quality-control

name	quality-control
description	Compute QC metrics, filter cells, detect doublets for single-cell RNA-seq data.
allowed-tools	Read, Bash
compatibility	scRNA-seq, snRNA-seq

Quality Control

When to Use

Before clustering to remove low-quality cells and doublets.

Functions

compute_qc_metrics(adata) -> dict

Computes standard QC metrics:

• n_genes: number of detected genes per cell
• total_counts: total UMI counts per cell
• pct_mito: percentage of mitochondrial genes
• pct_ribo: percentage of ribosomal genes

Returns metrics added to adata.obs: n_genes, total_counts, pct_mito, pct_ribo

Mitochondrial genes detected by prefix "MT-" (human) or "mt-" (mouse). Ribo genes by prefixes "RPS", "RPL" (ribo protein) and "RPSA".

filter_cells(adata, min_genes=200, max_genes=5000, min_counts=500, max_counts=50000, max_mito_pct=20) -> AnnData

Filters cells based on QC thresholds. Returns subsetted AnnData.

• Cells with too few genes = likely empty droplets
• Cells with too many genes = likely doublets
• High mito percentage = dying cells

detect_doublets(adata, expected_cells=None) -> list[int]

Heuristic doublet detection based on gene count outliers. Cells with n_genes > Q3 + 1.5*IQR flagged as potential doublets. Returns list of cell indices flagged as doublets.