MDGAE-DTI: Drug-Target Interactions Prediction Based on Multi-information Integration and Graph Auto-Encoder

Computational strategies for identifying drug-target interactions (DTIs) can improve the efficiency of drug development. With the advancement of modern biotechnology, a vast amount of biomedical data has been accumulated. There are multiple sources of data for DTIs research, and it has become a challenge to integrate multiple sources of data. By integrating multiple sources of information, we can more easily predict DTIs and identify targets more quickly. To efficiently reposition drugs, we propose MDGAE-DTI method, a network-based computational technique that precisely predicts DTIs throughout a multi-information network. The proposed method makes use of a couple of similarity matrices for drugs and targets to seize their respective similarity. These matrices are subsequently integrated to generate feature vectors for drug and target nodes in a graph-based neural network. The preliminary facets are based primarily on node similarity, and a multi-layer perceptron is employed to update node features. Finally, the representations of drugs and target proteins are combined in a bilinear decoder to enable accurate prediction of DTIs. MDGAE-DTI achieved satisfactory prediction results, surpassed the performance of other comparison methods, and effectively predicted DTIs.