Studies show that mutations or abnormalities in micRNAs can lead to many diseases, and the identification of disease-associated microRNAs (miRNAs) can help diagnose and treat related diseases. However, it is costly and long-term to obtain accurate correlations through biological experiments. Therefore, this paper proposes a machine learning method (HNDLM) that uses network topology information to predict disease-miRNA associations. HNDLM avoids the construction of similarity networks, but applies the network embedding method proposed in recent years to biological networks. Experimental results show that HNDLM performs better than MIDPE, MIDP, WBSMDA, RLSMDA, CPTL, HDMP classical algorithms in accuracy and AUC value. In case study, the top 30 candidate miRNAs recommended by HNDLM can be confirmed by previous experiments. HNDLM can discover the potential disease-miRNA relationship and help to further study the pathogenesis of the disease, promote the development of bioinformatics. © 2020, Chinese Institute of Electronics. All right reserved.
