Functional similarities between microRNAs inferred from biomedical texts

Abstract

MicroRNAs miRNAs are involved in multiple biological processes, such as tumorigenesis and differentiation. The functions of most miRNAs still remain elusive. Measuring functional similarity between miRNAs is an important step to predict the functions of novel miRNAs and further identify disease-related miRNAs. In this study, we applied a biomedical text-mining method to assess miRNA functional similarities. According to validations, miRNA functional similarities inferred from biomedical texts are reliable and have the potential to distinguish disease miRNA pairs from random ones. Therefore, we further applied this set of similarity scores to uncover disease-related miRNAs, and achieved a high AUC of 0.941. Compared with existing methods, our set of miRNA functional similarity scores has higher reliability, larger coverage, and superior performance in prioritising disease-related miRNAs. We also conducted the case studies examining four common diseases and found that majority of the top ten candidates have been validated by experimental evidence.