What problem does it solve? Pooled CRISPR screen analysis involves many fragile decisions—baseline choice, control classes, copy-number artifacts, and method selection—where a wrong-but-silent commitment invalidates every downstream hit with no error thrown. This Skill orchestrates the entire pipeline so each step uses the correct reference, QC gate, and hit-calling method for the experimental design. ## Core Features & Use Cases - End-to-end orchestration: Runs guide counting with MAGeCK, six-stage QC (plasmid Gini, replicate Pearson, CEGv2 PR-AUC), copy-number correction with CRISPRcleanR or Chronos, and batch covariates via MAGeCK MLE. - Design-matched hit calling: Selects among MAGeCK RRA/MLE, BAGEL2, drugZ, JACKS, and Chronos based on screen type, then builds tier-based consensus (Tier 1 = 3-method agreement) across methods. - Specialized branches: Hands off to dedicated workflows for Perturb-seq, combinatorial paralog screens, base-editor and prime-editor variant screens, and in vivo bottleneck-aware screens. - Use Case: Given FASTQ files from a Brunello dropout screen with vehicle and drug arms, the Skill counts guides, verifies QC gates, runs drugZ and MAGeCK in parallel, and outputs tier-2 consensus hits at FDR < 0.05. ## Quick Start Ask the agent to run the end-to-end CRISPR screen pipeline on your pooled screen FASTQ files, specifying the baseline, library control classes, and screen type so it can count guides, run QC, and call consensus hits.
