Protein function prediction based on protein-protein interaction together with sequence and structural similarity information

Abstract

Recent advances in experimental biology makes large amounts of protein-protein interaction (PPI) data available. Thus, using PPI data to functionally annotate proteins has been extensively studied. But if there is not enough information about annotation available in the network, most existing network-based approaches do not work well. In a recent interaction network based research work proposal has been made to combine PPI data and sequence similarity information to boost up the prediction performance. But we know that structural similarity is much more affective for predicting protein functions, because protein structure is far more conserved than sequence. Here we have proposed to use structural similarity information together with PPI data and sequence similarity information for predicting protein function. Our method divides function prediction into two phases: first, the original PPI network is enriched by adding a number of implicit edges that are inferred from protein sequence and structural similarity information. Second, a collective clas­sification algorithm is employed on the new network to predict protein function. The experimental results support our assumption and provide better function prediction results than method with PPI and sequence similarity information only.