LHNHLDA: A Novel Approach Based on LHN-2 Algorithm for Predicting Associations Between LncRNAs and Diseases

Abstract

With rapid development of high-throughput technique in the field of life science, LncRNAs are found to be inextricably linked to many diseases that seriously endanger human health. However, traditional experiment-based methods for inferring unknown diseases-LncRNA associations are time-consuming and laborious, therefore, it has become an effective way to adopt computational models to predict potential LncRNA-disease associations in recent years. In this article, a novel prediction model called LHNHLDA has been proposed. In LHNHLDA, based on known LncRNA-disease associations downloaded from benchmark databases, a heterogeneous LncRNA-disease network is built first by integrating LncRNA-LncRNA similarities with disease-disease similarities. And then, through adopting the LHN-2 algorithm, the path-based similarities between different nodes in the newly constructed heterogeneous network are obtained, which can be utilized to infer potential associations between LncRNAs and diseases. Finally, in order to evaluate the performance of LHNHLDA, intensive experiments have been done, and experimental results show that LHNHLDA can achieve reliable AUCs of 0.8155, 0.8281 and 0.8569 under the frameworks of 2-Fold CV, 5-Fold CV and LOO-CV respectively. Furthermore, case studies on lung cancer and leukemia illustrate that there are 10 and 7 potential LncRNAs out of the top 10 related LncRNAs of leukemia and lung cancer predicted by LHNHLDA having been confirmed to be linked to these two kinds of diseases by latest studies separately. Hence, due to the satisfactory prediction performance achieved by LHNHLDA, it is obvious that LHNHLDA may be a useful tool for future researches in the field of bioinformatics.