Fig. 5

Cohort-specific characterization of gene CNAs with a high impact on survival signature. All gene deletions and amplifications in each TCGA cohort (GBM, OV, and LUAD) with a high impact on the corresponding survival signature were grouped based on their mutation frequency shown in log-scale along the x-axes. a, b Percentage of genes in each bin belonging to the cohort-specific genes with a high impact on survival signature genes. More frequently mutated genes are more likely high-impact genes (p<0.03 for each cohort, one-sided correlation test). c, d Median impact of high-impact genes in each frequency bin. The median impact quantifies the contribution of all high-impact gene CNAs in a bin to the variation of the expression levels of all cohort-specific survival signature genes. Average percentages of explained variance of survival signature expression computed for all high-impact gene CNAs were used to determine the median impact per bin. e, f Proportion of known cancer genes among the high-impact genes in each bin. Known tumor suppressor and oncogenes are enriched among more frequently mutated genes with high survival impact (p<0.005 except for LUAD deletions, one-sided correlation test). CNA copy number alteration, GBM glioblastoma multiforme, LUAD lung adenocarcinoma, OV ovarian serous cystadenocarcinoma