Boosting Drug-Target Binding Affinity Predictions with a Novel Three-Branch Convolutional Neural Network Approach

The process of discovering new drugs is costly and time-consuming, with safety concerns often arising. Deep learning has become a mainstream ap-proach in computer-aided drug design, with convolutional neural networks CNN and graph neural networks GNN playing a significant role in drug-target affinity DTA prediction. This paper introduces a novel method for predicting DTA using a combination of graph convolutional networks and a three-branch multiscale CNN, leading to significant improvements in predic-tion accuracy.