Prioritizing candidate disease-related long non-coding RNAs by walking on the heterogeneous lncRNA and disease network

Accumulated evidence has shown that long non-coding RNAs (lncRNA) act as a widespread layer in gene regulatory networks and are involved in a wide range of biological processes. The dysregulation of lncRNA has been implicated in various complex human diseases. Although several computational methods have been developed to predict disease-related lncRNA, this still remains a considerable challenging task. In this study, we tried to construct an lncRNA-lncRNA crosstalk network by examining the significant co-occurrence of shared miRNA response elements on lncRNA transcripts from the competing endogenous RNAs viewpoint. As expected, functional analysis showed that lncRNA sharing significantly enriched interacting miRNAs tend to be involved in similar diseases and have more functionally related flanking gene sets. We further proposed a novel rank-based method, RWRHLD, to prioritize candidate lncRNA-disease associations by integrating three networks (miRNA-associated lncRNA-lncRNA crosstalk network, disease-disease similarity network and known lncRNA-disease association network) into a heterogeneous network and implementing a random walk with restart on this heterogeneous network. We used leave-one-out cross-validation to test the performance of this rank-based method in this study based on known experimentally verified lncRNA-disease associations and obtained a reliable AUC value of 0.871, which is much higher than RWR merely based on an lncRNA network, hypergeometric test and random situation. Furthermore, several novel lncRNA-disease associations predicted in case studies of ovarian cancer and prostate cancer have been confirmed in new studies by literature surveys.