To demonstrate the quality of their dataset, Cheung et al. subdivided their dataset along the mutational status of KRAS, BRAF and PIK3CA, genes frequently mutated in human cancers. Cells harboring such activated oncogenes frequently depend on their continued activity to maintain a malignant phenotype, a phenomenon called ‘oncogene addiction’ [ 15]. Reassuringly, comparing the phenotypes of mutant and wildtype cell lines consistently pinpointed the known oncogene – KRAS, BRAF or PIK3CA, respectively – as specifically required Compound Library supplier for cell growth only in the presence of the activating mutation. Next, the researchers split their dataset according to the cell lines’ tissue

of origin instead. Searching for genes required specifically for proliferation and/or survival of ovarian cancer cells revealed a set of ∼600 genes, a subset of which had previously been reported to be amplified or overexpressed in ovarian tumors (9.5%, 55/582). The differential phenotype of one of them, the transcription factor PAX8, was tested in eight ovarian cancer cell lines: six of them relied on PAX8 expression for continued growth. In an independent study, Brough et al. employed a similar strategy to identify differential growth and viability phenotypes in a panel of 34 breast ATM inhibitor cancer cancer cell lines [ 16•]. They recorded the effects

of targeting ∼700 kinases with pooled siRNAs and then split the dataset according to the cell lines’ genetic markers, including common amplification events (e.g. of the ERBB2 locus), known

mutations (e.g. in Inositol oxygenase PIK3CA) or clinical subtypes (e.g. ER+/ER−). The researchers identified multiple RNAi phenotypes specifically associated with cancer-associated genetic aberrations: For example, cells lacking functional copies of the tumor suppressor gene PTEN were particularly dependent on genes controlling the mitotic spindle assembly checkpoint and showed synthetic lethality with siRNAs as well as small molecule inhibitors targeting the checkpoint kinase TTK [ 16• and 17]. These examples highlight how the phenotypic differences within a panel of cell lines can reveal shared dependencies of tumor subtypes, potentially providing a highly selective set of candidate drug targets. Recently, this approach has also been applied to address a long-standing challenge in cancer research: how to kill tumors carrying mutations in the gene most frequently affected in human cancers – RAS? More than 30% of tumors carry mutations in members of the RAS small GTPase protein family, making NRAS, KRAS and HRAS the most commonly affected genes in human cancers [18]. Many cancer cell lines have also remained addicted to constant activity of the Ras-signaling pathway for maintaining a malignant phenotype, rendering RAS (and other pathway members including, for example, its downstream effector BRAF) highly attractive drug targets [19••].