypharmacology of natural minus-trans-CBD; Sofuoglu (Yale) + Compton (Penn) opioid-withholding human laboratory paradigm during methadone OAT in 23 OUD-with-chronic-pain patients — randomized double-blind crossover of oral CBD 400/800/1200 mg across pre-OAT and post-OAT phases with conditioned pain modulation and temporal summation of pain as outcomes — phase- and dose-dependent CBD effects, methodological frame of dosing relative to opioid trough versus peak as the transferable advance; Jakubik (Czech Academy of Sciences) KH-5 muscarinic antagonist shows dualsteric behavior simultaneously engaging M2 orthosteric pocket and extracellular allosteric vestibule (saturable probe-dependent at M1, MD + site-directed mutagenesis at M2 vestibule residues) — under-explored chemical lane for xanomeline-style M1-preferring antipsychotics without broad-muscarinic GI liability; Shadani (Monash) 1.5 mg/kg psilocybin in C57BL/6J mice acutely enhances huddling and hypothermia selectively in females and post-acutely (4h) enhances novelty-seeking and social investigation — sex- and window-stratified analyses required for psilocybin trials in female-prevalent indications; Jang (Korea University Guro Hospital) baseline DMN resting-state functional connectivity (vMPFC/dMPFC/PCC seeds) prospectively predicts CDRS-R response to 8-week escitalopram in 70 medication-naive adolescents 12-17 — biomarker that survives in a real adolescent cohort for contested pediatric SSRI prescribing; Gozes (Tel Aviv) davunetide (NAP, 8-AA neuroprotective peptide from ADNP) + low-dose estrogen intranasal combination outperforms either monotherapy on cognitive and motor preservation in OVX female mice — CNS-targeted HRT alternative for the Alzheimer-menopause overlap population; Wanunu (Northeastern) NanoCortex multi-agent framework on Gemini ADK with task parsing/code generation/self-correction/biological interpretation agents for nanopore sequencing — offline-runnable generated code addresses audit/reproducibility for closed-API agentic systems, agentic value is orchestration plus interpretation not first-principles generation; closing with ETIS (Cergy Paris) spiking neural comparison of disinhibition vs direct control in dopamine-driven RL — both reach asynchronous-irregular firing in a 3-armed bandit but direct is more sensitive to disruption and learns less reliably, disinhibition is a computational advantage with implications for benzodiazepine/GABAergic/next-gen depression pharmacology effects on dopamine value learning. Receptor & Reason for May 26, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate opens with the cleanest CNS receptor-pharmacology paper of the week — Steinfeld/Niswender/Javitch (Columbia) deploy CODA-RET, a BRET-based assay that measures direct Gq recruitment to defined full-length receptor pairs, to resolve the long-standing class C GPCR question of which protomer carries the signal in a metabotropic glutamate one/five heterodimer. The answer is mGlu1 — signaling flows predominantly through mGlu1, with domain-swap chimeras localizing the dominance to the transmembrane domain. The drug-discovery consequence is allosteric inversion: cis-acting mGlu5 PAMs and NAMs that work at the homodimer flip behavior at the heterodimer because mGlu5 is not driving Gq there, and the mGlu5-selective NAM MTEP is silent at the heterodimer. The conceptual offer for medicinal chemistry is that a trans-acting mGlu5 PAM, allosterically perturbing the opposite protomer, would in principle be homomer-selective — the selectivity profile that has been hard to engineer pharmacologically and that matters because mGlu5 homomer pharmacology has spent two decades trying not to drag in mGlu1 effects in cerebellum and elsewhere. Reframes how to read existing PAM/NAM datasets in tissues where heterodimers are abundant and gives medicinal chemistry a coordinate — protomer identity plus cis-versus-trans engagement — to design selectivity across the homomer-heterodimer space. Then Gundersen/Bramham (Bergen) with Greger (Cambridge) target the synaptic AMPA receptor stabilization layer with a TAT-fused 21-amino-acid peptide (TARP-pep) mimicking the TARP carboxy-terminal motif that competitively occupies the ARC N-lobe binding pocket — disrupts TARP-ARC interaction, increases surface stabilization of AMPA receptors at the synapse in cultured hippocampal neurons, and via acute intrahippocampal infusion rescues the post-weaning social-isolation fear-memory deficit (a behavioral phenotype that has been resistant to pharmacological intervention because the underlying receptor-trafficking lesion was not directly targetable). Destabilization of synaptic AMPA receptors is implicated transdiagnostically in cognitive dysfunction across schizophrenia, depression, and neurodevelopmental syndromes — most existing AMPA pharmacology (ampakines, PAMs) works on gating rather than surface stabilization, so a PPI inhibitor at the TARP-ARC interface is a different lane, and the in vivo proof of concept moves the strategy from speculative to a chemistry program that could plausibly evolve into a small-molecule or stapled-peptide therapeutic. Then Yang/Sun (Peking University National Institute on Drug Dependence) translate a prior cross-substance GWAS signal (ANKS1B as shared genetic risk locus across heroin/methamphetamine/alcohol dependence) into mechanism — ANKS1B expression drops in nucleus accumbens after extended cocaine access in rats, manipulation modulates escalation of cocaine intake in a long-access self-administration paradigm (the closest preclinical correlate of loss-of-control), and the molecular chain runs through ANKS1B interaction with the histone acetyltransferase CBP controlling H3K27 acetylation at FoxO3 target loci to repress FoxO3 transcriptional output during escalation. ANKS1B itself is hard to drug directly, but CBP and FoxO3 each have tractable chemical biology, and specificity to escalation rather than acquisition is the right profile for substance-use-disorder indications historically dominated by replacement and aversive agents. The cross-substance GWAS signal also makes ANKS1B-pathway pharmacology a candidate for polysubstance phenotypes current treatments do not address well. Then Gusinskaia/Martin-Garcia (Maldonado lab, Universitat Pompeu Fabra) — female mice meet more cannabinoid-addiction-like criteria and at higher severity in a WIN 55,212-2 (CB1/CB2 full agonist) operant self-administration model across persistence-during-drug-free-periods, compulsive-responding-under-punishment, and progressive-ratio motivation criteria, with medial prefrontal cortex microRNA profiling identifying a sex-specific signature in susceptible females converging on synaptic plasticity and glutamatergic signaling targets. The methodological caveat that WIN is a full agonist rather than the partial agonist delta-9-THC is acknowledged; even so, female-vulnerable miRNA signature convergence on glutamatergic plasticity is consistent with cannabis use disorder sharing more biology with stimulant-class addictions than opioid-class, with implications for the pharmacology lane — CB1 neutral antagonists and allosteric modulators rather than agonist replacement. Then Saleki/Kienlen-Campard (UCLouvain) dissect the presenilin 2 N141I familial AD mutation outside amyloid processing — both presenilin 2 deletion and N141I reduce neuronal lipid content, but exogenous lipid supplementation rescues only the N141I phenotype (partial loss-of-function signature); N141I neurons also have reduced OPA1 mitochondrial fusion regulator, restored by lipid; RNA-sequencing identifies Gbf1 (Golgi-specific guanine nucleotide exchange factor) as selectively downregulated in N141I but not knockout tissue, with Gbf1 knockdown in fibroblasts recapitulating the N141I lipid profile. Proposed presenilin-2-Gbf1-lipid-mitochondria axis sits outside amyloid processing entirely — the kind of presenilin-2-specific liability the upcoming gamma-secretase modulator wave needs to be aware of, with Gbf1 trafficking or its downstream lipid axis as a separable disease-modifying lever with potential read-through to sporadic cases where presenilin 2 function is altered. Then Abdelrahman/Gabr (Weill Cornell) built a high-throughput cellular thermal shift assay platform — CETSA reads target engagement via drug-induced thermal stabilization in cells, the right format for membrane-receptor target validation that is hard to do biochemically — and screened approximately 40,000 compounds against LILRB4/ILT3, an inhibitory immune receptor whose signaling on microglia restrains amyloid clearance and amplifies neuroinflammation. IB15C is a submicromolar binder validated by microscale thermophoresis, surface plasmon resonance, docking, and site-directed mutagenesis; functionally in human iPSC-derived microglia, IB15C disrupts LILRB4-ApoE interaction, reduces SHP1/SHP2 phosphorylation, suppresses inflammatory cytokine secretion, and enhances amyloid uptake. First small-molecule tool for microglial disinhibition in AD beyond antibody approaches — chemistry now in hand, in vivo PK is the next question. Then Epuran/Albrecht (Fribourg) use floxed Per1 mice with viral Cre delivery to selectively delete Per1 in either the suprachiasmatic nucleus (master circadian pacemaker) or the lateral habenula (central to depressive behaviors and antidepressant response) — SCN Per1 loss blunts light-induced phase advances at CT22 without affecting light-induced improvement in despair-like behavior; LHb Per1 loss does the opposite (intact phase shifting, lost light-induced antidepressant-like effect). The two effects of light run through anatomically separable Per1-dependent mechanisms — antidepressant strategies engaging habenular Per1 induction or its downstream signaling could capture the mood benefit of light therapy without the constraints of timed exposure, predicting the antidepressant response to light should be partly orthogonal to its sleep-and-circadian effects (suspected clinically, not previously mechanistically grounded). Then Zhang/Bowman (Penn) attention-enabled AE-PocketMiner uses a graph neural network with attention to simultaneously predict cryptic pocket locations and their allosteric coupling to the rest of the protein from a single input structure — outperforms prior methods on benchmark datasets, recapitulates known allosteric interactions, and experimentally confirms newly predicted cryptic pockets plus mutations that allosterically control pocket opening. For CNS drug discovery (large multi-domain proteins like voltage-gated ion channels, transporters, class C GPCRs where allosteric modulation is the dominant selectivity strategy), a single-structure method that predicts both pocket and allosteric link in one pass collapses a multi-step MD-based pipeline and gives medicinal chemistry a basis for prioritizing mechanistically coupled pockets over transient ones. Then Li/Lu/Wang (SAI-Lab NYU) LipoAgent is a safety-aware multi-agent LLM framework for lipid nanoparticle design — toxicity is treated as a decision-level prerequisite (if toxic, efficiency prediction is clinically irrelevant), encoded with a conditional prediction objective where the model must clear a toxicity gate before producing an efficiency prediction; multiple fine-tuned LLM agents propose and verify, with human oversight invoked only on agent disagreement. Across foundation models, the framework delivers approximately 32% relative improvement in mRNA transfection efficiency prediction over previously reported lipid-design models, with wet-lab validation translating virtual rankings to biological outcomes. For CNS-relevant therapeutic delivery (blood-brain-barrier-crossing LNPs for siRNA and ASO programs against neurodegenerative targets), the agentic architecture with safety-as-gating-constraint is closer to how medicinal chemistry triages candidates, and the multi-agent disagreement-flagging gives a tractable human-in-the-loop interface that single-model screening pipelines tend to skip. Closing with Stephen/Jha (Princeton) knowledge-graph-driven expert-level reasoning for neuroscience — dual-LLM validation pipeline constructs a high-quality KG from a single authoritative neuroscience textbook, masked-LM expands it over the KG topology, multi-hop QA items are generated from graph traversal, supervised fine-tuning runs on those items, and reinforcement learning uses path-derived KG signals as implicit reward models. Resulting fine-tuned model surpasses the underlying large language models on neuroscience expert-level reasoning while using orders of magnitude fewer parameters. Two reasons it lands as a closer: methodological argument that domain superintelligence does not require web-scale heterogeneous training (structured-knowledge-distillation route is competitive, with implications for cost, auditability, and contamination avoidance), and the released artifact — a KG-derived synthetic neuroscience curriculum readers can quiz themselves on — is a clean demonstration that the KG itself, not just the model, is a useful pedagogical object. The kind of bottom-up superintelligence approach that could plausibly produce a hallucination-resistant neuroscience-specific clinical-reasoning tutor for drug mechanisms. