Application of Multi-Feature Fusion Method Based on Neural Networks in Drug-Target Prediction
Zhijing Li, Hong Li, Yiping Jiang · 2024
Drug-target prediction plays a crucial role in drug discovery, aiding in better understanding of drug mechanisms and improving the efficiency and success rate of new drug development. However, traditional single-feature models have limitations in prediction, failing to fully exploit the complex relationships between drugs and targets. To address this issue, this study designs a multi-feature fusion system that combines bipartite graph, drug features, target features, and drug-target interaction information to enhance prediction performance. Specifically, multidimensional features including drug molecular fingerprints and descriptors, drug similarity, target protein sequence, structure, and functional annotation are employed. By leveraging these features comprehensively, our system achieves promising results on ECFP6 features, with an AUC value of 0.773. Compared to systems without multi-feature fusion, our system demonstrates a significant improvement, with an increase of approximately 0.05 in AUC. These findings underscore the effectiveness of multi-feature fusion in enhancing drug-target prediction accuracy and performance, providing valuable insights and support for advances in drug discovery and design. Our study offers new insights and methods for a deeper understanding of the complex relationships between drugs and targets, accelerating the process of new drug development.