Prediction of lncRNA-miRNA Interactions via an Embedding Learning Graph Factorize Over Heterogeneous Information Network.

An increasing number of studies show that identification of lncRNA-miRNA interactions (LMIs) helps the researchers to understand lncRNAs functions and the mechanism of involved complicated diseases. However, biological techniques for detecting lncRNA-miRNAs interactions are costly and time-consuming. Recently, many computational methods have been developed to predict LMIs, but only a few can perform the prediction from a network-based point of view. In this article, we propose a novel computational method to predict potential interactions between lncRNA and miRNA via an embedding learning graph factorize over a heterogeneous information network. Specifically, a large-scale heterogeneous information network is built by combing the associations among proteins, drugs, miRNAs, diseases, and lncRNAs. Then, a graph embedding model Graph Factorization is employed to learn vector representations for all miRNA and lncRNA in the heterogeneous network. Finally, the integrated features are fed to a classifier to predict new lncRNA-miRNA interactions. In the experiment, the proposed method performed good prediction results with AUC of 0.9660 under five-fold cross-validation. The experimental results demonstrate our method as an outperform way to predict potential associations between lncRNAs and miRNAs.