Combinatorial CRISPR Screens - Usage Guide

Overview

Decision-grade design and analysis of combinatorial CRISPR screens. Covers paired-Cas9 (Big Papi, Najm 2018), enhanced AsCas12a multiplex (enCas12a, DeWeirdt 2021), in4mer 4-guide arrays (Esmaeili Anvar 2024), and Inzolia paralog library; paralog buffering detection (Dede 2020, Thompson 2021) for synthetic-lethal targets; genetic-interaction (GI) scoring; and the MAGeCK MLE interaction-term approach.

Prerequisites

conda install -c bioconda mageck   # not on PyPI
pip install pandas numpy scipy matplotlib
# Inzolia library annotation: bioconductor or hart-lab
# Inzolia annotation: Nat Commun 15:3577 supplement, or Addgene #209551-2

# Cas12a-aware analysis tools
# No dedicated Cas12a helper package; use pandas/numpy with the library annotation

Required inputs: - Cassette-level count matrix (rows = cassettes, columns = samples) - Library annotation: cassette_id -> gene_A, gene_B (and gene_C, gene_D for in4mer) - Singleton controls in library - Vehicle / Day 0 / endpoint samples

Quick Start

Tell the AI agent what to do: - "Design an Inzolia paralog-pair library covering 600 paralog pairs in the kinase / transcription factor / DDR families with all-singleton controls" - "Analyze my Cas12a multiplex paralog screen: compute per-pair GI scores; identify synthetic-lethal pairs at GI z-score <-2" - "Compare Big Papi paired-Cas9 vs Inzolia Cas12a for the same paralog-pair set" - "Diagnose: my paired-Cas9 cassettes seem to act as single perturbations -- diagnose dual-cassette cloning failure" - "Run MAGeCK MLE with explicit interaction terms on my 4-condition combinatorial screen (NT, A_KO, B_KO, A_B_KO)"

Example Prompts

Library Design

"Build an Inzolia-style enAsCas12a 4-guide array library covering 400 receptor tyrosine kinase paralog pairs. Include singletons (gene A alone, gene B alone, double-NTC) so we can compute GI = double_LFC - sum(single_LFC). Output library.tsv."

"Design Big Papi constructs for 150 specific DDR pathway pairs using the pPapi architecture: SaCas9 sgRNA and SpCas9 sgRNA expressed from the U6 and H1 promoters, with SpCas9 supplied by the cell line. Verify by amplicon sequencing of clones."

Per-Pair GI Analysis

"Apply GI scoring to my paired-Cas9 screen. From paired_lfc.tsv and single_lfc.tsv, compute observed_double - (single_A + single_B). Z-normalize across pairs. Classify: SL (z <-2), rescue (z >2), no-interaction (otherwise)."

"From my in4mer Inzolia screen counts, aggregate cassette-level LFCs by paralog pair. Compute per-pair GI z-scores. Identify 24+ synthetic-lethal pairs as in Dede 2020."

MAGeCK MLE with Interactions

"Build MAGeCK MLE design matrix for a 4-condition combinatorial screen: NT, A_KO, B_KO, A_B_KO. Include interaction column. Run mageck mle; output per-pair interaction|beta and |fdr. Flag synthetic-lethal pairs at FDR <0.05 with negative beta."

Method Comparison

"Compare GI z-scoring vs MAGeCK MLE interaction-term approach on the same screen. Investigate disagreement: which method is more sensitive for small-effect interactions?"

"Validate Big Papi (Cas9) vs Inzolia (Cas12a) on the same 50 paralog pairs. Quantify concordance of synthetic-lethal calls."

Diagnostics

"My Cas12a in4mer screen has 20% of cassettes editing <50% per Cas12a editing assay. Investigate: locus-specific inefficiency? Filter and re-analyze."

"Many GI z-scores are positive (synthetic rescue) for essential pairs (e.g., RPS3 + RPL11). Diagnose: linear-space saturation; switch to log-space (LFC) GI scoring."

"My paired-Cas9 cassettes show as single perturbation despite design. Diagnose: recombination between repeated U6/tracrRNA elements; re-clone in pPapi (orthologous SaCas9 + SpCas9, U6 + H1); re-clone with validated dual-cassette protocol."

What the Agent Will Do

Identify screen architecture: paired-Cas9 vs Cas12a multiplex vs in4mer
Verify library composition: paired cassettes + singletons + NTCs
Run standard pooled-screen counting and QC (see [[screen-qc]])
For Cas12a: verify per-locus editing efficiency >50%
Compute per-cassette LFCs vs control
Aggregate to per-pair LFCs across replicate cassettes
Extract single-gene LFCs from singleton-control cassettes
Compute GI = observed_double - expected_additive
Z-normalize GI across all pairs; apply z = -2 cutoff for synthetic lethal
Cross-validate via MAGeCK MLE interaction-term analysis
Annotate hits against COSMIC, paralog databases, and known synthetic-lethal pairs
Output: per-pair GI table, synthetic-lethal candidates, MAGeCK MLE comparison, validation strategy

Tips

For genome-scale paralog screens, Inzolia (Cas12a, 4-guide arrays) is preferred. It is ~30% smaller than a typical monogenic Cas9 library while additionally covering ~4,000 paralog pairs, which makes it more cost-effective at scale.
Always include singleton controls in your library. Without them, you cannot compute GI; you only have paired LFCs without baseline.
Per-gene per-pair, include 4+ cassettes for stable GI scoring. Single-cassette per pair gives noisy z-scores.
Cas12a editing efficiency is locus-specific. Pilot at representative loci before screening; flag low-efficiency loci.
For drug-target nomination, validate synthetic-lethal hits via:
Orthogonal chemistry (Cas9 if originally Cas12a)
Arrayed knock-out validation
CRISPRi orthogonal (no DNA damage)
Pharmacological inhibition with drugs
Linear-space GI saturation: for essential-pair tests where both genes are independently essential, the additive expectation overestimates the saturable double-effect. Use log-space (LFC) GI scoring.
For combinatorial Perturb-seq (single-cell readout of multi-perturbation), see [[perturb-seq-analysis]].
Dual-sgRNA constructs with repeated U6/tracr elements recombine in lentivirus and collapse to a single perturbation. Use the pPapi architecture and verify by amplicon sequencing of clones. Verify by amplicon sequencing of clones before screening.

Decision Cheat Sheet

Goal	Architecture	Library
Genome-scale paralog buffering	Cas12a multiplex	Inzolia (4-guide arrays)
Specific pairwise GI	Paired-Cas9	Big Papi
3-way / 4-way KO	Cas12a in4mer	Custom in4mer
Combinatorial Perturb-seq	Cas9 multi-cassette + scRNA	Custom
Drug + KO interaction	Standard library + drug	drugZ or MAGeCK MLE

Thresholds

Threshold	Value	Rationale
Synthetic lethal	GI z <-2	Standard
Synthetic rescue	GI z >2	Standard
Cassettes per pair	4-6	Stable GI
Singletons per gene	4-6 cassettes	For expected_additive
Cas12a editing efficiency	>50%	Per-locus inclusion
Library coverage at infection	500x cassettes/cells	Standard pooled

crispr-screens/library-design - Inzolia / in4mer / Big Papi design
crispr-screens/screen-qc - Cassette-level QC
crispr-screens/mageck-analysis - MAGeCK MLE with interaction terms
crispr-screens/hit-calling - Cross-method validation
crispr-screens/perturb-seq-analysis - Combinatorial Perturb-seq
crispr-screens/copy-number-correction - Pre-correction for cancer-line screens
pathway-analysis/go-enrichment - Functional analysis of GI clusters