Inferring Pathways from Protein-protein Interaction Networks and Gene Expression Data

Biomedical researchers have long studied biological pathways in hopes of further understanding the effect of disease on critical events and interactions within these pathways. In recent years, new bioinformatic approaches to pathway prediction have been developed, taking advantage of the availability of vast amounts of microarray data and efficient machine learning techniques. Recent years have also led to the large-scale identification of protein-protein interaction (PPI) networks and transcription-factor-DNA (TF-DNA) interaction datasets. These large datasets each provide unique information on gene interactions, but when used alone can result in a high rate of false positive predictions. To address this problem, we present a novel approach to pathway prediction which integrates knockout microarray data with protein interaction networks by modeling the two data sets as an electrical circuit. The Electric Circuit Model achieved 99% specificity and up to 69% sensitivity in predicting the pheromone signaling pathway.