GraphMS: Drug Target Prediction Using Graph Representation Learning with Substructures

The prediction of drug–target interactions is always a key task in the field drug redirection. However, traditional methods predicting are either mediocre or rely heavily on data stacking. In this work, we proposed our model named GraphMS. We merged heterogeneous graph information and obtained effective node substructure based mutual embeddings. then learned high quality representations for downstream tasks, an end–to–end auto–encoder to complete link prediction. Experimental results show that method outperforms several state–of–the–art models. can achieve area under receiver operating characteristics (AUROC) curve 0.959 precise recall (AUPR) 0.847. found between graph–level contributes most index relatively sparse network. And node–level dense