Bipartite graph-based integrative method to detect consistent protein functional modules from multiple sources

A bipartite graph-based cluster ensemble method that integrates gene ontology (GO) and gene expression data with protein-protein interaction (PPI) networks is proposed. In this method, all different views of biological information and three basic clustering methods are contributed to a bipartite graph that comprehensively represents the relationships between the objects in this problem, including the proteins and the meta-clusters from the basic cluster methods. Furthermore, consistent modules are extracted using a symmetric non-negative matrix factorization (NMF)-based graph partition method and overlapping results are achieved. Extensive experimental results show that this method is superior to the baseline methods; further analysis is addressed to discuss the benefits of integrating multiple biological information sources and diverse clustering methods.