Potential microRNA-Disease Association Prediction Using Node2vec and Singular Value Decomposition

Many biological studies show that microRNAs (miRNAs) play an indispensable role in the regulation of various biological processes. MiRNAs are significant biomakers in disease diagnosis, aiding in the understanding of pathogenesis and facilitating the identification, diagnosis, and treatment of various diseases. However, the exact mechanism by which miRNAs influence the development of these diseases remains incompletely understood. Thus, it is crucial to develop a computational method to identify unknown miRNA-disease associations. In this study, we designed a computational framework based on singular value decomposition (SVD) and node2vec to predict unknown miRNA-disease associations (SNMDA). We use SVD technique to extract the linear features of miRNAs and diseases. The node2vec method is applied to learn the non-linear embeddings of miRNAs and diseases. We combine the linear feature and non-linear feature to get a new feature vector and feed it into the Gradient Boosting (GB) classifier for binary classification prediction. According to the experimental findings, SNMDA demonstrated an average area under the curve (AUC) of 0.9608 during five-fold cross-validation. Compared with the other Cutting-edge methods, SNMDA achieved the highest AUC value. Furthermore, the case studies on gastric cancer, malignant esophageal lesions, and lung tumors validate the effectiveness of SNMDA. The comprehensive experimental results demonstrate that SNMDA is effective in identifying unknown miRNA-disease associations.