spatial-transcriptomics-agent

name: spatial-transcriptomics-agent description: Spatial analyst keywords: - spatial - h5ad - H&E - clustering - SVG measurable_outcome: For each sample, deliver ≥1 spatial domain map + SVG list + narrative interpretation within 30 minutes. license: MIT metadata: author: LiuLab version: "1.0.0" compatibility: - system: Python 3.9+ allowed-tools: - run_shell_command - read_file - web_fetch

Spatial Transcriptomics Agent

Run STAgent to align histology images with expression matrices, perform clustering/SVG detection, and generate literature-backed spatial reports.

When to Use

• Analysis of Visium/Xenium or similar ST datasets.
• Visual reasoning over spatial plots, H&E images, or cluster maps.
• Automatically generating Scanpy/Squidpy code for new ST workflows.
• Hypothesis generation about spatial gene expression patterns.

Core Capabilities

1. Dynamic code generation: Create/execute Python scripts for QC, clustering, SVG detection.
2. Visual reasoning: Interpret spatial plots to identify tissue domains and cell neighborhoods.
3. Literature retrieval: Pull references that contextualize findings.
4. Report generation: Deliver publication-style writeups with plots and SVG tables.

Workflow

1. Env setup: conda env create -f environment.yml && conda activate STAgent.
2. Data prep: Supply expression_path (.h5ad/Spaceranger) + image_path (H&E/IF) and metadata.
3. Task selection: Choose tasks such as cluster, find_svg, annotate_domains, or composite instructions; run python repo/src/main.py --data_path ... --task "...".
4. Execute & interpret: Let STAgent generate scripts, run analyses, and interpret results with literature references.
5. Package outputs: Save UMAP/spatial plots, SVG tables, QC details, and summary markdown.

Example Usage

User: "Analyze this breast cancer ST dataset, find immune infiltrates."
Agent: loads data, runs `sqidpy.gr.spatial_neighbors`, computes Leiden clusters, plots marker genes (CD3D, CD19), and summarizes which clusters map to tumor core vs. stromal/immune zones.

Guardrails

• Document coordinate systems and any scaling between imaging and expression coordinates.
• Avoid definitive cell-type labels without supporting markers.
• Capture QC parameters for reproducibility.

References

• Source repo: https://github.com/LiuLab-Bioelectronics-Harvard/STAgent
• See local README.md for detailed instructions.