scrna-orchestrator

What problem does it solve?

Single-cell RNA-seq analysis is easy to misconfigure and hard to reproduce when steps like QC, normalization, clustering, and marker/DE testing are run ad hoc; this Skill provides a single local pipeline that enforces raw-count inputs and consistent defaults to produce reproducible results.

Core Features & Use Cases

• QC and Filtering: Mitochondrial percent filtering, minimum genes/cells thresholds, and early rejection of processed-like inputs.
• Preprocessing & Feature Selection: Total-count normalization, log1p transformation, and highly variable gene selection.
• Embedding, Clustering & Markers: PCA, neighbors graph, UMAP, Leiden clustering, and Wilcoxon-based cluster-vs-rest marker discovery.
• Optional Two-Group DE and Volcano Plot: Wilcoxon two-group DE on any obs column with optional volcano visualization and reproducibility bundle generation.
• Use Case: Run end-to-end QC, clustering, and marker discovery on a raw-count AnnData .h5ad file to produce figures, tables, and a reproducible commands/environment bundle for publication or review.

Quick Start

Run standard QC, clustering, marker discovery, and optional two-group differential expression on a raw-count .h5ad file and save the report, figures, and tables to the chosen output directory.