Identifying functional miRNA-mRNA modules based on hypergraph-based learning

1 Background and problem Analyzing functional relationships between microRNAs (miRNAs) and mRNAs is an important issue in biological process because it can give new insights into the pathogenesis mechanism of complex diseases including various cancers. Especially, miRNAs emerge recently as one of crucial molecules in post-transcriptional regulation and play critical roles in diverse processes such as tumorigenesis by regulating target mRNAs [1]. Thus, identifying a group of miRNAs and mRNAs as a module is essential for the discovery of their combinatorial effects on different physiological and pathological conditions. To investigate such issues, various approaches have been suggested to detect miRNA-mRNA regulatory relationships in recent years [2-3]. However, the discovery of relevant miRNA and mRNA regulatory modules still remains a major challenge in biology due to the complexity of the issue. Here we introduce a hypergraph-based model for identifying miRNA-mRNA modules underlying specific regulatory conditions from expression data. The proposed model represents explicit higher-order interactions among many features, thus facilitating the analysis of complex biological phenomena. We evaluate the proposed model on a prostate cancer dataset, and show the discovery of significant miRNA-mRNA regulatory modules involved in specific cancer processes. The biological significance of the identified miRNA-mRNA modules is confirmed by gene ontology analysis and literature reviews. Figure 1 outlines the proposed approach to identifying miRNA-mRNA modules on specific regulatory conditions.