TransGAT-DTI: Transformer and Graph Attention Network for Drug-Target Interaction Prediction.

Area of Science:

• Computational chemistry
• Pharmacology
• Bioinformatics

Background:

• Drug-target interaction (DTI) prediction is crucial for drug discovery, development, and repurposing.
• Experimental DTI prediction methods are costly and time-consuming.
• Existing machine learning approaches face challenges in representing spatial structures and modeling local interactions.

Purpose of the Study:

• To propose a novel framework for accurate and interpretable DTI prediction.
• To address limitations in representing drug and target spatial features.
• To effectively model local interactions for improved prediction and interpretation.

Main Methods:

• Developed TransGAT-DTI, a framework combining transformer and graph attention convolutional networks.
• Utilized benchmark datasets (BindingDB, BioSNAP, Human) for validation.
• Focused on representing spatial structures and modeling local interactions.

Main Results:

• TransGAT-DTI achieved high overall performance on three benchmark datasets.
• The framework accurately predicts putative drug-target interactions.
• Identified pivotal sequences contributing to positive predictions, enhancing interpretability.

Conclusions:

• TransGAT-DTI offers a powerful computational solution for DTI prediction.
• The approach demonstrates superior performance compared to existing methods.
• Provides a novel strategy for accelerating drug discovery and development.