Coessentiality And Cofunctionality: A Network Approach To Learning Genetic Vulnerabilities From Cancer Cell Line Fitness Screens

Abstract

Genetic interaction networks are a powerful approach for functional genomics, and the synthetic lethal interactions that comprise these networks offer a compelling strategy for identifying candidate cancer targets. As the number of published shRNA and CRISPR perturbation screens in cancer cell lines expands, there is an opportunity for integrative analysis that goes further than pairwise synthetic lethality and discovers genetic vulnerabilities of related sets of cell lines. We re-analyze over 100 high-quality, genome-scale shRNA screens in human cancer cell lines and derive a quantitative fitness score for each gene that accurately reflects genotype-specific gene essentiality. We identify pairs of genes with correlated essentiality profiles and merge them into a cancer coessentiality network, where shared patterns of genetic vulnerability in cell lines give rise to clusters of functionally related genes in the network. Network clustering discriminates among all three defined subtypes of breast cancer cell lines (basal, luminal, and Her2-amplified), and further identifies novel subsets of Her2+ and ovarian cancer cells. We demonstrate the utility of the network as a platform for both hypothesis-driven and data-driven discovery of context-specific essential genes and their associated biomarkers.