Additional file 14 of Systematic interrogation of mutation groupings reveals divergent downstream expression programs within key cancer genes

Additional file 14: Figure S13. Transferring mutation signatures across disease contexts. Models trained to predict the presence of mutations and their subgroupings in each cohort were applied to every other cohort in which the corresponding mutation was also present. (top) For each gene matching the criteria used in Additional file 13: Figures S12, we measured the performance of the gene-wide classifier according to the training cohort (x-axis) and the cohort they were transferred to (y-axis). (bottom) Where applicable, transfer AUC performance of the optimal subgrouping according to most frequent cv-significance relative to the gene-wide task across all cohorts in which subgroupings were enumerated. We include here these figures for TP53, PIK3CA, and NFE2L2 and for the remaining genes at our data portal under Figures/S13 - Gene Transfer AUC Heatmaps. The names of these figures have the format (gene)__transfer-aucs_Ridge.svg.