Key Takeaways
- Spatial transcriptomics preserves tissue location information that scRNA-seq loses — AI agents can automate the analysis pipeline
- Key analysis tasks: spot deconvolution, spatially variable gene detection, cell-cell communication, tissue architecture mapping
- K-Dense spatial skills provide the most coverage: SpatialAgent (Genentech), squidpy, cell2location workflows
- SpatialAgent by Genentech (~30 stars): Purpose-built for spatial transcriptomics with AI
- OmicsClaw includes spatial transcriptomics as part of multi-omics integration
- Challenge: spatial data is larger and more complex than scRNA-seq — analysis requires more compute
Key Analysis Tasks
| Task | What It Does | AI Skill Coverage |
|---|---|---|
| Spot deconvolution | Estimate cell type composition per spot | K-Dense ✓, SpatialAgent ✓ |
| Spatially variable genes | Find genes with spatial expression patterns | K-Dense ✓ |
| Cell-cell communication | Infer ligand-receptor interactions in space | K-Dense ✓ |
| Tissue architecture | Map tissue regions and boundaries | SpatialAgent ✓ |
| Integration with scRNA-seq | Reference-based annotation | K-Dense ✓, OmicsClaw ✓ |
Tools
| Tool | Focus | Stars |
|---|---|---|
| SpatialAgent (Genentech) | Purpose-built spatial transcriptomics AI | ~30 |
| OmicsClaw (Tsinghua) | Multi-omics including spatial | 200+ |
| MedgeClaw + K-Dense | Spatial skills within broader toolkit | 1,000+ |
FAQ
Q1: Can AI agents fully automate spatial transcriptomics analysis?
Partially. Spot deconvolution and spatial gene detection are well-automated. Tissue architecture interpretation still benefits from expert histological knowledge that current AI skills don't fully capture.
Q2: What spatial platforms are supported?
10x Visium is best supported. MERFISH and Slide-seq have growing but less complete skill coverage.
Summary
Spatial transcriptomics adds tissue context to single-cell data — and AI agents can automate much of the analysis. K-Dense spatial skills and SpatialAgent (Genentech) provide the best coverage. The field is earlier-stage than scRNA-seq automation, with opportunities for new skill development.