A novel computational model based on super-disease and miRNA for potential miRNA–disease association prediction

Abstract

In recent years, more and more studies have indicated that microRNAs (miRNAs) play critical roles in various complex human diseases and could be regarded as important biomarkers for cancer detection in early stages. Developing computational models to predict potential miRNA–disease associations has become a research hotspot for significant reduction of experimental time and cost. Considering the various disadvantages of previous computational models, we proposed a novel computational model based on super-disease and miRNA for potential miRNA–disease association prediction (SDMMDA) to predict potential miRNA–disease associations by integrating known associations, disease semantic similarity, miRNA functional similarity, and Gaussian interaction profile kernel similarity for diseases and miRNAs. SDMMDA could be applied to new diseases without any known associated miRNAs as well as new miRNAs without any known associated diseases. Due to the fact that there are very few known miRNA–disease associations and many associations are ‘missing’ in the known training dataset, we introduce the concepts of ‘super-miRNA’ and ‘super-disease’ to enhance the similarity measures of diseases and miRNAs. These super classes could help in including the missing associations and improving prediction accuracy. As a result, SDMMDA achieved reliable performance with AUCs of 0.9032, 0.8323, and 0.8970 in global leave-one-out cross validation, local leave-one-out cross validation, and 5-fold cross validation, respectively. In addition, esophageal neoplasms, breast neoplasms, and prostate neoplasms were taken as independent case studies, where 46, 43 and 48 out of the top 50 predicted miRNAs were successfully confirmed by recent experimental literature. It is anticipated that SDMMDA would be an important biological resource for experimental guidance.