DGCA-DTA: A Deep Graph Neural Network Based on Co-Attention for Drug Target Affinity Prediction

Accurately predicting drug-target affinity (DTA) is a critical step in drug discovery and design. By utilizing deep learning methods for DTA prediction, the drug development cycle can be shortened, and research and development costs can be reduced. Currently, graph neural networks (GNNs) are widely applied in DTA prediction. However, shallow GNNs are insufficient for capturing the overall structure and local features of compounds. Moreover, existing methods do not fully incorporate the interactions between drugs and their targets. To effectively process large-scale biomolecular data and capture intricate interaction patterns, we propose a deep graph neural network based on co-attention for DTA prediction, called DGCA-DTA. This model leverages a multi-scale graph neural network to extract drug features, enabling it to capture both the local and global structures of drug compounds. In addition, the model integrates a co-attention mechanism to learn higher-order interaction features between drugs and protein internal subspaces. Experimental results on two benchmark datasets show that the proposed model outperforms existing methods in DTA prediction.