Using Protein Interactome Similarity to Improve Random Walk with Restart Model for Drug Repurposing

Abstract

Development of new drugs has the limitations of high cost, high time requirement and low success rate. If an existing drug can be used to treat a drug-seeking disease, we can reduce these limitations. The process of using existing drugs to treat new diseases is called drug repurposing. Since the drugs are already approved for human use, the success rate becomes high. In recent years, numerous random walk models have been proposed on the disease-drug heterogeneous network and become a popular drug repurposing approach. The performance of random walk-based approach depends on the network similarity measures used to build the heterogeneous network. In this paper, we improve the network similarity measures by integrating the similarity between the disease and drug-specific protein interactomes in human Protein-Protein Interaction network. We then run a random walk with restart algorithm over the modified network to predict disease-drug relations. Our experiments reveal that performance of random walk model has improved after integrating protein interactome similarity.