Drug-Target Affinity Prediction Based on Graph Representation and Attention Fusion Mechanism

Predicting drug-target affinity is crucial in the field of drug discovery. To further improve the accuracy of predictions, this paper proposes a drug-target affinity prediction model, GRAM-DTA, based on graph representation and attention fusion mechanisms. The model represents the input features of drugs and targets as graph data and utilizes deep graph isomorphism networks and graph neural network modules combining graph convolutional networks and graph attention networks to process the feature information of drugs and targets, respectively. In the feature fusion stage, an attention mechanism is introduced to simulate the interactions between drug molecules and amino acids, dynamically adjust the importance of features, and capture the interaction patterns between drugs and targets. Experiments were conducted on the Davis and KIBA benchmark datasets, and the model was compared with current state-of-the-art models.The experimental results show that our model achieved a 3.1% and 3.4% improvement in the r_m^2 value compared to the best-performing baseline model, significantly outperforming other traditional methods and baseline models.