Drug-Target Interaction Prediction Based on Deep Feature Fusion

Abstract

Drug-target interaction prediction plays an important role in drug discovery and drug repositioning. Traditional experimental screening methods are expensive and time-consuming, while existing computational approaches still have limitations in molecular representation and feature fusion. To address these problems, this paper proposes a drug-target interaction prediction model based on deep feature fusion. In the GCT-DTI, a Graph Isomorphism Network is used to capture the topological structure information of drug molecules, and Morgan fingerprints are introduced to supplement global chemical features. For target proteins, a Transformer encoder is employed to learn long-range dependency information from amino acid sequences. In addition, a cross-attention mechanism is designed to enhance the interaction between drug features and protein features. Experiments on the Human and C.elegans benchmark datasets demonstrate that the GCT-DTI achieves good performance in terms of AUC, Precision, and Recall. The results show that deep feature fusion can effectively improve the prediction accuracy of drug-target interactions and provide useful support for computer-aided drug discovery.