# This file has all general filepaths and directories used in the scoring pipeline. The individual scripts may have # additional parameters specific to each method # See here for further details: https://github.com/OATML-Markslab/ProteinGym/issues/63 # Most of these files can be automatically downloaded # using the `proteingym.utils.download.download_resources()` function. # If so, you can set `PROTEINGYM_CACHE` here # to the same as `proteingym.utils.download.PROTEINGYM_CACHE` # By default, this will be: "$HOME/.cache/ProteinGym" export PROTEINGYM_CACHE="$HOME/.cache/ProteinGym" # ---- DMS zero-shot parameters ---- # # Folders containing the csvs with the variants for each DMS assay ## DMS benchmark - Substitutions 1.0GB DMS_ProteinGym_substitutions.zip export DMS_data_folder_subs="${PROTEINGYM_CACHE}/DMS_ProteinGym_substitutions/" ## DMS benchmark - Indels 200MB DMS_ProteinGym_indels.zip export DMS_data_folder_indels="${PROTEINGYM_CACHE}/DMS_ProteinGym_indels/" # Folders containing multiple sequence alignments and MSA weights for all DMS assays ## Multiple Sequence Alignments (MSAs) for DMS assays 5.2GB DMS_msa_files.zip export DMS_MSA_data_folder="${PROTEINGYM_CACHE}/DMS_msa_files/" ## Redundancy-based sequence weights for DMS assays 200MB DMS_msa_weights.zip export DMS_MSA_weights_folder="${PROTEINGYM_CACHE}/DMS_msa_weights/" # Reference files for substitution and indel assays ## Included in GitHub repo export DMS_reference_file_path_subs=../../reference_files/DMS_substitutions.csv ## Included in GitHub repo export DMS_reference_file_path_indels=../../reference_files/DMS_indels.csv # Folders where fitness predictions for baseline models are saved ## Zero-shot DMS Model scores - Substitutions 31GB zero_shot_substitutions_scores.zip export DMS_output_score_folder_subs="${PROTEINGYM_CACHE}/zero_shot_substitutions_scores/" ## Zero-shot DMS Model scores - Indels 5.2GB zero_shot_indels_scores.zip export DMS_output_score_folder_indels="${PROTEINGYM_CACHE}/zero_shot_indels_scores/" # Folder containing EVE models for each DMS assay ## This is where you would store your local copy of trained EVE models ## (eg., needed for scoring mutated sequences with EVE or TranceptEVE). ## Note that given the substantial size of all EVE model checkpoints that is not something that we make available for download by default. ## But the code to train these models from scratch is provided here: ## https://github.com/OATML-Markslab/ProteinGym/blob/main/scripts/scoring_DMS_zero_shot/training_EVE_models.sh export DMS_EVE_model_folder="${PROTEINGYM_CACHE}/DMS_EVE_models/" # Folders containing merged score files for each DMS assay ## you can specify any location of your choosing. This is where merged files will be stored if you run ## the score merge script (https://github.com/OATML-Markslab/ProteinGym/blob/main/scripts/scoring_DMS_zero_shot/merge_all_scores.sh). ## This merge script takes scores from all baselines in DMS_output_score_folder_subs (the ones we provide, ## plus new ones you may have computed yourself), and creates individual files per assay that includes ## scores for all baselines. export DMS_merged_score_folder_subs="${PROTEINGYM_CACHE}/merged_scores/subs/" ## Same thing for indels export DMS_merged_score_folder_indels="${PROTEINGYM_CACHE}/merged_scores/indels/" # Folders containing predicted structures for the DMSs (AF2=AlphaFold2) ## Predicted 3D structures from inverse-folding models 84MB ProteinGym_AF2_structures.zip export DMS_structure_folder="${PROTEINGYM_CACHE}/ProteinGym_AF2_structures/" # ---- Clinical parameters ---- # # Folder containing variant csvs ## Clinical benchmark - Substitutions 123MB clinical_ProteinGym_substitutions.zip export clinical_data_folder_subs="${PROTEINGYM_CACHE}/clinical_ProteinGym_substitutions/" ## Clinical benchmark - Indels 2.8MB clinical_ProteinGym_indels.zip export clinical_data_folder_indels="${PROTEINGYM_CACHE}/clinical_ProteinGym_indels/" # Folders containing multiple sequence alignments and MSA weights for all clinical datasets ## Clinical MSAs 17.8GB clinical_msa_files.zip export clinical_MSA_data_folder="${PROTEINGYM_CACHE}/clinical_msa_files/" ## Clinical MSA weights 250MB clinical_msa_weights.zip export clinical_MSA_weights_folder="${PROTEINGYM_CACHE}/clinical_msa_weights/" # Folder containing MSA weights for all clinical datasets ## use files in clinical_msa_weights.zip export clinical_MSA_weights_folder_subs="${PROTEINGYM_CACHE}/clinical_msa_weights/subs/" ## use files in clinical_msa_weights.zip export clinical_MSA_weights_folder_indels="${PROTEINGYM_CACHE}/clinical_msa_weights/indels/" # reference files for substitution and indel clinical variants ## Included in GitHub repo export clinical_reference_file_path_subs=../../reference_files/clinical_substitutions.csv ## Included in GitHub repo export clinical_reference_file_path_indels=../../reference_files/clinical_indels.csv # Folder where clinical benchmark fitness predictions for baseline models are saved ## Clinical Model scores - Substitutions 0.9GB zero_shot_clinical_substitutions_scores.zip export clinical_output_score_folder_subs="${PROTEINGYM_CACHE}/zero_shot_clinical_substitutions_scores/" ## Clinical Model scores - Indels 0.7GB zero_shot_clinical_indels_scores.zip export clinical_output_score_folder_indels="${PROTEINGYM_CACHE}/zero_shot_clinical_indels_scores/" # Folder containing EVE models for each clinical variant ## same as DMS_EVE_model_folder, but for proteins included in the clinical benchmarks. export clinical_EVE_model_folder="${PROTEINGYM_CACHE}/clinical_EVE_models/" # Folder containing merged score files for each clinical variant ## same as DMS_merged_score_folder_subs, but for the clinical benchmark. export clinical_merged_score_folder_subs="${PROTEINGYM_CACHE}/merged_scores/subs/" ## same as DMS_merged_score_folder_indels, but for the clinical benchmark. export clinical_merged_score_folder_indels="${PROTEINGYM_CACHE}/merged_scores/indels/"