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-26-receptor-and-reason Tue, 26 May 2026 12:00:00 +0000 1071 Daily roundup for May 26, 2026: Steinfeld/Niswender/Javitch (Columbia) CODA-RET BRET assay resolves mGlu1/5 heterodimer protomer asymmetry — signaling flows predominantly through mGlu1 with transmembrane-domain dominance, cis-mGlu5 PAMs/NAMs undergo allosteric inversion and MTEP is silent at the heterodimer, opening trans-acting mGlu5 PAM design space for homomer selectivity; Gundersen/Bramham (Bergen) + Greger (Cambridge) TAT-fused 21-AA TARP-pep disrupts TARP-ARC interaction, stabilizes synaptic AMPA receptors and rescues post-weaning social-isolation fear-memory deficit — protein-protein-interaction inhibitor at the receptor-trafficking layer bypasses ampakine-class gating pharmacology; Yang/Sun (Peking University NIDD) translate ANKS1B cross-substance GWAS signal to mechanism — ANKS1B in nucleus accumbens controls cocaine self-administration escalation via CBP H3K27ac-dependent FoxO3 repression; Maldonado (Pompeu Fabra) WIN 55,212-2 mouse self-administration shows female vulnerability to cannabinoid addiction-like behavior with PFC sex-specific microRNA signature converging on synaptic plasticity and glutamatergic targets — supports CB1 neutral-antagonist lane for CUD; Saleki/Kienlen-Campard (UCLouvain) PSEN2 N141I impairs neuronal lipid homeostasis and mitochondrial dynamics via selectively downregulated Gbf1 Golgi GEF — non-amyloid familial AD axis as separable disease-modifying lever; Abdelrahman/Gabr (Weill Cornell) HT-CETSA screens 40,000 compounds vs LILRB4/ILT3 — IB15C submicromolar binder disrupts LILRB4-ApoE, reduces SHP1/2, suppresses cytokines, enhances amyloid uptake in human iPSC-derived microglia (first small-molecule microglial disinhibition in AD beyond antibodies); Epuran/Albrecht (Fribourg) region-specific Per1 deletion dissociates light's circadian phase-shifting (SCN-gated) from its antidepressant-like effect (LHb-gated) — habenular Per1 pharmacology as mood-specific antidepressant orthogonal to chronobiology; Zhang/Bowman (Penn) AE-PocketMiner attention GNN simultaneously predicts cryptic pockets and allosteric coupling from a single input structure with wet-lab confirmation — single-pass method for CNS allosteric drug discovery; Li (NYU) LipoAgent safety-gated multi-agent LLM framework for lipid nanoparticle design with 32% relative improvement in mRNA transfection efficiency prediction and wet-lab validation — agentic architecture with toxicity-as-prerequisite; closing with Stephen/Jha (Princeton) KG-driven expert-level neuroscience reasoning from a single textbook surpasses frontier LLMs with orders-of-magnitude fewer parameters via dual-LLM KG construction + multi-hop QA generation + KG-path RL rewards — domain superintelligence without web-scale corpora and a synthetic neuroscience curriculum as released artifact. Receptor & Reason for May 25, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate opens with what looks like the most consequential mechanism paper of the week. Ding/Qiu (Songjiang Research Institute) show that Asxl3 haploinsufficiency in mice reduces cortical thickness and upper-layer projection neurons while expanding parvalbumin interneuron density and producing autism-like behavioral abnormalities — the mechanism is a chromatin handoff in which ASXL3 loss alters histone H2A monoubiquitination and derepresses the thyroid-hormone-inactivating enzyme DIO3, depleting brain thyroid hormone; conditional deletion of TH receptor alpha (Thra) in inhibitory progenitors phenocopies the PV expansion; neonatal but not adolescent TH supplementation rescues behavior (critical-window endocrine intervention); and an intein-based AAV system reconstitutes full-length ASXL3, normalizing cortical architecture and behavior in heterozygous mice while also driving efficient ASXL3 expression in non-human primate brain — three-modality target (chromatin, endocrine, inhibitory circuit) with a defined therapeutic window and a primate-validated gene-therapy delivery platform. Then Monroe/Bilbo (Duke Neurobiology) dissect a CGRP-receptor-gated immune engram from acute respiratory influenza A — TRAP2 chemogenetic capture in posterior insula at 3 days post-infection (Puerto Rico 8 strain) defines an ensemble whose reactivation produces both anxiety behavior and elevated pulmonary cytokine transcription (but no spleen change); airway-specific ablation of Calca-expressing cells (including pulmonary neuroendocrine cells) prevents the insular microglial synaptic-engulfment response; the FDA-approved small-molecule CGRP receptor antagonist rimegepant during engram formation blocks the behavioral arm of reactivation but spares the peripheral cytokine arm — clean pharmacological dissociation, migraine-drug-class repurposed to break infection-derived anxiety phenotypes (post-acute COVID and post-viral anxiety read-throughs) without disabling host defense. Then Taugher-Hebl/Wemmie (University of Iowa) — global ASIC1A disruption increases CO2-evoked jumping while reducing freezing with elevated basolateral amygdala c-Fos; direct BLA acidification recapitulates the shift; CO2 normally suppresses firing in most BLA neurons but ASIC1A disruption enhances suppression of putative interneurons specifically and facilitates excitation of putative principal neurons — ASIC1A's pH-sensing function concentrates in the inhibitory population, and its loss shifts defensive response toward active (jumping) defense, identifying a non-monoamine pH-amygdala panic-disorder substrate for the SSRI/benzodiazepine-saturated market. Then Shilikbay/Ceman (UIUC) RNA helicase MOV10 brain-specific KO mice show enhanced fear memory aligned with diffusion-tensor-imaging-revealed impaired canonical fear-circuit structural connectivity, with fear-memory retrieval producing increased fMRI activity in cortical regions (not canonical subcortical fear circuit) — model: MOV10 loss elevates GABRA2 in hippocampus, reduces canonical anatomical connectivity, shifts fear-memory retrieval substrate to corticalized response; molecular frame for the intractability of certain PTSD memories and prediction that benzodiazepine-class GABAA modulators may be the wrong intervention for this subtype, with substance-dependence comorbidity potentially sharing the corticalized-learning architecture. Then Pusuluri/Calhoun (Georgia State) introduce DESINE (Dynamic Estimation of Spatially Interactive Networks) — joint-density-distribution-based 4D spatial network framework applied to 91 network pairs in 508 subjects (315 controls, 193 schizophrenia) showing schizophrenia patients have a markedly constrained dynamic repertoire (lower variability of spatial-interaction metrics, particularly in high-activity regions) with metrics correlating to cognitive performance, PANSS positive/negative symptom scores, and chlorpromazine-equivalent antipsychotic dose — global voxel-agnostic biomarker shape (cross-site reproducibility friendly) with antipsychotic-response readout potential. Then Byrne/Le (Genentech) benchmark cDNA-PCR-seq vs direct RNA sequencing on nanopore long reads and show direct RNA sequencing achieves higher accuracy/sensitivity across neuropathological gene sets, then build a multi-omic atlas in iPSC-derived neurons combining DRS transcript abundance + m6A status + poly-A tail length with ribosome profiling and mass spectrometry — DRS+Ribo-seq synergistically predict protein abundance, and m6A modification is inversely related to mRNA abundance in a manner gated by ribosomal A-site engagement (mechanistic link between RNA modification and translation efficiency); applied to post-mortem AD brain, m6A profiles distinguish early- vs late-stage disease — drug-discovery-grade resolution for the modification layer of neurodegeneration-relevant transcripts. Then Finol-Urdaneta/Ng (Wollongong) automated patch-clamp framework calibrated against 25 benign/likely-benign + 16 pathogenic/likely-pathogenic CACNA1G variants (encoding Cav3.1 T-type calcium channel relevant to absence epilepsy, cerebellar ataxia, neurodevelopmental phenotypes): >80% of P/LP variants show reduced current density, deactivation kinetics add discriminatory information for preserved-CD variants, and 2 of 5 Australian patient VUS are reclassified as likely pathogenic via voltage-sensor-perturbation signatures under ACMG/AMP guidelines — clinical-grade ion-channel functional-evidence pipeline templated for other channelopathy genes. Then Pritchard/Bulmer (Cambridge) show hyper-IL-6 (IL-6 complexed with soluble receptor) and Oncostatin M elicit calcium mobilization in a subset of murine colonic sensory neurons attenuated by pan-JAK inhibition or selective TYK2 inhibition alone, with TRPV1 or TRPA1 inhibition also reducing the OSM-responsive population — IBD chronic visceral pain becomes an indication-expansion candidate for the deucravacitinib-class TYK2-selective allosteric inhibitor mechanism, operating through a peripheral nociceptor-sensory pathway distinct from the immune-suppression action of existing TYK2 development. Then Thi Hai Nguyen/Lee (Korea Research Institute of Bioscience and Biotechnology) — VPS26B (retromer subunit regulating surface GluA1 trafficking) is reduced in primary motor cortex of MPTP-induced PD mice with parallel surface GluA1 and synaptic protein loss, VPS26B overexpression partially attenuates these losses, and VPS26B-deficient mice show unstable accelerating-rotarod performance after MPTP — VPS26B overexpression improves rotarod in both wild-type and knockout backgrounds; model: PD motor symptoms are not exclusively a striatal-dopaminergic problem but include cortical glutamate-trafficking dysfunction correctable independent of the dopaminergic substrate, giving retromer-stabilizing chaperone-class small molecules an additional rationale in PD motor symptoms beyond Alzheimer. Closing on the mechanistic-interpretability tier — sparse autoencoder decomposition of GPT-2 XL and Llama-3 8B into 16K-32K interpretable features per layer (with human-validated semantic taxonomy κ ≥ 0.74) recovers 94% of peak brain-LLM encoding performance from semantic features alone with five a priori semantic subcategories mapping to distinct cortical regions (Spearman ρ = 0.72) — the residual mystery of LLM-brain alignment is largely a semantic-features story, and decomposing the language-model representation into sparse features lets you isolate which semantic categories are perturbed in schizophrenia or autism from naturalistic-language data rather than relying on opaque whole-layer alignment scores. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-25-receptor-and-reason Mon, 25 May 2026 12:00:00 +0000 954 Daily roundup for May 25, 2026: Ding/Qiu (Songjiang) Asxl3 haploinsufficiency mouse defines ASXL3-DIO3-thyroid-hormone-receptor-alpha axis depleting brain TH and expanding parvalbumin interneurons, with neonatal (not adolescent) TH supplementation rescue and intein-based AAV full-length ASXL3 reconstitution normalizing cortical architecture/behavior plus NHP-brain expression — primate-validated gene-therapy delivery for syndromic ASD; Monroe/Bilbo (Duke) TRAP2 insular engram of acute influenza A PR8 infection drives anxiety on reactivation, with pulmonary Calca-cell ablation and FDA-approved CGRP antagonist rimegepant blocking the behavioral but not peripheral-cytokine arm — migraine class repurposed to break infection-derived anxiety; Taugher-Hebl/Wemmie (Iowa) ASIC1A disruption in basolateral amygdala shifts CO2-evoked defensive response from freezing to jumping via enhanced interneuron suppression — pH-amygdala axis as non-monoamine panic substrate; Shilikbay/Ceman (UIUC) MOV10 KO enhances fear memory via impaired canonical fear-circuit connectivity, corticalized re-exposure response, and hippocampal GABRA2 upregulation — molecular frame for PTSD/substance-dependence intractability; Pusuluri/Calhoun (Georgia State) DESINE joint-density-distribution 4D spatial network framework on 508 subjects reveals schizophrenia as constrained dynamic repertoire correlated with cognition/PANSS/chlorpromazine-equivalent dose; Byrne/Le (Genentech) nanopore direct RNA sequencing of iPSC-derived neurons + post-mortem AD brain reveals inverse m6A-abundance relationship gated by ribosome A-site and stage-distinguishing m6A profiles in AD; Finol-Urdaneta/Ng (Wollongong) automated patch-clamp framework calibrated against 41 reference CACNA1G variants reclassifies 2/5 Australian VUS as likely pathogenic via voltage-sensor perturbation — ACMG/AMP functional-evidence pipeline for Cav3.1 channelopathies; Pritchard/Bulmer (Cambridge) pan-JAK and selective TYK2 inhibition attenuates hyper-IL-6 + Oncostatin M calcium mobilization in colonic sensory neurons funneled through TRPV1/TRPA1 — deucravacitinib-class TYK2-selective IBD pain indication expansion; Thi Hai Nguyen/Lee (KRIBB) VPS26B retromer subunit reduced in MPTP mouse M1 with surface GluA1/synaptic protein loss and rotarod instability rescued by VPS26B overexpression — cortical glutamate-trafficking substrate in PD motor symptoms; closing with sparse-autoencoder decomposition of GPT-2 XL and Llama-3 8B showing semantic features alone recover 94% of brain-LLM encoding performance with five a priori semantic subcategories mapping to distinct cortical regions (Spearman 0.72) — mechanistic-interpretability substrate for computational psychiatry naturalistic-language phenotyping. Receptor & Reason for May 24, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate opens with Lerner (Northwestern) using simultaneous fiber photometry of striatal dopamine (GRAB-rDA3m) and thalamic axon calcium activity (gCaMP8m) in mice learning the accelerating rotarod, showing that dopamine transients in dorsomedial striatum — but not dorsolateral striatum — are frequently coupled to thalamic axon activity peaks via an acetylcholine-dependent mechanism, with the coupling emerging across learning, gated by task engagement, and dynamically regulated by the recent history of dopamine activity itself. The result resolves a long-standing field controversy by showing that local nicotinic-driven dopamine release operates during natural behavior in a regionally targeted, learning-dependent, history-gated way — defining a physiological niche for nicotinic agonists and PAMs in cognitive disorders that prior diffuse-modulation framings missed. Then Hohmeister/Jhou (University of Maryland Baltimore) combine retrograde labeling from VTA with RNAscope detection of oprm1 mRNA and GABAergic markers across multiple distal brain regions, finding mu-opioid receptor expression in both GABAergic and non-GABAergic VTA-projecting neurons (disinhibition is not the whole story — direct mu-opioid actions on glutamatergic/peptidergic inputs are real) but with subcortical GABAergic afferents expressing higher oprm1 on average than cortical inputs, which are largely opioid-insensitive regardless of GABAergic identity — implications for input-source-biased opioid antagonist design and for why prefrontal top-down control of VTA dopamine is a separable engineering problem from the opioid pharmacology of subcortical addiction circuits. Then Jacquot/David (VECT-HORUS, Marseille) demonstrate camelid-derived single-domain-antibody (VHH) conjugates against transferrin receptor 1 deliver tool small interfering RNAs to deep brain structures across wild-type mice, hTfR1-transgenic mice, and non-human primates with potent RNAi activity at low nanomolar in neural cells, rapid TfR-dependent distributional clearance, and 50-80% knockdown of mRNA/protein with ED50 below 1 mg/kg — a translational data point in the antisense-oligonucleotide-class clinical dosing range that combines with yesterday's Wyss-Institute 10x inter-individual TfR1 variance result to argue receptor-expression diagnostics + RNAi-conjugate platforms are halves of the same product. Then LaCroix (28bio, Maple Grove MN) tests 66 small-molecule drugs (30 seizure-associated, 36 comparators) in human iPSC-derived CNS-3D organoids with neuronal/astrocytic populations expressing the right neuroactive-receptor/ion-channel programs and reproducible spontaneous calcium oscillations, at concentration ranges anchored to reported clinical Cmax with ML integration of calcium time-series and chemistry features — AUROC 0.872, 83.3% sensitivity, 88.9% specificity for clinical seizure liability with graded clinical-prevalence stratification, beating published in-vitro and preclinical models. Pharmacokinetically-anchored functional human-tissue safety pharmacology with order-of-magnitude discrimination is now real for CNS, with the bottleneck moving from "do we have a model" to "do we have labeled clinical datasets to train against" for cardiovascular, attention/arousal, and idiosyncratic neuropsychiatric adverse events. Then Kwon (Lieber Institute) applies spatial transcriptomics at laminar, microenvironment, and cellular scales in postmortem human dorsolateral prefrontal cortex showing schizophrenia transcriptional changes are strongest in glia-enriched domains (layer 1/meninges and white matter) with prominent microglia-associated gene downregulation — anatomically dissociated from neuronal-rich gray-matter localization of genetic risk. Within microenvironments, neuropil shows the most prominent down-regulation of activity-dependent synaptic genes and inhibitory-neuron markers, and at the cellular level these reflect intrinsic cell-type alterations. All three scales converge on BDNF-TrkB signaling and inhibitory-circuit dysfunction — TrkB-targeted programs in TRD become candidates for indication expansion to negative-symptom/cognitive-symptom schizophrenia, and polygenic risk scores are insufficient biomarkers (you want spatial-transcriptomic surrogate signatures of the disease-state compartment). Then Bethus/Mari (Institut de Pharmacologie Moléculaire et Cellulaire, Valbonne) characterize AETA — an APP-derived peptide distinct from amyloid-beta that modulates NMDA-receptor activity in healthy brain — finding significantly elevated AETA in human AD hippocampal/prefrontal-cortex tissue with female-predominant elevation, and building an AETA-m transgenic mouse with chronically increased brain AETA that recapitulates female-biased hippocampal transcriptional changes paralleling human AD vulnerable regions, plus impaired NMDA signaling, dendritic spine loss, and memory deficits (males spared). AETA enters the AD target conversation as a third APP-derived peptide with its own receptor pharmacology and antibody-capture/secretase-modulation/NMDA-axis-modulation entry points, with female-biased phenotype tracking human prevalence in a way amyloid-beta-centric models have not. Then Cruces (Montreal Neurological Institute, McGill) uses the [18F]MK-6240 tau-PET tracer in 28 TLE patients vs 28 controls — markedly increased uptake in bilateral superior/medial temporal and parietal regions following functional and structural connectivity (spatial spread along network paths consistent with tau propagation), with phospho-tau immunohistochemistry confirmation in 5/6 operated cases. Reframes temporal lobe epilepsy as a tauopathy-overlap syndrome with seizure-driven (non-Alzheimer) tau substrate for the cognitive decline, opening TLE cognitive-decline cohorts as a cleaner test population for anti-tau immunotherapies and aggregation inhibitors that have struggled in mild AD. Then Madamanchi/Melkani (UAB) Drosophila glial-specific human tau and amyloid models show disease-induced gut microbiome shifts (Lactobacillus, Acetobacter) and axenic microbiome depletion markedly worsens sleep-circadian rhythms, memory, and locomotor function via amplified neuroinflammatory signaling, increased apoptotic gene expression, lipid dysregulation, and altered synaptic markers — mechanistically gated through energy-stress induction (elevated phospho-AMPK) with failed proteostasis restoration (ubiquitinated-protein and Ref2p autophagy-adaptor accumulation, impaired autophagic flux) driving tau, phospho-tau, and Aβ42 buildup. Time-restricted feeding, autophagy enhancers (trehalose/spermidine), and circadian-tuned dosing of existing AD drugs become more biologically motivated than under correlational gut-AD framings. Then Fang/Piray (USC) Bayesian decomposition of volatility vs stochasticity across three tasks and two large online samples identifies three computational phenotypes — intact, stochasticity-blind (treat noise as change, over-update), and volatility-blind (treat change as noise, under-update) — with double dissociation against transdiagnostic psychiatric dimensions: stochasticity-blind learners score higher on internalizing (anxiety/depression), volatility-blind learners on externalizing (behavioral addiction/compulsivity). Distinct symptom dimensions correspond to opposite failures of inference about uncertainty, motivating computational-phenotype-stratified psychiatric trial design and predicting differential responses to learning-rate-modulating pharmacotherapy (SSRIs slowing updates aligned with anxious-depressed; stimulants speeding updates aligned with compulsive-behavioral). Closing with DeepMind (Kasenberg, Daw, Stachenfeld and collaborators) DataDIVER — an automated discovery pipeline that searches over short computer programs jointly optimized to fit behavioral data and to be interpretable, producing readable RL/cognitive-modeling-family programs that surface novel mechanistic insights into human and animal learning confirmed by re-analysis of existing data. Broadens the hand-crafted Rescorla-Wagner/Q-learning/hierarchical-Bayesian model family and could plausibly discover volatility-stochasticity-style dissociations directly from raw behavior without modeler-imposed Bayesian structure — the methodological complement to item 9. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-24-receptor-and-reason Sun, 24 May 2026 12:00:00 +0000 1289 Daily roundup for May 24, 2026: Lerner (Northwestern) simultaneous photometry of striatal dopamine (GRAB-rDA3m) and thalamic axon calcium (gCaMP8m) during rotarod learning shows thalamus drives acetylcholine-dependent local dopamine release in dorsomedial (not dorsolateral) striatum, learning-emergent and dynamically history-gated — defines a physiological niche for nicotinic modulators previously treated as diffuse cognitive enhancers; Hohmeister/Jhou (Maryland Baltimore) retrograde-RNAscope cell-type-resolved oprm1 mapping across VTA afferents finds higher mu-opioid receptor expression in subcortical GABAergic inputs than cortical (largely opioid-insensitive) inputs, supporting input-source-biased opioid antagonist design; VECT-HORUS camelid TfR1-VHH conjugates deliver tool siRNAs to deep brain across mice/hTfR1-transgenics/NHPs with 50-80% knockdown at ED50 <1 mg/kg — translational dosing range, complementing yesterday's Wyss 10x inter-individual TfR1 variance constraint; 28bio iPSC-derived CNS-3D organoid calcium-network + ML predicts clinical seizure liability across 66 pharmacokinetically-anchored drugs at AUROC 0.872, sensitivity 83.3%, specificity 88.9% — order-of-magnitude functional-tissue CNS safety pharmacology with graded clinical-prevalence stratification; Lieber Institute spatial transcriptomics at laminar/microenvironment/cellular scales in dlPFC localizes schizophrenia transcriptional change to glia-enriched layer-1/WM domains (dissociated from neuronal genetic-risk locus), neuropil down-regulation of activity-dependent synaptic and inhibitory-marker genes, with multi-scale convergence on BDNF-TrkB and inhibitory-circuit dysfunction — TrkB-targeted TRD programs are indication-expansion candidates; Bethus/Mari (Valbonne) APP-derived AETA peptide enters AD target list — elevated in female AD brain, transgenic AETA-m mouse recapitulates female-biased hippocampal transcriptional changes plus NMDA-signaling impairment, spine loss, and memory deficits with male sparing; McGill MNI [18F]MK-6240 PET in 28 TLE patients shows markedly increased bilateral temporal and parietal tau following connectivity with phospho-tau IHC confirmation — reframes TLE as tauopathy-overlap, opening TLE cognitive-decline cohorts as cleaner test populations for anti-tau therapeutics; UAB Melkani Drosophila glial-tau/amyloid + axenic microbiome depletion shows worsened sleep-circadian/memory/locomotor outcomes via energy stress (p-AMPK up) and impaired autophagic flux driving tau/phospho-tau/Aβ42 buildup — gut-microbiome axis gated through autophagy/circadian; USC Piray Bayesian volatility-vs-stochasticity decomposition identifies stochasticity-blind learners (higher internalizing — anxiety/depression) vs volatility-blind learners (higher externalizing — compulsivity/behavioral addiction), a double dissociation motivating computational-phenotype-stratified trial design; closing with DeepMind DataDIVER automated discovery of short interpretable computer-program cognitive models from behavioral data surfacing novel mechanistic insights confirmed by re-analysis — broadens the RL/cognitive-modeling family beyond hand-crafted Rescorla-Wagner variants. Receptor & Reason for May 23, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate opens on the most consequential CNS-biologics paper of the week — Gorman/Ingber (Wyss Institute, Harvard) quantify 11 canonical brain-shuttle receptor targets (transferrin receptor, CD98 heavy chain, basigin, LRP8, and 7 others) at gene and protein level across 11 large cohorts spanning normal brain, Alzheimer's, Parkinson's, Huntington's, and ALS in isolated human brain microvascular endothelial cells and brain microvessels. Two-part result — reassuringly, expression is remarkably stable across brain regions, sexes, ages, and all four neurodegenerative diseases (disease state is not collapsing or amplifying the shuttle target pool at the BBB), but between individuals, expression of the most-used shuttle targets varies by roughly an order of magnitude (a fixed-dose transferrin-receptor-shuttle biologic will produce up to 10x differences in brain exposure across patients invisible in plasma PK). Implications: dose-finding studies need to be powered for inter-individual receptor-expression variability, peripheral diagnostics for shuttle-receptor expression become a real product opportunity, healthy-volunteer-to-patient dose extrapolation is risk-reduced. Then the most substantial tau paper of the week — Lasagna-Reeves (Baylor) single-nucleus RNAseq of aged tau-knockout cortex against age-matched wild-type and on a cerebral amyloid angiopathy background revealing tau ablation is not neutral in health but actively rewires excitatory neuron identity (distinct subtype with unique glutamatergic signaling), drives astrocytes into synaptoprotective states (not reactive A1), pushes oligodendrocytes into connectivity/plasticity-supporting programs, and engages microglial structural remodeling rather than disease-associated inflammatory signatures — with disease-model arm showing tau ablation pulls cell types back toward the healthy-tau-KO protective configurations, and that cell-state restoration tracks with functional rescue. Implications for tau-lowering therapeutics (ASOs/destabilizers/immunotherapies): benefit may be substantially about releasing glia into protective states rather than removing toxic tau from neurons, motivating glial cell-state CSF/imaging biomarkers instead of soluble tau/NfL and predicting flatter clinical dose-response than tau-engagement assays suggest. Then a Parkinson's mechanism paper from Szelechowski (Toulouse INFINITY) — the stress-response transcription factor Foxo3 is rapidly induced in dopaminergic-like cells in response to complex-I inhibition and α-synuclein aggregation, but neuron-specific Foxo3 deletion only modestly attenuates early Parkinson-like phenotypes whereas microglia-specific Foxo3 deletion produces substantially larger rescue — microglial Foxo3 is the more important determinant of dopaminergic neuron survival, switching between homeostatic and neurotoxic microglial states. A microglia-selective Foxo3 modulator would be a different shape of asset than broad anti-inflammatories or LRRK2-pathway compounds — closer to immuno-oncology cell-state-selective TF pharmacology — and explains LRRK2 trial failures as wrong-node attacks on the right pathway. Then an Edinburgh astrocyte paper from Hardingham — cortical astrocyte co-culture triggers widespread DNA methylation and chromatin accessibility remodeling in developing cortical neurons distinct from the neuron-intrinsic program, with the load-bearing mechanism being transcriptional repression of inwardly-rectifying potassium channels (Kir family) in the cortical neuron, flipping intrinsic excitability into a pro-excitatory zone. Neurons start firing earlier in development without first building functional excitatory synapses, with cell-autonomous Kir homeostasis bringing excitability back. Disease-genetics layer: astrocyte-induced neuronal gene program is significantly enriched for risk genes from schizophrenia, epilepsy, ADHD, and Alzheimer's — a substantial fraction of