Abstract

In recent years there has been a growing trend towards the inclusion of diverse genomic information to support comprehensive large-scale prediction of protein–protein interaction networks. The Gene Ontology (GO) is one such functional knowledge resource, which consists of three hierarchies to describe functional attributes of gene products: Molecular function, biological process, and cellular component. Using Bayesian networks, this paper presents a framework for the probabilistic combination of semantic similarity knowledge extracted from the three GO hierarchies for analysis of protein–protein interaction networks and demonstrates its application in yeast. The results indicate that by integrating information encoded in the GO hierarchies a better result can be achieved in terms of both statistical prediction capability and potential biological relevance.