Prediction of Protein Function Using Gaussian Mixture Model in Protein-Protein Interaction Networks

Predicting protein function is one of the most important problems in the post-genomic era. Recent high-throughput experiments have determined proteome-scale protein physical interaction maps for several organisms. In this paper, a new method, which is based on Gaussian Mixture Model, is introduced to predict protein function from protein-protein interaction data. In the proposed method, A global information are taken into account by representing a protein using all the functional annotations of all proteins assigned with that term and have a shortest path with target protein in the all protein interaction network. We apply our method to a constructed data set for yeast and fly based upon protein function classifications of GO scheme and upon the interaction networks collected from IntAct protein-protein interaction. The results obtained by leave-one-cross-validation test show that the proposed method can obtain desirable results for protein function prediction and outperforms some existing approaches based on protein-protein interaction data.