psychiatric risk-gene expression in developing cortex is being set by the astrocyte through this Kir switch, motivating Kir-channel modulators as developmental-onset psychiatric candidates and showing neuron-only iPSC drug screens miss the astrocyte-driven half. Then two pharmacological-rescue papers in different channelopathies — Adney (Northwestern) iPSC-derived excitatory neurons carrying the SCN2A-M1879T pathogenic variant show gain-of-function sodium current and AP-shape changes on patch clamp and all-optical EP, with selective Nav1.2 blockers at clinically achievable concentrations restoring firing properties toward isogenic-control baseline — a clean in-vitro proof of concept for precision SCN2A pharmacology and a throughput platform for variant-and-compound screening; and Kang (Vanderbilt) four-phenylbutyrate (PBA), an approved chemical chaperone for urea-cycle disorders, rescues GABRA1 misfolding variants by increasing total and surface α1 subunit expression and restoring GABA-A currents in HEK cells with in vivo extension in a Gabra1-A322D knockin mouse, generalizing the chaperone-rescue paradigm across GABA-A subunit variants beyond the prior gamma2-Q390X work and opening GABRA1-associated DEEs to a PBA-based program. Then two complementary depression papers — Shukla (Wyoming) with Sibille and McCullumsmith reanalyze postmortem sgACC RNAseq to distinguish progressive (linear-trend-across-clinical-stages) from state-specific (episode-vs-remission) depression signatures, finding a separable progressive component enriched in microglial inflammatory programs, mitochondrial metabolism, and specific GPCR-signaling cassettes with repositionable drug candidates — treatment-resistant depression is not just a more severe version but has a partially distinct molecular fingerprint where unmet pharmacological need lives; and Nagy (McGill) longitudinal transcriptomic/epigenomic/regulatory-network analysis across adolescence reveals immediate-early-gene cassette (c-Fos/Arc/Egr1) showing strongest sex-biased adolescent signals with males/females on opposing trajectories, with males skewing toward activity-dependent neuronal signatures and females toward stress-responsive/immune signatures, and the female adolescent stress/immune cassette overlapping the molecular signature of adult depressed brain — adolescent prevention pharmacology is closer to viable than the field assumed, and biomarker-stratified trial design would tease apart female-biased antidepressant efficacy. Then a theoretical paper from Averbeck (NIMH) and Brunel — computational models of prefrontal cortical attractor dynamics with three-factor learning rules show that synaptic pruning and myelination crossing threshold during late adolescence push the attractor landscape into a regime where strongly encoded memories (aversive/paranoid/depressed) become deeply embedded and resistant to extinction, producing psychiatric-like perseverative/rigid behavior as a generic consequence of the structural transition. 5-HT2A psychedelic agonists and NMDA-antagonist ketamine-class drugs are reframed as attractor-state-softening levers rather than receptor-deficiency repairs, predicting synergy between synaptic-plasticity-enhancing and attractor-softening drug combinations. Closing on the agentic-AI side — Han SMDD-Bench is the first multi-turn, long-horizon, agentic small-molecule drug-design benchmark hard enough to be useful (502 task instances, 5 task types: 2D pharmacophore identification, interaction-point discovery, scaffold hopping, lead optimization, fragment assembly across 102 unique protein targets, each guaranteed solvable). Frontier closed-source models solve the easier task types but degrade sharply on lead optimization and fragment assembly (multi-step chemical reasoning), open-source chemistry-fine-tuned models do better on narrow tasks but lose on cross-domain transfer — the benchmark differentiates models in a way previous-generation drug-design benchmarks did not, providing the public hard multi-turn benchmark the agentic drug-discovery field has needed for the small-molecule side. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-23-receptor-and-reason Sat, 23 May 2026 12:00:00 +0000 1251 Daily roundup for May 23, 2026: Wyss/Gorman 11-cohort quantification of brain-shuttle receptor targets shows expression is stable across brain regions, sexes, ages, and all four major neurodegenerative diseases — but varies by roughly an order of magnitude between individuals (transferrin-receptor variance produces up to 10x differences in brain exposure invisible in plasma PK), motivating receptor-expression-aware dose-finding and peripheral diagnostics for CNS-biologics; Lasagna-Reeves Baylor single-nucleus RNAseq maps tau-knockout cell-state rewiring across excitatory neurons (distinct subtype with unique glutamatergic signaling), astrocytes (synaptoprotective not reactive A1), oligodendrocytes (connectivity/plasticity), and microglia (structural remodeling not disease-associated inflammatory), with disease-model arm showing the protective configurations recapitulate in cerebral amyloid angiopathy and cell-state restoration tracks functional rescue — tau-lowering benefit may be substantially about releasing glia into protective states, motivating glial cell-state biomarkers over soluble tau/NfL; Szelechowski Toulouse microglia-specific Foxo3 deletion produces substantially larger dopaminergic rescue than neuronal deletion in PD models — microglial Foxo3 is the more important determinant of neuron survival, reframing LRRK2 trial failure as wrong-node attack; Hardingham Edinburgh astrocyte-cortical-neuron co-culture triggers epigenome-wide remodeling via Kir-channel repression with disease-risk-gene enrichment in schizophrenia/epilepsy/ADHD/AD — neuron-only iPSC drug screens miss the astrocyte-driven half, Kir modulators become developmental-onset psychiatric candidates; Adney Northwestern SCN2A-M1879T gain-of-function iPSC neurons rescued by selective Nav1.2 blockers with all-optical-EP throughput; Kang Vanderbilt 4-phenylbutyrate chemical chaperone rescues GABRA1 misfolding variants in HEK + Gabra1-A322D knockin mouse, generalizing chaperone paradigm across GABA-A subunits and opening GABRA1-DEEs to repositioned PBA; Shukla Wyoming progressive-vs-state RNAseq decomposition of MDD in sgACC identifies separable progressive component enriched in microglial/mitochondrial/GPCR pathways with repositionable candidates — treatment-resistant depression has partially distinct molecular fingerprint; Nagy McGill sex-biased adolescent IEG regulatory architecture with female stress/immune cassette overlapping adult depressed-brain signature — prevention pharmacology in adolescents is biomarker-accessible; Averbeck/Brunel NIMH computational model — late-adolescent pruning + myelination push attractor landscape into rigid regime producing psychiatric-disorder onset window, reframing 5-HT2A and NMDA-antagonist psychedelics/ketamine as attractor-softening levers; closing with SMDD-Bench multi-turn 502-task long-horizon drug-design benchmark (2D pharmacophore, interaction-point discovery, scaffold hopping, lead optimization, fragment assembly across 102 protein targets) revealing LLM agents degrade on lead-opt and fragment-assembly multi-step chemical reasoning. Receptor & Reason for May 22, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate is anchored on the freshest paper on bioRxiv — Shehreen/Hassani/Womelsdorf/Pawliszyn (University of Waterloo) miniaturized solid-phase-microextraction probes simultaneously sample dopamine, acetylcholine, serotonin, glutamate, and GABA in nonhuman primate prefrontal cortex and dorsolateral striatum during a cognitive set-shifting task, revealing area-specific multi-neurochemical fingerprints of cognitive engagement: GABAergic and cholinergic changes predict the rest→engaged transition in PFC; dopaminergic and cholinergic changes do the same in striatum; glutamate co-modulates with acetylcholine across states in both structures (coordinated excitatory-cholinergic axis); serotonin moves in parallel in PFC and striatum consistent with common external drive (raphe in synchrony rather than independent regional tone) — a fingerprint-aware framework for drug action on a network whose state is jointly defined by five neurochemicals at once with area-specific dominance. Then a serotonergic-psychedelic-mechanism paper from Hai/Haber/Banks (UW-Madison) — acute DOI exposure on rat primary cortical cultures on MEAs drives mean firing rate increase, burst-timing acceleration, and rate-corrected spike-time-tiling-coefficient path-length shortening (a topology shift surviving firing-rate correction, addressing the principal confound in psychedelic network-rewiring claims), with ketanserin blocking the population-bursting effects but a path-length component persisting beyond 5-HT2A occupancy. Then a temporal-cortex astrocyte paper from Heimbach/Melzer (Medical University of Vienna) — chemogenetic DREADD-based pathway-isolated activation shows Gi-coupled (not Gs- or Gq-) astrocytic GPCR signaling enhances fear-memory retrieval and attenuates cue-evoked astrocyte Ca2+ transients via in vivo fiber photometry, with a broader Gi-coupled receptor repertoire in cortical astrocytes than previously appreciated (α2-adrenergic, CB1, kappa-opioid, several 5-HT subtypes — historically interpreted as neuronal — may operate substantially through astrocyte Gi). Then a Parkinson's lipid-cytokine paper from Kim/Zoghbi (Baylor) — astrocyte-enriched long-chain acyl-CoA synthetase ACSBG1 KO crossed into Thy1-α-Syn PD mice rescues behavior, reduces astrogliosis, and decreases total and pS129-α-synuclein; mechanistically KO astrocytes mount attenuated cytokine-induced transcriptional responses with reduced IL-6/RANTES/MIP-3 and reduced sphingosine 20:1 secretion (exogenous sphingosine 20:1 alone induces neuronal pS129) — astrocyte lipid metabolism as a fresh therapeutic surface in PD. Then an oligodendrocyte angle on Angelman syndrome from Yang/Huang (Brown) — UBE3A loss in iPSC-derived OPCs collapses ERβ signaling and impairs self-renewal; UBE3A sustains ERβ protein levels in OPCs; selective ERβ agonists restore OPC self-renewal and myelination in Angelman patient-iPSC systems and rescue behavioral abnormalities in mouse models — a stage-specific OPC-homeostasis pharmacological entry distinct from differentiation. Then a schizophrenia mitochondrial paper from Kambali/Rudolph (Illinois) — weak Complex I inhibitor CP2 (a tricyclic pyrone small molecule) at 8-week dosing reverses Y-maze spontaneous-alternation, startle-habituation, and three-chamber social-novelty deficits in 4×Gldc schizophrenia mice, with Western blot restoration of PGC-1α (mitochondrial biogenesis master regulator) and VDAC1 — repositions weak mtCI inhibition from neurodegeneration into schizophrenia therapeutics. Then a GLP-1-for-OUD preclinical data point from Rojas/Nguyen (Baylor) — semaglutide (0.1 mg/kg s.c.) dose-dependently reduces fentanyl IVSA rewards under both short- and long-access conditions in female Wistar rats, significantly reduces progressive-ratio breakpoint (motivation), and attenuates withdrawal-like behavior in the short-access cohort (but not long-access) — a clean female-only-design dataset bridging the Psychiatric Times GLP-1-for-AUD review covered yesterday into the opioid side. Then an addiction-circuit therapy paper from Edwards/Grella (Loyola) — TetTag tagging of dorsal-dentate-gyrus ensembles encoding either first or fourth cocaine exposure followed by CPP/extinction/reinstatement reveals first-exposure (but not fourth-exposure) engram reactivation attenuates cocaine-prime reinstatement, with saline-engram reactivation protective in females only — the early ensemble carries latent protection against later cue-driven relapse, motivating reconsolidation-blocker and engram-stabilization pharmacological strategies. Then a delirium repurposing paper from Aoyama/Shinozaki (Stanford) — zuranolone, the FDA-approved GABA-A PAM neuroactive steroid for postpartum depression, dose-dependently attenuates BSEEG slowing in LPS systemic-inflammation and postoperative-delirium mouse models across young/aged/super-aged groups (with prophylactic dosing improving super-aged survival after POD induction), and reduces Iba-1/CD68 microglial activation markers at 24h post-insult — a neuroimmune-active GABA-A PAM signature with a clear regulatory path into postoperative delirium prophylaxis. Closing with a computational-pharmacology highlight from Moeller/Nygaard (Gubra and consortium) — an end-to-end AlphaFold2-hallucination ColabDesign pipeline generates >5,000 linear and head-to-tail cyclic MC4R agonist candidates, with 74% of linear and 23% of cyclic peptides showing measurable activity in functional screening; identifies a cyclic agonist with EC50 of 340 nM lacking the canonical His-Phe-Arg-Trp activation motif; deep mutational scanning identifies an alternative APWR activation motif and a single P5 substitution (E5P variant) reaches 6.7 nM hMC4R potency; central administration of E5P (10 nmol) reduces acute food intake in mice — proof that the de novo pipeline produces real pharmacology and that the canonical melanocortin pharmacophore is one solution rather than the only one. Production note: arXiv was 429-blocked from this IP throughout the run (continued cooldown from the upstream biomedical-agentic-ai run); MC4R cyclic-peptide design item covers the de novo computational drug-design tier. