Construction of multi-relation protein networks and its application

Considering the different influences of different types of interactions in protein function prediction, we construct a multi-relation interaction network by integrating protein-protein interaction (PPI) network and proteins' domain information and provide different priorities for different types of interactions. We propose a protein function prediction method, named functions prediction based on multi-relation networks (FPM) to infer protein functions. Given an unannotated protein, all interactions with highest priority associated with the protein are visited to form a candidate neighbor nodes set. Finally, we sort all functions of the neighbor nodes and select part of them to predict an unknown protein. Comparson with other methods show that our method outperforms other function prediction methods. Copyright © 2015 Acta Automatica Sinica. All rights reserved.