Fig 1. Data fusion framework.

Sequencing data are collapsed to calculate their mutational loads using four ROIs, namely genes, pathways, domains and PPIs. This allows studying ROI-phenotype associations along the four correspondent axes. Each element tested for association then becomes a feature for a prediction model. Single ROI types are combined to create data sets. Each data set is split into a training and test set. The training set is used to tune the learning parameters of a RF model and then select the best set of features, while the test set is used to measure the prediction performances.