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-22-receptor-and-reason Fri, 22 May 2026 12:00:00 +0000 1276 Daily roundup for May 22, 2026: Womelsdorf/Pawliszyn (Waterloo) miniaturized solid-phase-microextraction probes simultaneously sample dopamine/acetylcholine/serotonin/glutamate/GABA in NHP PFC and striatum during cognitive set-shifting — area-specific multi-neurochemical fingerprints (GABA+ACh predict PFC engagement, DA+ACh predict striatal engagement, glutamate-ACh co-modulation, common-raphe serotonin drive across both regions); Haber UW-Madison acute DOI on rat cortical MEAs drives hyperexcitability and rate-corrected STTC path-length shortening with a ketanserin-resistant topology component — psychedelic network rewiring survives the firing-rate-confound correction; Melzer MedUni Vienna chemogenetic activation shows astrocytic Gi-coupled (not Gs/Gq) GPCR signaling in temporal cortex enhances fear memory retrieval and attenuates cue-evoked astrocyte Ca2+ — α2-adrenergic/CB1/kappa-opioid/several 5-HT subtypes may operate substantially through astrocyte Gi; Zoghbi Baylor astrocytic ACSBG1 KO in Thy1-α-Syn PD mice rescues behavior and reduces pS129 α-syn via reduced sphingosine-20:1 cytokine signaling — astrocyte lipid metabolism as a PD therapeutic surface; Huang Brown UBE3A→ERβ axis in iPSC-derived OPCs — UBE3A sustains ERβ protein, loss collapses OPC self-renewal in Angelman syndrome, selective ERβ agonists restore self-renewal/myelination/behavior in patient iPSCs and mouse models; Rudolph Illinois weak Complex-I inhibitor CP2 reverses Y-maze/startle/social-novelty deficits in 4×Gldc schizophrenia mice via PGC-1α/VDAC1 restoration — mtCI partial inhibition repurposed from neurodegeneration into schizophrenia; Nguyen Baylor semaglutide (0.1 mg/kg) dose-dependently reduces fentanyl IVSA rewards and PR breakpoint in female Wistar rats under short- and long-access, attenuates withdrawal in short-access only — clean female-only-design preclinical data point bridging the GLP-1-for-AUD review into opioids; Grella Loyola TetTag dDG engram reactivation — first-exposure (not fourth-exposure) cocaine engram attenuates cocaine-prime reinstatement, with saline-engram protection in females only, motivating reconsolidation-blocker/engram-stabilization strategies; Shinozaki Stanford zuranolone (GABA-A PAM neuroactive steroid, FDA-approved for PPD) mitigates BSEEG delirium-slowing and microglial activation across young/aged/super-aged LPS and POD mouse models with improved super-aged survival — neuroimmune-active GABA-A PAM signature; closing with Gubra AlphaFold2-hallucination de novo cyclic MC4R agonist pipeline — 74%/23% linear/cyclic functional hits, novel APWR motif, deep-mutational-scanning-matured E5P variant at 6.7 nM hMC4R, central administration reduces mouse food intake. Receptor & Reason for May 21, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate is centered on causal human neuromodulation and the modalities that surround it. Opening with the rarest kind of pharmacological evidence — Yebra/Rutishauser (Cedars-Sinai + Caltech + Universidad Politecnica de Madrid) focal microstimulation of human substantia nigra pars compacta in awake PD DBS surgery during a gambling task with interleaved happiness ratings — stimulation at outcome reduces risk-seeking specifically in the loss domain, improves choice efficiency, increases cumulative earnings, and selectively scales outcome-related happiness with decision value and reward prediction error (no tonic mood elevation), with computational modeling localizing the change to loss-utility weighting while gain processing is intact. Then the Anikeeva-lab (MIT) therapeutic demonstration of magnetoelectric nanodiscs (MENDs) for subthalamic-nucleus DBS in a PD mouse model — injectable nanoscale discs convert weak external magnetic fields into local electric polarization, motor improvements match clinical electrode-based DBS via a new GaitPattern gait-feature pipeline, with substantially reduced local inflammation and reduced brain oxidative stress (the latter hinting at a disease-modifying signature beyond symptomatic benefit) — a credible challenger DBS modality. Then a Psychiatric Times pipeline update — Synendos Therapeutics SYT-510, a first-in-class selective endocannabinoid reuptake inhibitor (SERI), enters Phase 2 dosing in DSM-5 generalized anxiety disorder after a 60-volunteer Phase 1 showed CNS exposure at pharmacological levels and EEG readouts consistent with established anxiolytics — context-conditional endocannabinoid amplification as the indirect strategy after direct CB1 agonism failed for anxiety; the endocannabinoid axis becomes a real therapeutic surface if Phase 2 efficacy holds. Then the Psychiatric Times GLP-1-for-AUD review consolidating the past year's evidence — NIH-led RCT in 108 AUD + obesity patients with 26 weeks of weekly semaglutide on top of CBT reducing heavy-drinking days, lab self-administration, weekly craving, and cigarettes-per-day; a 120M+ patient EHR trial emulation showing tirzepatide-vs-DPP-4 association with reduced AUD incidence; mesolimbic GLP-1 receptors as the mechanism converging on VTA/NAc reward modulation. Then the Vaidya group (TIFR Mumbai) reverse-genetic early-adversity model — postnatal hM3Dq chemogenetic activation of CaMKII+ forebrain excitatory neurons produces durable adult anxiety-like behavior plus bidirectional transcriptional regulation of GFAP, Aldh1l1, S100β, Eaat1/Eaat2/Eaat3 (down in neocortex, up in hippocampus) with no astrocyte density change, and 1H-(13C) NMR showing reduced astrocyte-specific glutamate/GABA turnover in adults — astrocyte transcriptional reprogramming as durable consequence and candidate intervention surface. Then the Georgakopoulos/Robakis Mt Sinai mechanism paper explaining γ-secretase trial failure — ADAM17 cleaves extracellular VEGFR2 to produce membrane-embedded substrate VCTF1, normally cleared by γ-secretase; PS1 FAD mutants and pharmacological γ-secretase inhibitors block this clearance, VCTF1 accumulates, and accumulated VCTF1 binds full-length VEGFR2 monomers preventing productive dimerization, suppressing brain angiogenesis and neuroprotection — γ-secretase inhibition actively suppressed an angiogenic program FAD patients are already losing, suggesting ADAM17 inhibition or VCTF1-monomer-interaction blockade as the corrective strategy. Then a methodology audit — Donald lab (Duke) 3,255 black-box AlphaFold 3 predictions of D-peptide binders in heterochiral complexes with explicit D-stereocenter inputs find a 51% chirality-violation rate (11× the headline number), random-chance accuracy in residue chirality assignment, systematically wrong folds and binding poses, no rescue from increased seeds, and confidence metrics that fail to distinguish correct from chirality-violating predictions — AlphaFold 3 has a structural blind spot in heterochiral systems and is not the workflow for D-peptide design today. Then the cleanest demonstration this year of why the Model Context Protocol is the right abstraction for life-science data infrastructure — Yamamoto-lab (DBCLS Japan) TogoMCP system uses LLMs as protocol-driven inference engines over the DBCLS RDF Portal (60 aggregated life-science databases) with MIE YAML files dynamically supplying schema/semantic context and a two-stage entity-resolution-then-SPARQL workflow, achieving Cohen's d = 1.82 over an unaided baseline on 50 biologically-grounded questions across 23 databases — externalize schemas into protocol-accessible context, compose specialized tools through a standard interface. Then the Bittencourt lab (USP São Paulo) atlas-style neuropeptide-pharmacology contribution — integrative scRNA reanalysis + FACS-sorted GAD67-GFP single-cell profiling identifies two transcriptionally distinct Pmch-expressing GABAergic populations in the medial preoptic area that emerge during mid-to-late lactation co-expressing prodynorphin (kappa-opioid ligand precursor), islet amyloid polypeptide, and prolactin receptor — pharmacological implications for MCHR1 antagonists in postpartum mood/maternal behavior and for kappa-opioid antagonists in postpartum depression. Closing on a tool-compound paper for an underserved indication — Zegarra-Valdivia/Torres Aleman (Achucarro Basque Center) AIK3a305 IGF-1 sensitizer in mild controlled-impact mouse TBI normalizes sensorimotor (adhesive-removal), neurological severity score, Y-maze cognition, and anxiety-like behavior; increases circulating IGF-1 via sensitization feedback; holds IL-6 at control levels; bulk transcriptomics shows dampened bidirectional injury-response gene programs rather than a new program — the right pharmacological signature for a sensitizer, opening sensitizers as a class for protein-deficiency-of-response neurological indications. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-21-receptor-and-reason Thu, 21 May 2026 12:00:00 +0000 1275 Daily roundup for May 21, 2026: Yebra/Rutishauser focal microstimulation of human substantia nigra pars compacta in awake PD DBS surgery causally reshapes loss processing and outcome-locked happiness (no tonic mood change) during a gambling task, with computational modeling localizing the effect to loss-utility weighting; Anikeeva MIT magnetoelectric nanodiscs (MENDs) for STN-DBS match clinical electrode-based motor recovery in a PD mouse via a GaitPattern computational pipeline, with reduced inflammation and oxidative stress hinting at disease-modifying potential beyond symptomatic benefit; Synendos SYT-510 first-in-class selective endocannabinoid reuptake inhibitor (SERI) enters Phase 2 DSM-5 GAD dosing after Phase 1 EEG anxiolytic-like readouts, opening context-conditional endocannabinoid amplification as the indirect path after direct CB1 agonism failed; Psychiatric Times GLP-1-for-AUD review consolidates the NIH semaglutide RCT (108 patients, reduced heavy-drinking days/craving/cigarettes), the 120M+ patient tirzepatide trial-emulation signal vs DPP-4 inhibitors, and mesolimbic VTA/NAc GLP-1R as mechanism; Vaidya TIFR reverse-genetic early-adversity model — postnatal hM3Dq Gq activation of CaMKII+ forebrain excitatory neurons drives durable adult astrocyte transcriptional reprogramming (GFAP/Aldh1l1/S100β/Eaat1-3) and reduced astrocyte glutamate/GABA turnover by NMR; Georgakopoulos Mt Sinai explains γ-secretase Alzheimer trial failure — PS1 FAD mutants and γ-secretase inhibitors accumulate ADAM17-derived VCTF1 fragment of VEGFR2 that blocks VEGFR2 monomer dimerization and suppresses brain angiogenesis/neuroprotection, suggesting ADAM17 inhibition or VCTF1-monomer blockade as corrective; Donald Duke AlphaFold 3 evaluation finds 51% D-peptide chirality violation (11× headline rate), random-chance residue accuracy, systematically wrong folds/poses in heterochiral complexes, and useless confidence metrics — AF3 is not the D-peptide design workflow; Yamamoto DBCLS TogoMCP demonstrates schema-guided LLM + MCP architecture for natural-language SPARQL over 60 life-science databases with Cohen's d = 1.82 improvement on 50 benchmark questions; Bittencourt USP identifies two Pmch-expressing GABAergic populations in lactating-dam MPOA co-expressing prodynorphin/IAPP/prolactin receptor with pharmacological implications for MCHR1 and kappa-opioid antagonists in postpartum disorders; closing with Zegarra-Valdivia/Torres Aleman AIK3a305 IGF-1 sensitizer reversing sensorimotor/cognitive/anxiety deficits in mouse mild TBI, opening sensitizers as a pharmacological class for protein-deficiency-of-response neurological indications. Receptor & Reason for May 20, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate clusters on the neurodegeneration-and-inflammation axis with three coupled methodology threads (ligand-optimization controls, GPCR-atlas validation, protein-language-model engineering). Opening with the Phase 2 MINDFul readout on XPro1595 (pegipanermin), a dominant-negative soluble-TNF biologic for inflammation-defined early Alzheimer's — trends toward cognitive and biomarker improvement in the biomarker-enriched population, with strategic implications for whether the field's amyloid-PET enrichment template gains a parallel inflammatory-biomarker arm. Then the Zoghbi lab's GLP-1 mechanism paper in SncaG51D knock-in Parkinson's mice — transcriptomic and proteomic profiling shows a GLP-1 analog partially restores inflammatory, mitochondrial, and intercellular signaling networks toward wild-type, anchoring the lixisenatide/exenatide Phase 2 motor signals in a genetic-model mechanism and setting up the LRRK2/GBA-model generalization test. Then the Groppa group's single-cell PD work in mouse plus post-mortem human brain identifying oligodendrocyte transcriptional subtypes that track motor progression rate rather than just disease presence — Parkinson's as a glial disease with neuronal manifestations, with therapeutic implications. Then the Sheng/Stevens-affiliated Broad team's GPR34 knockout work showing this lysophosphatidylserine-responsive microglial GPCR restrains the disease-associated-microglia transition in an AD amyloid model — a druggable, selectively-microglial-enriched target with biased-agonist/PAM tractability. Then a Shoichet-Roth-Manglik-Fraser-Basbaum consortium on random property-matched backgrounds for ligand optimization across opioid, cannabinoid, and other GPCR campaigns — finding a substantial fraction of optimization "gains" are statistically indistinguishable from drawing random property-matched compounds, with major implications for how computational pharmacology campaigns report and validate progress. Then the Steele/Leinninger group's methodological pushback on dopamine-neuron GPCR atlas assignments — Nmur1 and Cckar Cre lines fail to provide functional adult-DA-neuron genetic access, with broader implications for the transcriptomic-to-driver-line gap as GPCR atlases proliferate. Then the Hérault group's long-horizon Down syndrome prenatal-pharmacology paper showing transient prenatal DYRK1A inhibition with Leucettine L41 produces durable postnatal cognitive and neurodevelopmental improvements, motivating window-of-treatment thinking for next-generation DYRK1A inhibitor programs. Then a counterintuitive Yassa/Bornstein clinical-psychology finding — anhedonia buffers (rather than compounds) the effects of early-life unpredictability on adult threat-reward decision-making, with implications for which symptoms fast-acting antidepressants should be targeting in patients with adverse-childhood-experience histories. Then a Boyden/Forest team's protein-language-model-suggested mutation — ChrimsonR-E300, identified directly from sequence context without structural modeling or screening — that amplifies optogenetic photocurrent response, demonstrating that PLM-driven single-residue engineering is starting to materially improve real working tools. Closing on topological pharmacokinetics — a methodology provocation arguing that scalar PK summaries (AUC, Cmax, clearance, t1/2) discard mechanistically informative shape information in concentration-time curves, with topological data analysis revisiting a fifty-year-old summary-statistic question now that dense PK sampling is routine. Production note: arXiv was 429-blocked from this IP throughout the run; today's slate is bioRxiv- and Psychiatric Times-driven. