Fig. 2

A An example of a simulated modulatory process: T>C mutation features are modulated downwards and T>A mutations are modulated upwards. B Simulated versus predicted mutation counts by additive model in one dataset (affected by the example modulatory signature). C Simulated versus predicted-modulated with the modulatory process mutation counts in the example dataset (corrected with the inferred modulatory process). D The cosine similarity of simulated and inferred modulatory signatures. E Simulated versus inferred modulatory process activity values for all 50 datasets (each dataset is colored differently)