CoRe: A robustly benchmarked R package for the identification of core fitness genes in genome-wide pooled CRISPR knock-out screens

Abstract CRISPR-Cas9 recessive genome-wide pooled screens have allowed systematic explorations of weaknesses and vulnerabilities existing in cancer cells, across different tissue lineages at unprecedented accuracy and scale. The identification of novel genes essential for selective cancer cell survival is currently one of the main applications of this technology. Towards this aim, distinguishing genes that are constitutively essential (invariantly across tissues and genomic contexts, i.e. core-fitness genes) from those whose essentiality is associated with molecular features peculiar to certain cancers is of paramount importance for identifying new oncology therapeutic targets. This is crucial to assess the risk of a candidate target’s suppression impacting critical cellular processes that are unspecific to cancer. On the other hand, identifying new human core-fitness genes might also elucidate new mechanisms involved in tissue-specific genetic diseases. We present CoRe: an open-source R package implementing established and novel methods for the identification of core-fitness genes based on joint analyses of data from multiple CRISPR-Cas9 screens. In addition, we present results from a fully reproducible benchmarking pipeline demonstrating that CoRe outperforms other state-of-the-art methods, and it yields more reliable sets of core-fitness and common-essential genes with respect to existing reference sets and methods.