Gene-related multi-network collaborative deep feature learning for predicting miRNA-disease associations

A growing number of experiments have shown that microRNAs (miRNAs) play a key role in regulating gene expression, and their aberrant expression may lead to the development of specific diseases. Therefore, accurate identification of the associations between miRNAs and diseases is crucial for the prevention, diagnosis and treatment of miRNA-related diseases. However, existing models have limitations in accurately capturing biological information and comprehensively extracting features. To address this problem, we propose gene-related multi-network collaborative deep feature learning for predicting miRNA-disease associations (MNFLMDA). First, we constructed three heterogeneous networks, miRNA-gene, disease-gene and miRNA-disease, and mined the potential information of the heterogeneous networks using Auto-Encoder and Graph Attention Networks. Subsequently, this potential information was fused to form the final features. Finally, these features were used to predict the associations between miRNAs and diseases. To validate the effectiveness of the model, we conducted extensive experiments on the Human miRNA Disease Database and compared it with eight of the most representative models over the past two years, and the results showed that MNFLMDA exhibits excellent performance. In addition, case studies of breast tumors, colorectal tumors and hepatocellular carcinoma were conducted to further validate the predictive performance of MNFLMDA.