rarecell: reproducible isolation of rare single‑cell populations
It solves the problem of extracting rare cell types from scRNA‑seq data using a human‑in‑the‑loop workflow. Users provide an .h5ad file and a natural‑language description; the tool drafts marker panels via LLM, validates them, then runs a deterministic state‑machine to isolate cells and logs every decision for reproducibility. Designed for computational biologists and bioinformaticians needing transparent, reproducible rare‑cell analysis, it stands out by combining LLM assistance with a fully auditable pipeline, unlike typical clustering tools that lack provenance.
View on GitHub →PatrickJReed/rarecell