Abstract
Conventional approaches to predict protein involvement in cancer often rely on defining either aberrant mutations at the single-gene level or correlating/anti-correlating transcript levels with patient survival. These approaches are typically conducted independently and focus on one protein at a time, overlooking nucleotide substitutions outside of coding regions or mutational co-occurrences in genes within the same interaction network. Here, we present CancerHubs, a method that integrates unbiased mutational data, clinical outcome predictions and interactomics to define novel cancer-related protein hubs. Through this approach, we identified TGOLN2 as a putative novel broad cancer tumour suppressor and EFTUD2 as a putative novel multiple myeloma oncogene.