A novel information fusion strategy based on a regularized framework for identifying disease-related microRNAs

Abstract

Abnormal microRNA (miRNA) expression can induce various complex human diseases. Thus, revealing the underlying relationship between miRNA and human diseases contributes to the early diagnosis and treatment of diseases. Utilizing a computational approach in selecting the most likely miRNA candidates related to a given disease for further biological experimental validation can save time and manpower costs. In this study, we propose a novel information fusion strategy called RLSSLP, which is based on a regularized framework, for discovering the underlying associations between miRNAs and diseases. RLSSLP integrates two submodels to construct effective prediction frameworks and quantify the similarities between miRNAs and diseases by fully using multiple omics data, which include verified associations, particularly miRNA–disease, miRNA–gene, and weighted gene–gene network associations. The 10-fold cross-validation and case studies for lung cancer, hepatocellular carcinoma and breast cancer indicate that RLSSLP performs well in predicting miRNA–disease interactions.