Add scAgent: natural-language scRNA-seq analysis agent

[deepmind11]

Submission: scAgent

Repo: https://github.com/deepmind11/scAgent

Description: An AI agent that lets wet-lab biologists run complete single-cell RNA-seq analysis through natural language conversation — no programming required.

Key details:

• Full scRNA-seq pipeline via plain English: QC → normalization → batch integration (Harmony, scVI, BBKNN, Scanorama) → clustering → cell type annotation (CellTypist) → differential expression (pseudobulk DESeq2/edgeR) → pathway enrichment (GSEA)
• 85.7% on SC-Bench — a benchmark of 394 verifiable scRNA-seq problems (vs 52.8% top baseline)
• State-aware data inspector: determines what analysis steps have already been run on any loaded .h5ad
• Dependency resolution: automatically plans the minimal prerequisite steps for any requested analysis
• W3C PROV-O provenance tracking + auto-generated reproducibility packages (methods prose + replay.py)
• Long-term memory via ChromaDB across sessions
• Runs on Feynman or Claude Code

Language: Python · License: MIT

[Jenqyang]

Decision: Keep open.

Reason:

1. There is a public OSS repository here and the topic looks in scope for the Applications section.
2. We still need a formal PR in repository format with one concise neutral line.

Next step: Please open a PR with the exact entry line and keep the wording technical and evidence-based.