CoRe: a robustly benchmarked R package for identifying core-fitness genes in genome-wide pooled CRISPR-Cas9 screens

Abstract

Background CRISPR-Cas9 genome-wide screens are being increasingly performed, allowing systematic explorations of cancer dependencies at unprecedented accuracy and scale. One of the major computational challenges when analysing data derived from such screens is to identify genes that are essential for cell survival invariantly across tissues, conditions, and genomic-contexts (core-fitness genes), and to distinguish them from context-specific essential genes. This is of paramount importance to assess the safety profile of candidate therapeutic targets and for elucidating mechanisms involved in tissue-specific genetic diseases. Results We have developed CoRe: an R package implementing existing and novel methods for the identification of core-fitness genes (at two different level of stringency) from joint analyses of multiple CRISPR-Cas9 screens. We demonstrate, through a fully reproducible benchmarking pipeline, that CoRe outperforms state-of-the-art tools, yielding more reliable and biologically relevant sets of core-fitness genes. Conclusions CoRe offers a flexible pipeline, compatible with many pre-processing methods for the analysis of CRISPR data, which can be tailored onto different use-cases. The CoRe package can be used for the identification of high-confidence novel core-fitness genes, as well as a means to filter out potentially cytotoxic hits while analysing cancer dependency datasets for identifying and prioritising novel selective therapeutic targets.

Gene essentiality in cancer is better predicted by mRNA abundance than by gene regulatory network-inferred activity

“…So far, we have analyzed gene inhibition sensitivity from CRISPR screens as a quantitative trait. However, in many cases the results of CRISPR screens are binarized, such that genes are deemed to be either essential or non-essential for survival 37,39,40 . Genes which are essential in a specific context might then be considered as suitable therapeutic targets.…”

Recovering false negatives in CRISPR fitness screens with JLOE

“…We also examined the gene expression profiles of these core essential genes and observed that the 314 genes at the intersection of the three approaches have higher median gene expression compared to genes that are uniquely core essential in each approach (median log2(TPM) 6.3 versus 5.1 for new common specific genes, 5.4 for CEGV2 and 4.7 for Sanger specific genes; Figure 3C ). An expanded comparison of the newly identified common essentials with the more recently defined core essential genes using CoRe AdaM ( 44 ) and CENtools ( 45 ) approaches showing the overlap between the datasets is shown in Supplementary Figure S6 . These observations are consistent with an increased false negative rate among essential genes with moderate levels of mRNA expression.…”

Highlights from the 1st European cancer dependency map symposium and workshop

“…They highlighted the opportunity to generate new models for rare and paediatric cancer, where there is a large unmet clinical need, and to map their dependencies [23]. Furthermore, they showcased studies using co-essentiality analysis to describe novel gene function [9], integration of pharmacological and CRISPR data to understand drug mode of action [10], opportunities to identify drug combinations [24], to interpret cancer gene mutations and gene fusions [25], and to define human essential genes [26], amongst other possible applications.…”