LCBNI: link completion bipartite network inference for predicting new lncRNA-miRNA interactions

Abstract

LncRNAs and miRNAs are two different kinds of non-coding RNAs and are both important for human in the field of health and disease. LncRNAs can interact with miRNAs, and the interactions play key roles in gene regulatory networks. Predicting IncRNA-miRNA interactions is an urgent and significant task and can help to explore the mechanism of involved complicated diseases, but very few computational methods are developed. In this paper, we introduce a computational method named link completion bipartite network inference (LCBNI) to predict the potential interactions between IncRNAs and miRNAs. LCBNI formulates the observed IncRNA-miRNA interactions as a bipartite network. Considering that there is no any known interaction for new IncRNAs or miRNAs, LCBNI calculates the sequence similarity and utilizes weighted nearest neighbor interaction information to construct new interaction scores for these IncRNAs and miRNAs. Then, we implement a resource allocation algorithm on the bipartite network to predict IncRNA-miRNA interactions. The experimental results demonstrate that LCBNI can effectively predict IncRNA-miRNA interactions with higher accuracy compared with other state-of-the-art methods and network-based methods, including random walk with restart (RWR), IncRNA-based collaborative filtering (LncCF) and miRNA-based collaborative filtering (MiCF). Furthermore, case studies are performed to demonstrate the prediction capability of LCBNI using real data.