Fig. 8

Correlation between data heterogeneity and clinical, molecular features on the predictable and unpredictable sets. (a) The pie chart shows the proportion of clinical and pathological features between the predictable and unpredictable sets in the pan-cancer dataset. The correlation between features is revealed through Spearman correlation coefficient and t-test. (b) The violin plot of the distribution of predictable and unpredictable sets for the same molecular pathological features. (c, d) The heatmap of the correlation between the features of the predictable and unpredictable sets. (e, f) The boxplot of feature importance scores calculated using random forest algorithm.