Predicting Gene Ontology Function of Human MicroRNAs by Integrating Multiple Networks

MicroRNAs (miRNAs) are small endogenous non-coding RNAs and take critical part in many human biological processes. Inferring the functions of miRNAs perhaps is an important strategy for understanding the pathogenesis of disease at the molecular level. In this paper, we propose an integrated model, PmiRGO, to infer the gene ontology (GO) functions of miRNAs by integrating multiple data sources including the expression profiles of miRNAs, the miRNA-target interactions and protein-protein interactions (PPI). PmiRGO starts with building a global network consisting of three networks. Then, it employs DeepWalk to learn latent representations as network features of the global heterogeneous network. Finally, the SVM-based models are applied to label the GO terms of miRNAs. Experimental results show that PmiRGO performs significantly better than the existing state-of-the-art method in terms of Fmax. Case study further demonstrates the feasibility of PmiRGO to annotate the potential functions of miRNAs.