EpiTracer - an algorithm for identifying epicenters in condition-specific biological networks

Abstract

Diseases in biological systems may result from small perturbations in a complex network of protein-protein interactions (PPIs). The perturbations typically affect a small set of proteins, which then go on to disturb a larger part of the network. Biological systems attempt to counteract these perturbations by launching a stress-response, resulting in a complex pattern of variations in the cell. We present an algorithm, EpiTracer which identifies the key proteins, termed epicenters, from which a large number of the changes in PPI networks ripple out. We propose a new centrality measure, ripple centrality, that measures how effectively a change at a particular protein can ripple across the network, by identifying condition specific highest activity paths obtained by mapping gene expression profiles to the PPI network. We perform a case study on a dataset (E-GEOD-61973) where the gene PARK2 was intentionally overexpressed in human glioma (U251) cell line and analyze the top 10 ranked epicenters. We find that EpiTracer identifies PARK2 as the most important epicenter in the perturbed condition. Analysis of the other top-ranked epicenters showed that all of them were involved in either supporting the activity of PARK2 or counteracting it, indicating that the cell had activated a stress-response. We also find that 5 of the identified epicenters did not have significant differential expression, proving that our method is capable of finding information that simple differential expression analysis cannot.