Gene Prioritization for Type 2 Diabetes in Tissue-specific Protein Interaction Networks∗

Computationally prioritizing disease genes by large-scale bio-experimental data can provide important insights into the underlying mechanism of complex diseases. Here, we explore the topology of the protein-protein interaction network and apply the PageRank algorithm to identify candidate genes relating to type 2 diabetes. Importantly, a novel idea is introduced to rank the disease genes in tissue-specific protein-protein interaction networks instead of the global protein interaction network. To this end, we extend the original PageRank algorithm by adopting a block-based strategy. The leave-one-out cross validation is conducted to evaluate the performances of all ranking algorithms. The resulting ROC curves show that the proposed method with tissue-specific information performs better than original PageRank algorithm in the global protein-protein interaction network and each subnetwork of single tissue. Finally, four candidate genes are highlighted for further experimental validation due to their higher scores.