New Approaches of Protein Function Prediction from Protein Interaction Networks

Functional annotation of proteins is vital to biological and clinical research, as well as other applications, due to the important roles proteins play in various biological processes. Although the functions of some proteins have been annotated via biological experiments, there are still many proteins whose functions are yet to be annotated due to the limitations of existing methods and the high cost of experiments. To overcome the experimental limitations, computational approaches have been rapidly developed for proteins function prediction. On the other hand, new high-throughput technologies have generated massive protein interaction data of many species, which makes it possible to use computational methods to analyse the relationships among proteins and predict protein functions. However, how to interpret the protein interaction data and then effectively predict protein functions still remains a big challenge. Although there have been some survey papers and books about PPI based protein function prediction, there are currently no information sources that systematically discuss the critical aspects of PPI network based protein function prediction, including semantically assessing the reliability of PPI data, measuring the functional similarity between proteins, dynamically selecting prediction domains and predicting functions, and establishing corresponding prediction frameworks. Provides innovative approaches and new developments targeting key issues in protein function predictionPresents heuristic ideas for further research in this challenging area