single-cell-rna-qc

single-cell-rna-qc performs automated quality control on single-cell RNA-seq data following scverse/scanpy best practices. It solves the common bioinformatics need to filter low-quality cells and assess dataset quality before downstream analysis, using MAD-based (median absolute deviation) outlier detection together with comprehensive before/after visualizations. It accepts .h5ad files (AnnData format from scanpy/Python workflows) and .h5 files (10X Genomics Cell Ranger output), auto-detecting the input format.

The skill offers two approaches. Approach 1 is a complete convenience pipeline (scripts/qc_analysis.py) that calculates QC metrics (count depth, gene detection, mitochondrial/ribosomal/hemoglobin content), applies permissive MAD-based filtering on counts/genes/MT%, filters lowly-detected genes, and generates before/after plots with threshold overlays. It exposes command-line parameters for output directory, per-metric MAD thresholds, a hard mitochondrial cutoff, gene filtering thresholds, and species-specific gene-name patterns, and writes a filtered dataset, a QC-annotated dataset, and PNG visualizations to a results directory. Approach 2 exposes modular building blocks from qc_core.py and qc_plotting.py (for example calculate_qc_metrics, detect_outliers_mad, filter_cells, plot_qc_distributions) for custom workflows such as applying different thresholds to cell subsets or running only metrics/visualization. It requires anndata, scanpy, scipy, matplotlib, seaborn, and numpy.

It targets computational biologists and bioinformaticians analyzing single-cell datasets who want a standards-aligned, "just works" QC step or reusable functions to embed in a larger pipeline. Use cases include standard QC on a new dataset, batch processing multiple datasets, exploratory quality assessment, and custom filtering logic (for example filtering neurons differently from other cells). It is distributed under the Apache License 2.0.