Prediction of missing common genes for disease pairs using network based module separation

Identifying genes that are associated with two or more diseases can shed lights on understanding the pathobiological mechanisms of these diseases. In this work we present a novel method to predict missing common genes for disease pairs. The method formulates searching for missing common genes as an optimization problem to minimize a network based module separation between two subgraphs formed by mapping the disease associated genes onto the interactome. Tested on a dataset of more than 600 disease pairs using cross-validation, it is shown that the method achieves an average ROC score of 0.95.