Cancer cell lines have numerous characteristics that make them favorable pre-clinical research models, yet they are notoriously poor at predicting drug response in the clinic. Here we sought to investigate the utility of synthetic lethality (SL) interactions discovered from large-scale CRISPR functional screens (i.e. the BROAD and Sanger Cancer Dependency Maps or "DepMap") as predictors of targets that validate in patients. Mutual exclusivity, the phenomenon where two genes are rarely mutated together in the same tumor, is a powerful clinical-stage readout that can be caused by synthetic lethality. We found that SL interactions discovered in DepMap are significantly more likely to be mutually exclusive in TCGA when they include a driver (tumor-suppressor/oncogene). These SL interactions represent high-value targeting opportunities with the advantage of clear patient selection criteria based on their driver mutation status. In an effort to identify drugs that target these proteins as potential repurposing opportunities, we found that pharmacogenomic inhibition rarely invokes the same target dependencies as a genetic deletion of the drug target. Nonetheless, we identified several dozen "clean" drugs with potential for repositioning and validated the top candidates in PDx. Although tumours are more heterogenous than cancer cell lines, we show that cell line viability readouts linked to single-gene/drug perturbations can yield accurate predictions of clinical efficacy when tied to tumor-driver biology.

Citation Format: Tomas Babak, Michael Vermeulen, Doris Coto Villa, Andrew Craig. Driver-gene dependencies reveal clinically actionable drug repositioning opportunities [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4035.