Consistent protein functional module detection from multi-view of biological data

Detecting functional modules from protein-protein interaction networks (PPINs) is an active research area with many practical applications. To date, multiple biological data sources are available such as gene expression data and gene ontology (GO). These data explain the biological roles of proteins from different views and provide additional information to alleviate false information in PPINs. This work focuses on extracting consistent information from diverse data sources. To address this problem, this work proposes a collective non-negative matrix factorization (CoNMF) method which efficiently integrates views of gene ontology, gene expression data and PPINs. In our method, the integration problem is reduced to optimimum approximations of multi-view data by the productions of their common matrix factor with basis matrices. As a result, the common matrix factor provides an intuitive interpretation of soft clustering. Extensive experiments show that CoNMF outperforms most of the baseline methods listed in the paper and is an effective method to extract functional modules in PPINs.