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-20-receptor-and-reason Wed, 20 May 2026 12:00:00 +0000 977 Daily roundup for May 20, 2026: Phase 2 MINDFul readout on XPro1595 (pegipanermin) — dominant-negative soluble-TNF biologic trends toward cognitive/biomarker improvement in inflammation-defined early AD, with implications for whether the amyloid-PET enrichment template gains a parallel inflammatory-biomarker arm; Zoghbi-lab GLP-1 analog partially restores inflammatory/mitochondrial/intercellular signaling networks in SncaG51D Parkinson's mice, anchoring lixisenatide/exenatide Phase 2 motor signals in a genetic-model mechanism; Groppa-group single-cell work in mouse + human post-mortem PD brain identifies oligodendrocyte transcriptional subtypes tracking motor progression rate (not just disease presence); Sheng/Stevens-affiliated Broad team's GPR34 KO crossed into AD amyloid model — this lysophosphatidylserine-responsive microglial GPCR restrains the disease-associated-microglia transition, a druggable microglia-selective target; Shoichet-Roth-Manglik-Fraser-Basbaum consortium proposes random property-matched backgrounds for ligand optimization, finding a substantial fraction of campaign "gains" are indistinguishable from drawing random property-matched compounds; Steele/Leinninger group's methodological rebuttal — Nmur1 and Cckar Cre lines fail to provide functional adult-DA-neuron access, exposing the transcriptomic-to-driver-line gap in GPCR atlases; Hérault-group Down syndrome paper — transient prenatal Leucettine L41 DYRK1A inhibition produces durable postnatal cognitive improvements, supporting window-of-treatment programs; Yassa/Bornstein counterintuitive finding — anhedonia buffers (not compounds) early-life-unpredictability effects on adult threat-reward decision-making, with implications for fast-acting antidepressants in ACE-history patients; Boyden/Forest team's protein-language-model-suggested ChrimsonR-E300 mutation amplifies optogenetic photocurrent without structural modeling; closing on topological pharmacokinetics — revisiting scalar PK summaries (AUC, Cmax, clearance) with topological data analysis, arguing concentration-time-curve shape carries mechanistic information current summaries discard. Receptor & Reason for May 19, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's slate has a dopamine center of gravity — psychedelic-driven, addiction, computational, and Parkinsonian — plus pharmacoepidemiology, a deuterated psilocin clinical update, a tRNA-fragment angle on Alzheimer's tau, and a process-level agentic-LLM benchmark. Opening with the Foldi lab's psilocybin striatal-dopamine work in female rats — in vivo fiber photometry shows psilocybin broadly amplifies nucleus accumbens dopamine transients time-locked to expected and unexpected outcomes, computational modeling identifies psilocybin-specific increases in learning rate and reductions in prior value weighting, and the translational hook is that calorie restriction and prior activity-based-anorexia exposure both moderate the pro-cognitive effects — likely reflecting reductions in cortical 5-HT2A availability, with direct implications for psilocybin anorexia trial designs. Then the Dong lab's clean astrocyte-versus-neuron dissociation on accumbens-shell D1 dopamine signaling in cocaine self-administration — selective knockdown of astrocytic (not neuronal) D1 inhibits silent-synapse generation, expedites extinction, and reduces cue-induced reinstatement, putting astrocytic D1 forward as a potentially more druggable handle for relapse prevention than the neuronal pool. Then the George lab's heterogeneous-stock-rat oxycodone self-administration showing prolonged withdrawal-induced allodynia for up to three weeks, with addiction-index severity prospectively predicting allodynia magnitude and duration — a prospective rather than reactive stratification biomarker that could feed buprenorphine/methadone induction protocols. Then the Paglini lab's p35-knockout ADHD model with acute methylphenidate, fluoxetine, and combined treatment normalizing region- and sex-dependent prefrontal-hippocampal c-Fos patterns — the right preclinical setup to probe the dopaminergic-serotonergic combination commonly prescribed for ADHD-depression comorbidity. Then a clinical update on Helus Pharma's HPL-003, a deuterated psilocin designed for narrower pharmacokinetic distribution and more manageable in-clinic session duration, with adjunctive Phase 1/2a benefit reported in inadequate responders without SSRI washout — paired with the broader federal psychedelic-research funding context. Then the Lancet Psychiatry meta-analysis of 37 studies, 600,000+ exposed pregnancies and ~25M unexposed — naive 35%/69% increases in ADHD/autism attenuate to null after adjustment for parental mental health and genetics, with paternal antidepressant use as a negative control showing the same direction of effect, supporting confounding rather than teratogenicity. Then a Harvard simultaneous-recording paper in olfactory tubercle and VTA showing trial-by-trial changes in striatal activity are more consistent with dopamine-driven representation learning than with frozen-representation value updating — convergence with machine-learning architectures that update features, not just weights. Then a Politecnico di Milano connectome-constrained multiscale mouse-brain model showing focal subcortical dopamine depletion is sufficient to produce widespread beta hypersynchrony via closed cortico-basal-ganglia-thalamic loops, with virtual loop ablations confining beta to subcortical generators when the loop is severed — a theoretical foundation for adaptive DBS using cortical beta as a control signal. Then the Krichevsky lab identifying stress-induced tRNA-derived fragments — specifically 5'GluCTC — as the most dysregulated small-RNA class in human Alzheimer's brains, with the fragment inducing tau S396 phosphorylation, oligomerization, and impaired neurite growth, plus extracellular secretion implicating it in cell-to-cell pathology spread. Closing on the agentic-AI side: BiomniBench-DA from Phylo — 100 expert-co-developed data-analysis tasks across 17 task types and five disease areas, scoring full agent trajectories against task-specific rubrics, finding frontier and open-weight models cluster within a few points, agent harness shifts scores more than successive model generations, and agents reliably ground claims in real sources but consistently fall short on method selection, biological interpretation, and scientific reasoning. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-19-receptor-and-reason Tue, 19 May 2026 12:00:00 +0000 1039 Daily roundup for May 19, 2026: Foldi-lab psilocybin amplifies accumbens dopamine transients and shifts learning-rate/prior-value parameters, with calorie restriction and prior activity-based anorexia attenuating pro-cognitive effects via reduced cortical 5-HT2A availability — direct implications for psilocybin-for-anorexia trial designs; Dong-lab astrocytic (not neuronal) D1 in accumbens shell is required for cocaine-induced silent synapses and reinstatement; George-lab HS-rat oxycodone study finds addiction-index severity prospectively predicts protracted withdrawal allodynia (up to 3 weeks); p35-KO ADHD model responds to acute methylphenidate/fluoxetine with sex-dependent prefrontal-hippocampal c-Fos normalization; HPL-003 deuterated psilocin Phase 1/2a adjunctive without SSRI washout, plus federal psychedelic funding context; Lancet Psychiatry meta-analysis (600K+ exposed pregnancies, ~25M unexposed) finds no causal antidepressant-pregnancy link to ASD/ADHD after adjustment, with paternal antidepressant negative control supporting confounding; Harvard olfactory-tubercle/VTA recordings support dopamine-driven representation learning over frozen-representation value updating; Politecnico di Milano multiscale connectome PD model shows focal subcortical DA depletion drives brain-wide beta via cortico-basal-ganglia-thalamic loop resonance; Krichevsky-lab 5'GluCTC tRNA fragment induces tau S396 phosphorylation, oligomerization, and propagation in AD; and BiomniBench-DA process-level agent evaluation shows harness matters more than model generation, with agents weak on method selection and biological interpretation. Receptor & Reason for May 18, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today's theme is receptor pharmacology at three different scales (psychiatric genetics, cancer neuroscience, deep-brain stimulation) plus a coherent skeptical-but-constructive arc on agentic AI in drug discovery. Opening: a spatial transcriptomics paper introducing the Gene Fraction Enrichment Score (GFES) statistic that controls for neuronal density — and finding ASD exome-sequencing risk genes most enriched in thalamus, schizophrenia GWAS genes in thalamus + hippocampus + cortex, which moves the thalamus from "creeping into both literatures" to "genuinely there at the regional level with neuron-density correction." Then a human-organoid tumor transplantation model (so-HOTT) showing AMPA/kainate-receptor inhibition shifts glioblastoma cells away from neuronal-like fates toward progenitor-proximal astrocytic/mesenchymal states via attenuated calcium signaling and reduced radial-glia plasticity — the cleanest mechanistic frame for perampanel's emerging adjunctive-GBM signal, and a case for actually developing kainate-receptor chemistry. Then combined fMRI + cerebellar TMS in humans establishing a causal role for cerebellum in working memory: cerebellar perturbation broadly degrades cortical spatial tuning and impairs recall at magnitudes matching frontoparietal perturbation — the cerebellum is doing the cognitive work, not just reflecting it, with implications for mGluR/GABAergic drug profiles previously dismissed as motor-side-effect risks. Then a sobering preregistered multilevel meta-analysis of 166 PD postmortem studies across 145 region-cell populations — only 4 are adequately powered, 82% of Allen Brain Atlas regions have never been quantified in PD, and where the data exist, locus coeruleus noradrenergic loss matches substantia nigra dopaminergic loss (>60% in both), supporting parallel-cell-population PD therapeutics rather than dopamine-centric ones. Then HaloNeu, a near-infrared-II chemo-optogenetic construct fusing cpHaloTag-TRPV1 with photothermal nanotransducers — non-invasive deep-brain stimulation at 1 cm with 60 mW/cm² and up to 5 cm under the standard skin safe-limit (~1 W/cm² at 1064 nm), >2 months stable in vivo, sustained VTA modulation and rescue of Parkinsonian motor symptoms in mice — the closest thing to a real challenger to electrode-based DBS I've seen at the bench. Then concerning neonatal opioid pharmacology: acute buprenorphine produces frank hypoventilation in P4-5 rat pups during eupnea and hypoxic-hypercapnic challenge, and neither standard naloxone hydrochloride nor peripherally-restricted naloxone methiodide fully reverses it — implying buprenorphine's neonatal respiratory effects extend beyond canonical mu opioid pharmacology, with practical implications for buprenorphine in NOWS management and rescue protocols. Then the agentic-AI methods arc: an open-LLM benchmark where Claude Opus authors variant-calling plans of increasing detail and six 2026-release open-weight implementers execute on desktop GPUs — qwen3.6:27b reproduces frontier accuracy on every plan and matches Opus cell-for-cell on a 36-cell error-injection matrix, runnable on a sub-$2k Jetson or Mac Mini, validating the plan-execute architectural split for agentic biomedical workflows. Then a comprehensive 156-fold-mean benchmark across ADME, toxicity, and bioactivity endpoints with random, Murcko-scaffold, and structure-separated splits — RF(ECFP4) and ExtraTrees(RDKit) win 116, GNNs (GIN, Ligandformer) win 25, pretrained sequence models (MoLFormer, ChemBERTa2) win 12, LLM-based SAR baselines win 3 — a corrective to scale-centered drug-discovery narratives. Then a clean ablation paper showing 7 recent deep-learning chemical-perturbation-prediction models with dedicated drug molecular encoders perform identically under cold-drug evaluation when drug input is zeroed or shuffled, and an MLP using only cell-line basal expression matches all of them — current architectures aren't using drug molecular features for generalization. Closing constructively: ToolMol, a multi-objective genetic algorithm with an agentic LLM operator using a curated RDKit toolbox of medicinal-chemistry transformations, achieving >10% stronger predicted binding affinity and 35% higher absolute binding free energy than existing methods on three protein targets — chain-of-thought traces show tool calling makes the LLM faithfully execute its planned modifications, with the LLM as planner-and-orchestrator on top of mature cheminformatics being the productive frame. Production note: arXiv was rate-blocked from this IP throughout the morning, again — today's only arXiv pick (ToolMol) was pulled via web search; Psychiatric Times had no fresh items in the 7-day window not already covered. