Disease phenotype similarity improves the prediction of novel disease-associated microRNAs

Abstract

Many studies have shown roles of miRNAs (microRNAs) on human disease and a number of computational methods have been proposed to predict such associations by ranking candidate microRNAs according to their relevance to a disease. Among them, network-based methods are becoming dominant since they well exploit the “disease module” principle in miRNA functional similarity networks. Of which, Random Walk with Restart (RWR) algorithm-based method on a miRNA functional similarity network, namely RWRMDA, is state-of-the-art one. The use of this algorithm was inspired from its success in predicting disease gene because “disease module” principle also exists in protein interaction networks. Besides, many other algorithms were also designed for prediction of disease genes. However, they have not yet been utilized for disease microRNA prediction. In this study, we proposed a method, namely RWRHMDA, for prediction of disease-associated miRNAs. This method was based on RWRH algorithm, which was successfully proposed for disease gene prediction on a heterogeneous network of genes and disease phenotypes. In particular, we used this algorithm to rank disease candidate miRNAs on a heterogeneous network of phenotypes and miRNAs, which was constructed by integrating a shared target gene-based microRNA functional similarity network and a disease phenotype similarity network. Comparing the prediction performance of RWRHMDA with that of RWRMDA on a set of 35 disease phenotypes, we found that RWRHMDA significantly outperformed RWRMDA irrespective of parameter settings since it better exploited “disease module” principle. In addition, using RWRHMDA method, we identified eight novel Alzheimer's disease-associated miRNAs.