NMGMDA: A Computational Model for Predicting Potential Microbe–Drug Associations based on Minimize Matrix Nuclear Norm and Graph Attention Network

Abstract

Abstract For drug research and development, the probable microbe-drug associations can be predicted with considerable utility. Deep learning-based techniques have recently found widespread use in the biomedical industry and have significantly improved identification performance. Additionally, the growing body of knowledge on germs and pharmaceutical biomedicine offers a fantastic potential for methods based on deep learning to forecast hidden associations between microbes and drugs. In order to infer latent microbe-drug associations, we developed a unique computational model in this publication called NMGMDA based on the nuclear norm minimization and graph attention network. We created a heterogeneous microbe-drug network in NMGMDA by fusing the drug and microbe similarities with the established associations between drugs and microbes. Then, in order to get predicted scores of potential microbe-drug associations, we used the nuclear norm minimization approach and a GAT-based auto-encoder, respectively. The final results, which are based on two datasets and weighted average of these two predicted scores, demonstrated that NMGMDA can outperform state-of-the-art competitive approaches. Case studies further demonstrated its capacity to reliably find fresh associations.