Enhancing Drug-Target Affinity Prediction Using Dynamic Attention Network

Drug-Target Affinity (DTA) prediction, which is crucial for drug discovery, computes the binding affinity between drugs and target proteins. While traditional experimental approaches are costly and time intensive, computational models are fast and accurate. In this study, we integrate a dynamic graph attention network (GATv۲) in combination with a graph convolutional network (GCN) to extract the most significant features from drug graphs. Moreover, proteins are modeled using a convolutional neural network (CNN). The complex relationships between drugs and targets are discovered based on deep learning techniques. The proposed model predicts the interaction strength between new drug-target pairs based on the computed representations of drugs and targets. Evaluated on the Davis dataset, our model achieves MSE of ۰.۲۴۹ and CI ۰.۸۹۱, outperforming several state-of-the-art models.

کلیدواژه ها:

Drug-target affinity prediction ، Dynamic Graph attention network ، Graph Convolutional Neural Network