spatial-transcriptomics

Files (14)

• SKILL.md
• gold_chain.example.json · configs/
• platform_configs.yml · configs/
• _common.py · scripts/
• spatial_clustering.py · scripts/
• spatial_deconvolution.py · scripts/
• spatial_inspect.py · scripts/
• spatial_io.py · scripts/
• spatial_niche_analysis.py · scripts/
• spatial_pipeline.py · scripts/
• spatial_plot.py · scripts/
• spatial_preprocessing.py · scripts/
• spatial_variable_genes.py · scripts/
• spatial_visualization.py · scripts/

SKILL.md (first lines)

# 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.

From HERRY423/BioNexus, Apache-2.0.

More from HERRY423/BioNexus