A random walk-based method for detecting essential proteins by integrating the topological and biological features of PPI network

The essential protein detection on protein–protein interaction (PPI) network can not only promote the research of life science, but also have important applications in disease diagnosis and drug target cell identifying. A large number of computation-based essential protein detection algorithms have been presented recently. Most of those methods detect the essential proteins according to the centrality measures of the nodes in PPI networks. Those centrality-based essential protein detection methods only consider the topological information of the PPI networks and neglect the biological features of the proteins which are crucial in recognizing the essential proteins. This paper presents a random walk-based method named EPD-RW to identify essential proteins by integrating network topology and biological information extracted from GO (gene ontology) data, gene expression profiles, domain information and phylogenetic profile. EPD-RW uses both the topological structure of the PPI and biological information of the proteins to guide the random walk for computing their essentialness. An iterative method is presented to efficiently integrate the topological and biological features at each step of the random walk. We test our method EDP-RW by experiments on yeast PPI datasets. We also compare the test results of EDP-RW with those of other methods. The experimental results demonstrate that EPD-RW can achieve the best performance among all the methods tested. The biological illustration of the results shows that our random walk-based method effectively increases the accuracy of essential proteins detecting results, and the biological features of the proteins can greatly enhance the performance of essential protein detecting.