Fig. 1
From: Clustering cancers by shared transcriptional risk reveals novel targets for cancer therapy
Unbiased genetic analyses identify three distinct cancer clusters which may be targetable in a cluster-specific manner. A. Dimensional reduction and clustering of cancer types (full names provided in Table S1) based on transcriptional hallmark pathway expression and correlation with patient survival identifies three cancer subpopulations. B. Summary of the detrimental genetic pathways enriched in the ‘inflammatory cluster’ (orange), the ‘metabolic cluster’ (blue), and the ‘proliferative cluster’ (black). C. 5-year overall KM survival curves for patients assigned to each cluster. D. Drug prediction statistics for the leading compound, AZ-628, which is predicted to specifically rescue the deleterious gene expression profile associated with inflammatory cancers (top subpanel). In vitro validation statistics (reversal score) for AZ-628 demonstrates benefit in a representative inflammatory breast cancer cell line (MDA-MB-231), but no impact on a representative proliferative lung cancer cell line (A549), nor a representative metabolic hepatocellular cancer cell line (HepG2, bottom subpanel). E. Propensity-matched pharmacovigilance studies (matched on demographics, smoking status, comorbid conditions, procedures, and therapeutics in the 6 months leading up to enrollment) demonstrate the 5-year incidence of each cancer cluster amongst individuals prescribed clopidogrel, an FDA-approved drug predicted to specifically reduce inflammatory cancers