Fig. 6

Heatmap displaying predictive performance for mutations in each of the 217 genes from the cancer-related gene set, across all six TCGA data modalities. Each cell quantifies performance for a target gene, using predictive features derived from a particular data type. Gray shaded dots indicate that the given data type provides significantly better predictions than the permuted baseline for the given gene; black inner dots indicate the same and additionally that the given data type provides statistically equivalent performance to the data type with the best average performance (determined by pairwise t-tests across data types with FDR correction)