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-18-receptor-and-reason Mon, 18 May 2026 12:00:00 +0000 1056 Daily roundup for May 18, 2026: spatial transcriptomics with neuron-density-corrected GFES implicates thalamus/cortex for ASD and schizophrenia; AMPA/kainate inhibition shifts glioblastoma cell fate via radial-glia plasticity in human organoid model; cerebellar TMS+fMRI establishes causal role of cerebellum in working memory at frontoparietal-equivalent magnitude; preregistered PD neuropathology meta-analysis finds locus coeruleus loss matches substantia nigra; HaloNeu cpHaloTag-TRPV1 + NIR-II photothermal nanotransducers enable non-invasive ~5 cm DBS with sustained VTA modulation; acute buprenorphine produces naloxone-refractory hypoventilation in neonatal rats; open qwen3.6:27b matches Claude Opus as agentic implementer for variant-calling plans on a sub-$2k Jetson; RF(ECFP4)/ExtraTrees(RDKit) beat GNNs, MoLFormer, ChemBERTa2, and LLMs in 156-fold drug-discovery benchmark; L1000 drug-encoder ablations go negligible (drug encoders aren't being used); and ToolMol combines a multi-objective genetic algorithm with an agentic-LLM operator and RDKit toolbox for >10% binding affinity gains. Note: arXiv was rate-blocked this run. Receptor & Reason for May 17, 2026 — your daily neuropsychopharmacology, computational pharmacology, and agentic-AI-in-biology roundup, curated for Alan Huebschen. Today's lead is the largest reward-imaging consortium pharmacology study to date — over 1,250 participants across five cohorts including IMAGEN and an Apulian psychosis sample — identifying two reproducible imaging phenotypes mirroring sign- and goal-tracking, with single- and repeated-dose D2/D3 antagonism (risperidone, haloperidol, amisulpride) selectively knocking down anticipatory ventral-striatal signaling in sign-tracker-like individuals (with parallel drops in self-reported energy), while aripiprazole partial agonism remaps goal-tracker responses; in the psychosis cohort, antipsychotic D2 affinity tracks blunted anticipatory signaling and higher negative symptoms — the closest thing to a usable stratification biomarker for antipsychotic response that reward imaging has produced. Then a clean causal MEG-pharmacology test from Düsseldorf showing GABAergic potentiation lengthens intrinsic cortical timescales across regions (strongest in frontal default mode and dorsal attention), shifts network occupancy probabilities, while NMDA manipulation produces no significant timescale effects — an asymmetry that should make us re-examine the NMDA-hypofunction-sets-timescales framing of schizophrenia. A pharmacology-of-psychosis-models paper using clinically-realistic LPS and restraint-stress doses suppresses methamphetamine sensitization via TLR4-dependent but mechanistically divergent COX-2 (peripheral) and microglia-derived TNF-α (psychological-stress) pathways — implying COX-2 inhibitors and TNF modulators aren't interchangeable as anti-inflammatory adjuncts in psychosis. Then a cell-type-resolved AD mechanism — long-term depression selectively impaired in dopamine D1R+ medium spiny neurons of the nucleus accumbens in APP/PS1 mice via mGluR1/5 dysfunction and persistent calcium-permeable AMPAR signaling, with CP-AMPAR blockade restoring LTD and a behavioral phenotype of palatable-reward overconsumption that maps onto early-AD neuropsychiatric symptoms. Then a presynaptic mechanism for the alcohol-impairs-speech-in-noise phenomenon — low-dose ethanol potentiates GABA-A currents specifically at the presynaptic endbulb of Held terminal, where high intra-terminal chloride drives a depolarizing GABA-A response activating low-threshold Kv1 channels and shunting the AP, limiting calcium influx and glutamate release. Neuromodulation pair: ALIC fiber topography is preserved in 18 TRD-OCD and 5 TRD-depression patients (diffusion tractography plus cerebro-cerebral evoked potentials), supporting individualized tractography for DBS contact selection; and the spatial-molecular para-subthalamic nucleus mapping shows para-STN wraps around STN and is partly embedded in the medial forebrain bundle, with an HTR2C gradient — geometry that engages all three major DBS targets simultaneously and opens a 5-HT2C-aware DBS-plus-pharmacology design space. Comp-pharm block: SIGMA-KG, a signed multi-omics knowledge atlas with FLASH (Fast Lightweight Architecture for Signed Heterogeneous GNN) as foundation model, preserving activating-versus-inhibitory polarity that unsigned graph baselines lose, with claimed 69.6% external clinical validation in inductive drug repurposing across four diseases. And from the Baker lab, de novo miniprotein agonists at MRGPRX1, NK1R, and CCR5 and antagonists at CXCR4, CCR5, OXTR, GLP-1R, GIP-R, GCGR, PTH-1R, and CGRPR, with cryo-EM atomic-level agreement between designed and experimental structures — opening designed miniprotein GPCR drug discovery as a real modality, with obvious extensions to 5-HT2A, NK1R, and OXTR for neuropsychiatric indications. Closing on the most clinically suggestive item — Johns Hopkins evaluation of BCG vaccination in the PS19 tauopathy mouse model, reproducing the bladder-cancer-cohort retrospective AD-prevention signal: hippocampal phospho-tau and microgliosis down, glutamate-weighted CEST-MRI up, hippocampal transcriptome normalized toward wild-type, peripheral myeloid CD80+ recruitment with enhanced tau-fibril phagocytosis, and improved novel-object recognition. Production note: arXiv was rate-blocked from this IP again this morning, so the agentic-AI side is thinner and Psychiatric Times had no fresh items in the window. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-17-receptor-and-reason Sun, 17 May 2026 12:00:00 +0000 892 Daily roundup for May 17, 2026: 1,252-participant reward-imaging consortium identifies sign- vs goal-tracker phenotypes with differential D2 antagonist response (a candidate antipsychotic stratification biomarker); GABA lengthens intrinsic cortical timescales while NMDA does not in healthy MEG; LPS and restraint stress both suppress methamphetamine sensitization via TLR4 but diverge into COX-2 vs TNF-α; LTD selectively lost in accumbal D1R+ MSNs in early AD via calcium-permeable AMPARs; ethanol shunts the endbulb of Held via presynaptic GABA-A and Kv1; ALIC fiber topography preserved in treatment-resistant OCD and depression; para-STN spatially wraps STN and MFB with an HTR2C gradient across DBS targets; SIGMA-KG sign-aware multi-omics drug-action foundation model with 69.6% external clinical validation; Baker-lab de novo miniprotein agonists/antagonists across 11 GPCRs; and BCG vaccination mitigates tau pathology in PS19 mice. Note: arXiv was rate-blocked this run; Tier 2 thinner today. Receptor & Reason Saturday bonus for May 16, 2026 — a toolbox-themed second drop after the daily roundup, focused on drug-target methodology and pharmacology tooling. Ten fresh items, all distinct from this morning's episode, all from the last week. Opening with a Rio Grande do Sul systematic review and meta-analysis of 264 controlled in-vivo studies of NMDA-receptor antagonists (dizocilpine, ketamine, phencyclidine) on rodent social behavior — confirming the social-withdrawal signal that underwrites the schizophrenia negative-symptom model, but with small-study effects in the social-preference literature and antipsychotic "rescue" commonly co-occurring with locomotor suppression. Then an IGBMC Strasbourg paper showing α5IA, a negative allosteric modulator at α5-containing GABA-A receptors that had rescued cognition in Ts65Dn, fails to restore Barnes-maze or pattern-dissociation performance in the more anatomically accurate Dp(16)1Yey Down syndrome mouse model — possibly explaining the disappointing α5-NAM clinical trials. Then a McGill hiPSC-derived neuron study (Khayachi/Milnerwood/Rouleau) identifying 10 proteins and 13 microRNAs in extracellular vesicle cargo that distinguish lithium responders from non-responders in bipolar disorder, with the bipolar neuronal secretome functionally altering activity in non-bipolar networks — a candidate liquid-biopsy biomarker for personalized lithium prescribing. Then a Max Planck Leipzig TMS-dosing paper combining resting-state fMRI, finite-element E-field simulation, and ECG-derived heart-brain coupling to argue that cortical electric-field dose plus DLPFC-to-sgACC connectivity explains far more variance in autonomic engagement than motor-threshold dosing — actionable for the field-and-connectivity-informed TMS targeting question. Two PET-tracer papers: from Emory, fluorine-18 P4B-2412 as a selective PDE4B radioligand for neuroinflammation imaging beyond TSPO; and from Rigshospitalet Copenhagen, tritiated JNJ-64413739 binding to P2X7R after intrahippocampal kainic acid in rats, showing delayed and region-progressive P2X7R upregulation peaking at day 30 — a candidate epileptogenesis-window biomarker. Then a Weill Cornell (Gabr) phage-display campaign identifying CAP1, a disulfide-constrained cyclic peptide ligand for the CAPON/NOS1AP adaptor protein, attenuating Aβ42 toxicity, NMDA-driven NO production, and pathological tau phosphorylation with favorable preliminary PK — the first credible chemical probe for a long-standing AD target. Then a KU Leuven Talavera-lab paper showing five clinically used statins activate human and mouse TRPA1 via two distinct putative pore-domain binding sites — non-electrophilic agonist sites separable from the canonical N-terminal cysteine site — with mechanistic implications for statin-associated myalgia and a path to agonist-specific TRPA1 antagonists. Then a King's College London integrated zebrafish-to-mouse pipeline for manganese chelators in slc39a14 and SLC30A10 loss-of-function models, with the H3PyC3A scaffold outperforming clinical CaNa2EDTA for selective manganese mobilization — addressing the unmet need in inherited manganism. Closing on agentic AI: MechAInistic from Nebraska (Helikar), an LLM-guided multi-agent system pairing an Architect-Reviewer pattern with constraint-based genome-scale metabolic models, demonstrated on rheumatoid arthritis and multiple sclerosis paired single-cell metabolic profiles (proposing devimistat and ivosidenib as repurposing candidates) — the right division of labor between language-model orchestration and mechanistic modeling. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-16-receptor-and-reason-toolbox Sat, 16 May 2026 18:00:00 +0000 908 Saturday bonus toolbox episode: NMDA-antagonist social-behavior meta-analysis (264 studies), α5IA fails in the Dp(16)1Yey Down syndrome mouse model, hiPSC-derived neuronal EV cargo as lithium-response biomarkers in bipolar disorder, E-field + sgACC-connectivity argues for replacing motor-threshold TMS dosing, fluorine-18 P4B-2412 PDE4B PET radioligand and tritiated JNJ-64413739 P2X7R tracer in epileptogenesis, CAP1 cyclic peptide ligand for CAPON in Alzheimer's, statins activate TRPA1 via two pore-domain non-electrophilic sites, novel H3PyC3A manganese chelator outperforms CaNa2EDTA in inherited manganism models, and MechAInistic multi-agent LLM system for genome-scale metabolic modeling. Receptor & Reason for May 16, 2026 — your daily neuropsychopharmacology, computational pharmacology, and agentic-AI-in-biology roundup, curated for Alan Huebschen. Two threads dominate the day: developmental neuropsychopharmacology and cannabinoid system pharmacology. Opening: a Neuropsychopharmacology fetal-MRI study of 182 pregnancies (62 on SSRI/SNRI, 120 controls) finding smaller fetal hippocampal volumes, reduced cortical gyrification, curvedness, and surface area with prenatal serotonin reuptake inhibitor exposure — with EPDS stratification dissociating the exposure signal from depression severity, and placental volume and diffusion both increased in the exposed group. Cleanest in-utero structural evidence to date for SRI pharmacology on fetal brain development. Then a bioRxiv mechanism paper showing maternal high-fat diet reduces microglial phagocytosis of accumbens serotonin projections during a critical postnatal window, producing persistent 5-HT hyperinnervation, elevated NAc serotonin release, and accelerated reward-motivated learning specifically in male offspring — a microglia-centered route from maternal metabolism to adult mesolimbic function. Cannabinoid block: an NPP paper showing the THC fear-memory-reconsolidation-disrupting effect operates through dorsal-hippocampal microglia with sex-specific receptor pharmacology — CB1 plus PPARγ in males, CB1 alone in females. Then the methodologically strongest translational item this week — first in-vivo PET imaging of fatty acid amide hydrolase dynamics during short-term cannabis abstinence using [11C]CURB, showing ~10% whole-brain FAAH upregulation by day 3-7 (largest in ventral striatum), with a pharmacodynamic-biomarker handle that the FAAH-inhibitor-for-CUD trial program has lacked. Then a striking cross-modal paper combining fMRI, PET-derived receptor/transporter atlases, and structural connectomics to map the neurochemical architecture of the human language network — cortico-cortical serotonergic/glutamatergic, thalamo-cortical cholinergic, anterior temporal GABAergic — with disrupted profiles in 239 post-stroke aphasics, and a post-hoc reanalysis of four pharmacological trials showing treatment-to-disruption matching predicts recovery. Then a sex-dependent Grin1 knockout in ALDH1A1+ dopamine neurons (the Parkinson's-vulnerable subtype) producing female-specific post-restriction overfeeding via VTA — a substrate for compulsive-eating side effects of dopaminergic PD therapy. Methods: dual-color mesoscopic imaging of cortical norepinephrine with calcium or acetylcholine in mice, demonstrating state-dependent NE patterns, NE-ACh coupling, and sleep-deprivation disruption. Repurposing: simvastatin reverses Aβ-induced cognitive deficit in mice by reducing hippocampal HDAC2, restoring H4 acetylation at the BDNF promoter, and restoring BDNF — with viral HDAC2 overexpression abolishing benefit, suggesting an epigenetic rather than cholesterol-lowering mechanism for the statin-and-dementia signal. Clinical reframe: a Psychiatric Times synthesis on low-dose lithium making the case for a distinct pharmacological category from mood-stabilizer doses, tying together the dementia-prevention literature, suicide-prevention ecological data, and GSK3β/autophagy/circadian mechanisms. Closing on agentic AI: NeuroAtlas, the largest EEG benchmark to date (42 datasets, 260k hours) for foundation-model evaluation across epilepsy, sleep medicine, brain age, and BCIs — finding EEG-specific FMs do not consistently outperform generic time-series FMs, that standard ML metrics miss clinical utility, and that domain-specific rankings vary substantially. