Reducing False Positives in CRISPR/Cas9 Screens from Copy Number Variations

Abstract

CRISPR/Cas9 knockout screens have been widely used to interrogate gene functions across a wide range of cell systems. However, the screening outcome is biased in amplified genomic regions, due to the ability of the Cas9 nuclease to induce multiple double-stranded breaks and strong DNA damage responses at these regions. We developed algorithms to correct biases associated with copy number variations (CNV), even when the CNV profiles are unknown. We demonstrated that our methods effectively reduced false positives in amplified regions while preserving signals of true positives. In addition, we developed a sliding window approach to estimate regions of high copy numbers for cases in which CNV information is not available.These copy number estimations can subsequently be used to effectively correct CNV-related biases in CRISPR screening experiments. Our approach is integrated into the existing MAGeCK/MAGeCK-VISPR analysis pipelines and provides a convenient framework to improve the precision of CRISPR screening results.peer-reviewed)

A benchmark of algorithms for the analysis of pooled CRISPR screens

“…Because of this, we believe that the most promising area for future research is in identifying issues specific to the experimental design of CRISPR-pooled screens. The recent work in identifying biases such as copy number-associated effects [30][31][32], structural rearrangement effects [33], and bottleneck effects [57] is exemplary of promising directions. We believe that there are further questions to be answered: for example, Are such biases generalizable to CRISPR interference and CRISPR activation screens?…”

“…Varying gene effect sizes can result in a bias towards finding only genes with large effects [29]. Cell death from excessive cutting in high copy number regions can lead to false positives in CRISPRko screens [30][31][32][33]. These issues complicate the data analysis and often cause both false positives and false negatives.…”

High‐resolution mapping of cancer cell networks using co‐functional interactions

“…Studies involving parallel screens are straightforward to design, but technical variation in how the screens are performed as well as copy number variation across cell backgrounds can confound the results (Zhang & Lu, ; Aguirre et al , ). Recent work has shown that copy number variation can underlie the strongest hits in CRISPR‐knockout screens, and multiple groups have proposed corrective algorithms to confront this problem (Pommier, ; Meyers et al , ; Data ref: Meyers et al , ; preprint: Wu et al , ). Additional heuristics aimed at increasing the quality of genetic interactions identified from parallel genetic screens have included discarding entire screens with noisy effect sizes, setting an effect size threshold for correlating genes, and capping the number of interactions per gene (Wang et al , ; preprint: Kim et al , ; Pan et al , ); however, reliance on these heuristics prevents truly unbiased genome‐wide analyses (McFarland et al , ).…”

High-resolution mapping of cancer cell networks using co-functional interactions

“…Studies involving parallel screens are straightforward to design, but technical variation in how the screens are performed as well as copy number variation across cell backgrounds can confound the results (Zhang & Lu, 2009;Aguirre et al, 2016). Recent work has shown that copy number variation can underlie the strongest hits in CRISPR knockout screens, and multiple groups have proposed corrective algorithms to confront this problem (Pommier, 2006;Meyers et al, 2017;Wu et al). Additional heuristics aimed at increasing the quality of genetic interactions identified from parallel genetic screens have included discarding entire screens with noisy effect sizes, setting an effect size threshold for correlating genes, and capping the number of interactions per gene Pan et al, 2018;Kim et al, 2018); however, reliance on these heuristics prevents truly unbiased genome-wide analyses.…”