By name: cellagent-annotation description: Cell tagger keywords: - single-cell - markers - annotation - confidence - tissue measurable_outcome: Label every provided cluster with a cell type + confidence + marker evidence (or "ambiguous") within 15 minutes per dataset. license: MIT metadata: author: CellAgent Team version: "1.0.0" compatibility: - system: Python 3.9+ allowed-tools: - run_shell_command - read_file
CellAgent Annotation
Use CellTypeAgent to interpret marker genes, annotate scRNA-seq clusters, and coordinate multi-agent workflows for downstream analysis.
When to Use
- Automated annotation of scRNA-seq datasets without manual curation.
- Multi-step workflows (QC → clustering → annotation → DE analysis).
- Integrating multiple batches requiring consistent labeling.
Core Capabilities
- Planning: Multi-agent planner decomposes analysis goals into steps.
- Tool execution: Generates Scanpy/Seurat code and runs it autonomously.
- Self-correction: Detects execution errors and retries with fixes.
Workflow
- Gather marker lists per cluster, plus species/tissue context and optional atlas references.
- Run CellTypeAgent (
pip install -r requirements.txtthenpython repo/main.py --data data.h5ad --goal annotate). - Review outputs for supporting markers; downgrade ambiguous clusters when signals conflict.
- Produce final table (cluster, label, confidence, supporting markers, notes) and cite references when used.
Example Usage
python3 Skills/Genomics/Single_Cell/CellAgent/repo/main.py --data "./data.h5ad" --goal "annotate"
Guardrails
- Avoid over-specific lineages if markers overlap; default to broader types.
- Flag clusters showing multiple signatures for manual review.
- Respect species/tissue differences when interpreting markers.
References
- README + upstream paper (Mao et al., 2025 / arXiv 2407.09811).