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-16-receptor-and-reason Sat, 16 May 2026 12:00:00 +0000 1007 Daily roundup for May 16, 2026: prenatal SSRI/SNRI fetal MRI shows altered hippocampal volume and cortical gyrification; maternal HFD remodels 5-HT reward circuits via microglia in male offspring; THC fear-reconsolidation-disrupting effect is microglial with sex-specific CB1/PPARγ pharmacology; first in-vivo PET of FAAH upregulation during cannabis abstinence; aphasia neurotransmitter-circuit mapping predicts pharmacological recovery; sex-dependent NMDA on ALDH1A1+ DA neurons and PD compulsive eating; mesoscale dual-color noradrenergic imaging; simvastatin rescues Aβ cognitive deficit via HDAC2-BDNF; low-dose lithium reframed as a distinct pharmacology; and NeuroAtlas EEG foundation-model benchmark. Receptor & Reason for May 15, 2026 — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Two threads carry the day: dopamine pharmacology getting reopened on multiple fronts, and a dense psychiatric pipeline week. Opening: a Neuropsychopharmacology paper using fast-scan voltammetry and fiber photometry shows acute fentanyl drives opposing changes in tonic versus phasic dopamine — sustained tonic elevation in the accumbens but suppressed phasic transients to reward-predictive cues, with the magnitude of phasic suppression predicting Pavlovian-conditioned-approach impairment. Reframes the opioid-versus-stimulant equivalence around computationally specific disruption of the reward-prediction-error channel while vigor is preserved or amplified. Then a bioRxiv mu/delta knockout dissection of tianeptine — confirming the mu requirement for spatial memory enhancement and hippocampal beta oscillations, while delta knockout has no effect; locomotor effects depend on both. A clean asymmetry that motivates safer biased mu agonist development. Molecular Psychiatry's "How muscle talks to brain" identifies apelin as a circulating peptide mediator of exercise-induced antidepressant effects, with apelin-receptor antagonism blocking the rescue and hippocampal AKT/mTOR signaling downstream — an exercise-mimetic pharmacology with an existing cardiometabolic apelin agonist already in clinic. Mol Psy nitrous oxide paper relocates the rapid-acting antidepressant action of N2O from generic NMDA antagonism to AMPA-receptor potentiation specifically at deep-cerebellar-nuclei terminals onto VTA dopamine neurons — site convergence with this week's earlier VGLUT2-trafficking depression circuit paper, structurally parallel to the ketamine rapid-AMPA framing at a different anatomical node. Then a Mol Psy adolescent-restricted Grin2a knockout in VTA dopamine neurons reproduces a schizophrenia-relevant phenotype with disrupted reward-prediction-error signaling — locking adolescent vulnerability, NMDA hypofunction, and dopamine RPE disruption into a single causal chain, with implications for prediction-error-restoring rather than dopamine-blockade interventions in early-course schizophrenia. Pipeline news: the first network meta-analysis of FDA-approved adjunctive MDD treatments places lumateperone first across all four efficacy outcomes (MADRS change, response, remission, CGI-S) versus aripiprazole, brexpiprazole, cariprazine, and quetiapine XR, with akathisia and weight signals at placebo levels — J&J-sponsored, NEI Spring Congress 2026, but quantitatively consistent with individual trials. FDA cleared NRX-101 (D-cycloserine/lurasidone fixed-dose combination) plus robotic-enabled TMS for treatment-resistant depression with suicidality — a three-arm 240-patient trial reframing TMS-plus-drug as a unified regulatory intervention. Translational method piece: focused-ultrasound-mediated transient BBB opening over primate perigenual ACC enables focal pgACC serotonin delivery in awake macaques, with the behavioral signature being reduced occupancy of high-motivation states and weakened reward-environment energization while immediate offer-value sensitivity is preserved — a route to causal regional pharmacology in primate cortex without chronic implants. Computational psychiatry piece: NPP paper using hierarchical Bayesian effort-discounting modeling in T2D patients on/off semaglutide finds elevated effort cost without altered reward sensitivity — opposite signature to what generalized anhedonia would predict, supporting a metabolically-driven motivational-specific account of GLP-1 mental health effects. Closing comp-pharm tool: Simpatico, a CLIP-style atomic-embedding virtual-screening method that preserves protein-pocket and ligand atom-level resolution rather than collapsing to single embeddings, with higher enrichment than DrugCLIP-class baselines at comparable speed. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-15-receptor-and-reason Fri, 15 May 2026 12:00:00 +0000 893 Daily roundup for May 15, 2026: acute fentanyl splits tonic vs phasic dopamine in Pavlovian learning, tianeptine's memory effect is mu- not delta-opioid, apelin as muscle-to-brain mediator of exercise antidepressant action, nitrous oxide enhances AMPA at DCN-to-VTA synapses, adolescent Grin2a-in-DA-neurons knockout produces schizophrenia-relevant prediction-error disruption, lumateperone tops first NMA of adjunctive MDD treatments, FDA clears NRX-101 + robotic-enabled TMS trial for TRD with suicidality, focused-ultrasound serotonin delivery into primate pgACC, computational effort-cost phenotyping on semaglutide, and Simpatico atomic-embedding virtual screening. Receptor & Reason for May 14, 2026 — your daily neuropsychopharmacology, computational pharmacology, and agentic-AI-in-biology roundup, curated for Alan Huebschen. The accidental theme today is dopamine pharmacology getting reopened at three different layers of organization in the same week. Opening: a bioRxiv biophysics preprint argues that the orphan glutamate delta receptors GluD1 and GluD2 are in fact ionotropic dopamine receptors — cryo-EM, lipid bilayer recordings, mutagenesis, and patch clamp showing dopamine drives clamshell closure and channel activation, with G-proteins acting as molecular switches (free Gβγ inhibits, Gα or inactive heterotrimers permit), uncoupled by a neurodegeneration-associated mutation. If reproducible, it forces a textbook rewrite — and there's a 2024 PNAS counter-claim arguing definitively against GluD ionotropic activity, which the field will need to resolve. Then a D3R-subtype paper identifying a D3-expressing subpopulation within NAc D1R neurons that defines the aversive component of quetiapine — a projection-target-resolved handle for D3-preferring antipsychotic chemistry like cariprazine. NTSR1 follows: SBI-553, the β-arrestin-biased allosteric modulator from Sanford Burnham, differentially regulating GABAergic transmission in central amygdala and BNST and suppressing motivated feeding — pushing the biased-agonism case for stimulant-use-disorder and binge-eating indications. Circuit pivot: chronic restraint stress drives activity-dependent VGLUT2 reduction at cerebellar deep-nuclei terminals onto VTA dopamine neurons, recasting the cerebellar contribution to depression as a presynaptic-trafficking-step druggable substrate. Clinical news: Vistagen's preliminary positive open-label-extension data from PALISADE-3 for fasedienol, the intranasal pherine in social anxiety — set against the failed randomized PALISADE-3 readout in December 2025 and the still-pending PALISADE-4 topline due this quarter. Psychedelics mechanism: a single dose of LSD in rats persistently reduces affective (but not sensory) pain via anterior cingulate cortex, with Neuropixels showing reduced stimulus-evoked nociceptive encoding and the cell-intrinsic-versus-network-level dissociation. Addiction: GLP1-ELP-FGF21, a long-acting bispecific GLP-1/FGF21 fusion, reducing voluntary alcohol consumption for 72 hours after a single dose in mice via NAc dopamine modulation — bridging the active semaglutide-for-AUD program and FGF21's alcohol-responsive central reward biology. Agentic AI block: Open-Rosalind reframes the bio-agent validation question around process-aware benchmarking — evidence-grounded outputs, trace completeness, workflow-constrained execution, explicit tool mediation for factual claims — complementary to ClaroAI-Bench's end-to-end reproduction framing covered earlier this week. Then "Cell-Level Virtual Screening" asks the basic operational question of whether virtual-cell ranking actually predicts downstream cellular phenotype rather than just other predictions. Closing lightest: ramelteon, the MT1/MT2 melatonin agonist sold for insomnia, facilitates hippocampal sharp-wave ripple occurrence and amplitude during NREM in mice, reframing the long-whispered cognitive-enhancement signal around a memory-consolidation substrate that's translationally measurable. Production note: arXiv hard-blocked this IP throughout the run with rate-exceeded responses; Tier 2 is fully bioRxiv-derived today. Voiced by Mistral voxtral-mini-tts-2603 with en_paul_neutral, ElevenLabs Bella as fallback. 2026-05-14-receptor-and-reason Thu, 14 May 2026 12:00:00 +0000 864 Daily roundup for May 14, 2026: GluD as ionotropic dopamine receptor, D3R subtype in quetiapine aversion, NTSR1 biased agonism with SBI-553 in extended amygdala, cerebellum-VTA VGLUT2 in stress depression, Vistagen fasedienol PALISADE-3 OLE in social anxiety, LSD-in-ACC affective pain mechanism, dual GLP-1/FGF21 in voluntary alcohol intake, Open-Rosalind tool-first bio-agents, Cell-Level Virtual Screening, and ramelteon facilitating hippocampal ripples. Note: arXiv rate-blocked this run; Tier 2 is fully bioRxiv-derived. The debut episode of Receptor & Reason — your daily roundup of neuropsychopharmacology, computational pharmacology, and agentic AI in biology, curated for Alan Huebschen. Today (May 13, 2026) is heavy on receptor pharmacology. Leading off: the Christopoulos and Thal groups at Monash report a 2.7-angstrom cryo-EM structure of the second-generation M4 muscarinic positive allosteric modulator MK-97 bound to the M4 mAChR with acetylcholine — resolving the allosteric vestibule, mapping the rodent-to-human species-variability swap MK-97 reaches past, and providing a template for biased M4 PAM design relevant to schizophrenia, cognitive impairment in schizophrenia, and Alzheimer's psychosis. Then "It's not all DAT": a Neuropsychopharmacology paper using OCT3 knockout mice and in vivo voltammetry to show organic cation transporter 3 selectively mediates amphetamine reinforcement and amphetamine-evoked DA release, opening a wider therapeutic-index window than DAT-blocking strategies for stimulant use disorder. Wake Forest follows with a phospho-resolved KOR–DAT mechanism — Thr53 phosphorylation on the dopamine transporter is necessary for KOR-driven cocaine effects on nucleus accumbens DA and for the negative-affective signature of withdrawal, a candidate stratification handle for KOR antagonist programs in stimulant use disorder. The opioid block continues with BMS-986122, a mu opioid PAM from Michigan tested in rat spared-nerve-injury and tibial-neuroma models, potentiating low-dose morphine and oxycodone analgesia in chronic neuropathic pain; and a Duke paper using multisite LFP and interpretable ML to identify two distinct large-scale oscillatory networks for fentanyl exposure versus fentanyl withdrawal — a candidate translatable EEG biomarker substrate for withdrawal-targeted pharmacology. From Vrije Universiteit Brussel, azo-THC-3, a photoswitchable delta-9-THC derivative that converts from inactive trans to active cis under violet light, demonstrated for focal seizure suppression in a temporal-lobe epilepsy model via implanted optic fiber. Two agentic AI items in the middle. From Cedars-Sinai: robotic perturbation proteomics paired with AI-agent data analysis — 172 compounds, 1,232 proteomes, more than 8,700 quantified proteins in about three weeks, agentic AI handling MoA-deconvolution at scale. From Sanofi: Talk2QSP, a human-in-the-loop multi-agent framework that translates unstructured clinical-scenario descriptions in the QSP literature into executable, simulation-ready model interventions — attacking the rate-limiting step in QSP operationalization, dose selection, and trial simulation. Korea University circuit piece: a self-limiting orexin–habenula circuit (lateral hypothalamic orexin neurons engaging Ddc-positive D-neurons in lateral habenula via OX2R) that mobilizes active coping and NAc dopamine during stress, then self-suppresses — arguing for stress-context-sensitive rather than chronic OX2R blockade in seltorexant/suvorexant-class programs. Closing on policy: HHS launched a federal initiative to reduce psychiatric "overprescribing" with explicit antidepressant deprescribing emphasis, sitting uncomfortably alongside the late-April FDA national priority vouchers issued to Compass Pathways, the Usona Institute, and Transcend Therapeutics for psilocybin and methylone programs — two federal agencies pulling in different directions on psychiatric pharmacotherapy. Note: arXiv hard-blocked this run with HTTP 429 from this IP, so Tier 2 backfilled via web-search rather than the listing API. Production note: Receptor & Reason is voiced by Mistral voxtral-mini-tts-2603 with the en_paul_neutral voice, with ElevenLabs Bella as fallback. 2026-05-13-receptor-and-reason Wed, 13 May 2026 12:00:00 +0000 680 Daily roundup for May 13, 2026: M4 PAM cryo-EM (MK-97), OCT3 vs DAT in amphetamine reinforcement, KOR–DAT Thr53 phosphorylation in cocaine, BMS-986122 MOR PAM in neuropathic pain, fentanyl withdrawal brain state, azo-THC-3 photoswitchable cannabinoid for epilepsy, agentic AI for proteomic MoA discovery, Sanofi's Talk2QSP, orexin–habenula resilience circuit, and the HHS antidepressant overprescribing initiative. Note: arXiv was rate-blocked this run; Tier 2 backfilled via web-search.