Predicting MicroRNA-Disease Associations Using Kronecker Regularized Least Squares Based on Heterogeneous Omics Data

MicroRNAs (miRNAs) play critical roles in many biological processes. Predicting the miRNA-disease associations will aid in deciphering the underlying pathogenesis of human polygenic diseases. However, existing in silico prediction methods typically utilize a single or limited data sources for disease-related miRNA prioritization and most of the methods are biased toward known miRNA-disease associations. Due to the insufficient number of experimentally validated interactions as well as no experimentally verified negative samples, obtaining remarkable performances is still challenging for these methods. In this paper, we present a semi-supervised method of Kronecker regularized least squares for predicting the potential or missing miRNA-disease associations (KRLSM). KRLSM integrates different omics data to assist various diseases or miRNAs with sparsely known associations to make predictions, and combines the disease space and miRNA space into a whole miRNA-disease space by Kronecker product. Finally, the semi-supervised classifier of regularized least squares is adopted to identify disease-related miRNAs. The experiment results demonstrate that the proposed method outperforms the other state-of-the-art approaches. In addition, case studies of several common diseases further indicate the effectiveness of KRLSM to identify potential miRNA-disease associations.