Improving Identification of Drug-Target Binding Sites Based on Structures of Targets Using Residual Graph Transformer...

Area of Science:

• Computational biology
• Drug discovery
• Bioinformatics

Background:

• Accurate identification of drug-target binding sites is crucial for efficient drug screening and design.
• Current prediction models face challenges due to insufficient multimodal information fusion and imbalanced datasets.

Purpose of the Study:

• To develop a novel deep learning framework, RGTsite, for enhanced drug-target binding site prediction.
• To overcome limitations of existing methods by effectively integrating diverse target information.

Main Methods:

• Utilized a Residual Graph Transformer Network (GTN) integrated with a residual 1D convolutional neural network (1D-CNN) and ProtT5 for feature extraction.
• Combined local, global sequence features, and physicochemical properties as graph vertex features.
• Incorporated edge features and applied GTN for comprehensive feature extraction and binding site classification.

Main Results:

• RGTsite demonstrated superior performance compared to state-of-the-art methods on benchmark datasets.
• Achieved higher scores in key metrics like F1-score (F1) and Matthews Correlation Coefficient (MCC).
• Interpretability analysis confirmed RGTsite's effectiveness in practical drug-target binding site identification.

Conclusions:

• RGTsite offers a significant advancement in predicting drug-target binding sites.
• The framework effectively addresses multimodal information fusion and dataset imbalance challenges.
• RGTsite holds promise for accelerating the drug development pipeline through accurate binding site identification.