DeepWalk-Based Graph Embeddings for miRNA-Disease Association Prediction Using Deep Neural Network

Background: In recent years, micro ribonucleic acids (miRNAs) have been recognized as key regulators in numerous biological processes, particularly in the development and progression of diseases. As a result, extensive research has focused on uncovering the critical involvement of miRNAs in disease mechanisms to better comprehend the underlying causes of human diseases. Despite these efforts, relying solely on biological experiments to identify miRNA-disease associations is both time-consuming and costly, making it an impractical approach for large-scale studies. Methods: In this paper, we propose a novel DeepWalk-based graph embedding method for predicting miRNA-disease association (DWMDA). Using DeepWalk, we extracted meaningful low-dimensional vectors from the miRNA and disease networks. Then, we applied a deep neural network to identify miRNA-disease associations using the low-dimensional vectors of miRNAs and diseases extracted via DeepWalk. Results: An ablation study was conducted to assess the proposed graph embedding modules. Furthermore, DWMDA demonstrates exceptional performance in two major cancer case studies (breast and lung), with results based on statistically robust measures, further emphasizing its reliability as a method for identifying associations between miRNAs and diseases. Conclusions: We expect that our model will not only facilitate the accurate prediction of disease-associated miRNAs but also serve as a generalizable framework for exploring interactions among various biological entities.