Sequence-derived linear neighborhood propagation method for predicting lncRNA-miRNA interactions

Abstract

RNA-miRNA interactions play key roles in gene regulatory networks. Predicting novel lncRNA-miRNA interactions can advance the progress of understanding the functions of lncRNAs and the mechanism of involved complicated diseases, but very few computational methods have been developed for predicting lncRNA-miRNA interactions. In this paper, we propose a computational method named sequence-derived linear neighborhood propagation method (SLNPM) to predict the novel interactions between lncRNAs and miRNAs, especially for lncRNAs and miRNAs which do not have any known interaction. First, SLNPM fully exploits lncRNA sequences, miRNA sequences and known interactions to calculate lncRNA-lncRNA similarities and miRNA-miRNA similarities by using fast linear neighborhood similarity approach. Then, SLNPM integrates multiple lncRNA-lncRNA similarities and multiple miRNA-miRNA similarities respectively to make use of diverse information, and constructs the integrated lncRNA similarity-based graph and the integrated miRNA similarity-based graph. Finally, SLNPM implements the label propagation process on graphs to score lncRNA-miRNA pairs, and adopts the linear combination of their outputs as final predictions. The experimental results demonstrate that SLNPM can predict lncRNA-miRNA interactions with higher accuracy compared with other state-of-the-art methods. We also analyze the prediction capability of SLNPM for lncRNAs or miRNAs without any known interaction, and the results indicate that SLNPM helps to find novel interactions, which do not exist in our dataset. Moreover, our studies reveal that the known interactions provide the most important information for the lncRNA-miRNA interaction prediction, and the incorporating sequence information further improves the performance.