---
name: spatial-transcriptomics
description: squidpy spatial gold chain on SpatialData .zarr or AnnData .h5ad with obsm['spatial']. Use when the user has Visium/Slide-seq/generic spots or cells with coordinates. Builds a knn spatial graph, Moran SVGs, and spatial_scatter plots. Multi-table SpatialData requires --table. Does not run Cell2location, BayesSpace, SpaGCN, or vendor HD/Xenium pipelines. Numeric clusters only.
---

# Spatial transcriptomics gold chain (squidpy)

Default path is **squidpy**, not the legacy fused-graph / NNLS helpers.

```bash
python scripts/doctor.py   # need ready.spatial_ready
python skills/spatial-transcriptomics/scripts/spatial_inspect.py visium.h5ad
python skills/spatial-transcriptomics/scripts/spatial_pipeline.py visium.h5ad -o spatial_out.h5ad
python skills/spatial-transcriptomics/scripts/spatial_plot.py spatial_out.h5ad -o figures/ --color leiden
# SpatialData with several tables:
python skills/spatial-transcriptomics/scripts/spatial_inspect.py data.zarr --table table_name
```

Accepts SpatialData `.zarr` (requires `spatialdata`) or `.h5ad` with `obsm['spatial']`. Multiple tables **refuse** unless `--table` is set.

| Step | Script | Backend |
|---|---|---|
| inspect | `spatial_inspect.py` | AnnData / SpatialData I/O |
| pipeline | `spatial_pipeline.py` | `spatial_neighbors_knn` + Moran |
| deconvolve | `spatial_deconvolution.py` | Tangram optimal transport (`tangram-sc` / PyTorch) |
| plot | `spatial_plot.py` | `squidpy.pl.spatial_scatter` → `spatial_{color}.png` |

**Refuses** if squidpy is missing for Moran's I. For deconvolution, Tangram deep learning is used with transparent NNLS fallback.