Topology-based scoring method for identification of responsive protein-protein interaction subnetwork

In this article, we propose a novel topology-based scoring and searching approach to extract protein protein interaction (PPI) subnetworks responsive to conditions being investigated according gene expression profiles. Each subnetwork is scored using both the activity of individual vertices and the network topology, instead of summing up only the activity of vertices as done in previous works. Using simulated data we demonstrate the advantage of the proposed method when the subnetworks contain highly significant hub genes, or high number of co-activated gene pairs. When applied to the sample data galFiltered in Cytoscape, our algorithm identified several biologically meaningful subnetworks that contain genes missed by vertex based scoring approach. Lastly we apply the new method to a human prostate cancer dataset and show that it can efficiently identify disease-relevant protein interactions. The new method is implemented as a Cytoscape Plugin jActiveModuesTopo. It is publicly available and can be used on most types of gene expression data.