Skip to main content
. 2021 Aug 20;7(34):eabh1275. doi: 10.1126/sciadv.abh1275

Fig. 7. Clinically relevance of predicted tumor dependencies.

Fig. 7

(A) Top 10 dependencies associated with chemoresponse in BRCA. Statistical significance is assessed by the t test comparing predicted tumor dependency scores of each DepOI between samples achieving CR and PD after chemotherapy. (B) NDUFS5, the most significant chemoresistance-associated dependency in BRCA. (C) Associations of chemoresistance-associated dependencies (Chemo Deps) with GO terms of mitochondrion and OXPHOS in BRCA. Significance is assessed by a one-tailed Fisher’s exact test for enrichment. (D and E) Top OS-associated gene dependencies and their hazard ratios (HR) identified by a univariate Cox proportional hazard regression model (Benjamini-Hochberg FDR, <0.2). A DepOI with a more negative dependency score, which indicates stronger dependency, associated with better (or worse) survival is termed as a protective (or adverse) dependency. Thus, a dependency score with a hazard ratio of >1 indicates a protective dependency. Full results of the univariate and multivariate Cox models are provided as table S16. (F and G) Pan-cancer significance of IL2 and Kaplan-Meier curves of uveal melanoma (UVM) and BRCA (see fig. S16 for other cancer types). Cancers with a Cox FDR of <0.2 are labeled in bold. CI, confidence interval; mth, month. (H and I) Significance of SMAD4 dependency and Kaplan-Meier curves of lower-grade glioma (LGG) and kidney chromophobe (KICH) (see fig. S17 for other cancer types).