version: "1.0.0" name: crispr-screen-analysis description: Analyze CRISPR pooled screens using MAGeCK. Accepts a raw count table or FASTQ files. Runs MAGeCK RRA (pairwise test) or MAGeCK MLE (multi-condition). Produces gene-level hit rankings, volcano plots, sgRNA rank plots, and hit summary figures for both positive and negative selection. Supports genome-wide KO screens, CRISPRi, and CRISPRa experiments.

CRISPR Screen Analysis (MAGeCK)

Purpose

Run a complete CRISPR pooled screen analysis pipeline using MAGeCK. Starting from either a raw count matrix or FASTQ files, identify positively and negatively selected genes, and generate publication-quality figures.

Use when

• the user has a CRISPR screen count table and wants to identify hits
• the user wants to find essential genes (negative selection) or enriched genes (positive selection)
• the user wants to compare treatment vs control conditions in a pooled screen
• the user has multiple conditions or time points (use MLE mode)
• the user has raw FASTQ files from a pooled screen and wants to go from reads to hits
• the user mentions MAGeCK, RRA, beta score, sgRNA count, or pooled screen analysis
• the user wants a volcano plot, rank plot, or gene score plot from screen data

Do not use when

• the user wants arrayed screen analysis (not pooled)
• the user wants single-cell CRISPR screen analysis (Perturb-seq / CROP-seq)
• the user wants to design sgRNAs (use design-sgrnas-by-gene instead)
• the user has only a gene hit list without raw count data

Supported modes

Inputs

• --count-table — TSV with columns: sgRNA, gene, sample1, sample2, ... (required for test/mle)
• --treatment — comma-separated sample column names for the treatment group
• --control — comma-separated sample column names for the control group
• --mode — test (default), mle, count, or all
• --fastq — space-separated FASTQ paths (for count/all mode)
• --library — sgRNA library file: TSV with columns sgRNA_id, sequence, gene (for count mode)
• --design-matrix — tab-separated design matrix file (for mle mode)
• --norm-method — median (default), total, control, or none
• --fdr-cutoff — FDR significance threshold (default 0.25)
• --lfc-cutoff — |log2FC| threshold for volcano plot labeling (default 1.0)
• --gene-lfc-method — median or alphamedian (default median)
• --top-n — number of top hits to label in plots (default 20)
• --prefix — output file prefix (default screen)
• --outdir — output directory (default ./crispr_screen_results/)

Outputs

Execution

# Pairwise test (most common)
python scripts/crispr_screen_analysis.py \
  --count-table counts.txt \
  --treatment Day14_rep1,Day14_rep2 \
  --control Day0_rep1,Day0_rep2 \
  --mode test \
  --outdir results/
 
# MLE with design matrix
python scripts/crispr_screen_analysis.py \
  --count-table counts.txt \
  --mode mle \
  --design-matrix design.txt \
  --outdir results/
 
# From FASTQ files
python scripts/crispr_screen_analysis.py \
  --fastq sample1.fastq.gz sample2.fastq.gz \
  --library sgrna_library.txt \
  --treatment sample1 \
  --control sample2 \
  --mode all \
  --outdir results/

Parameter decision guide

MAGeCK installation note

MAGeCK must be installed separately. The recommended method is conda:

conda install -c bioconda mageck

Or via pip (Python-only implementation):

pip install mageck

This skill will automatically detect whether MAGeCK is available and fall back to a Python re-implementation of the core RRA algorithm if it is not installed.