Genome-aware annotation of CRISPR guides validates targets in variant cell lines and enhances discovery in screens

Abstract

AbstractPooled CRISPR-Cas9 genetic knockout screens are powerful high-throughput tools for identifying chemo-genetic, synthetic-lethal and synthetic-viability interactions and are used as a key step towards identifying disease-modifying knockout candidates and informing drug design and therapeutic regimens. CRISPR guide libraries are commercially available for purchase and have been widely applied in different cell lines. However, discrepancies between the genomes used to design CRISPR libraries and the genomes of the cells subjected to CRISPR screens lead to loss of signal or introduction of bias towards the most conserved genes. Here, we present an algorithm, EXOme-guided Reannotation of nuCleotIde SEquences (Exorcise), which uses sequence search and CRISPR target annotation to adapt existing CRISPR libraries to user-defined genomes and exomes. Applying Exorcise on 55 commercially available CRISPR-spCas9 knockout libraries for human and mouse, we found that all libraries have mis-annotations, and that design strategy affects off-target effects and targeting accuracy relative to a standard reference sequence. In simulations on synthetic data, we modelled common mis-annotations in CRISPR libraries and found that they adversely affected recovery of the ground truth for all genes except for those with the strongest signals. Finally, we reanalysed DepMap and DDRcs CRISPR screens with Exorcise annotations and found that strong hits were retained, and lower-confidence hits were strengthened. Use of Exorcise on DepMap with exomes inferred from transcriptomic expression data demonstrated that cell-line–aware reannotation is possible without whole-genome sequencing. Taken together, our results show that Exorcise is a powerful reannotation tool that focuses existing CRISPR libraries towards the cell line genome under investigation and allows post-hoc reanalysis of completed CRISPR screens. Exorcise is open-source software licenced under a Creative Commons Zero Universal 1.0 licence and is available at <https://github.com/SimonLammmm/exorcise>.