A network of human functional gene interactions from knockout fitness screens in cancer cells

Abstract

Genetic interactions mediate the emergence of phenotype from genotype. The systematic survey of genetic interactions in yeast showed that genes operating in the same biological process have highly correlated genetic interaction profiles, and this observation has been exploited to infer gene function in model organisms. Such assays of digenic perturbations in human cells are also highly informative, but are not scalable, even with CRISPR-mediated methods. As an alternative, we developed an indirect method of deriving functional interactions. We show that genes having correlated knockout fitness profiles across diverse, non-isogenic cell lines are analogous to genes having correlated genetic interaction profiles across isogenic query strains and similarly imply shared biological function. We constructed a network of genes with correlated fitness profiles across 276 high-quality CRISPR knockout screens in cancer cell lines into a “coessentiality network,” with up to 500-fold enrichment for co-functional gene pairs, enabling strong inference of gene function and highlighting the modular organization of the cell.

Coessential Genetic Networks Reveal the Organization and Constituents of a Dynamic Cellular Stress Response

“…Previous reports have demonstrated that genes with shared functions, particularly those which encode components of obligate protein complexes, have convergent effects on fitness when deleted in cancer cell lines (Kim et al, 2019;Pan et al, 2018). Consistent with prior analyses performed in few cell lines, network analysis of the strongest coessential relationships in the genome (r>0.6; 1532 pairings) successfully identifies modules of biological relevance with highly correlated fitness profiles ( Figure S1B).…”

“…Great insights have been gained from pairwise genetic screens in model organisms, most prominently budding yeast (Costanzo et al, 2016), but the larger human genome has proved challenging for genome-wide combinatorial study of genetic perturbations. With the advent of high quality CRISPR-Cas9 screening libraries in recent years, the coessentiality approach has emerged as an alternative to pairwise genetic perturbations for the discovery of novel genes and genetic interactions (Boyle et al, 2018;Kim et al, 2019;Pan et al, 2018). These first coessentiality studies detail a top-down approach which effectively resolves major protein complexes and identifies functional clusters in the genome.…”

“…If not, what mechanisms underlie the requirement for these factors in some cell lines but not others, and which other genes share this context-specific phenotype? Regarding the latter question, recent work has demonstrated that correlated patterns of gene essentiality ("coessentiality") across many cancer cell lines can indeed identify functional relationships between genes (Boyle et al, 2018;Kim et al, 2019;Pan et al, 2018;Wang et al, 2017). These studies utilize a top-down approach, organizing the genome into coessential clusters which have already proven to yield valuable insights.…”

Matching cell lines with cancer type and subtype of origin via mutational, epigenomic and transcriptomic patterns

“…The tissue that originated a tumor is well known to be a major determinant of drug responses -- including drugs targeted to certain genetic mutations -- both in vitro (11,12) and also in vivo (15,16). Tissue of origin is an important factor in shaping the networks of genetic interactions in cancer (17) and also determines the phenotypes resulting from genetic perturbation (18). Therefore ascertaining the tissue/cell type identity of cell lines is relevant for interpreting results of various experiments.…”

Biases and Blind-Spots in Genome-Wide CRISPR Knockout Screens

“…Targeted small molecules often don't discriminate, or discriminate poorly, between closely related paralogs, and it may be their promiscuity rather than their specificity that renders them effective. For example, MEK inhibitor trametinib effectively targets the protein products of both MAP2K1 and MAP2K2, redundant kinases downstream of RAS/RAF oncogenes, but the functional redundancy of these genes renders them both invisible to monogenic CRISPR screens, even in RAS/RAF backgrounds (Kim et al, 2019). A systematic survey of paralogs families, in particular those with only two family 460 members that are constitutively expressed, for synthetic lethality could unmask this genetic buffering and yield greater insight into the constellation of targetable genes in a tumor.…”