AIGO-DTI: Predicting Drug-Target Interactions Based on Improved Drug Properties Combined with Adaptive Iterative...

Area of Science:

• Computational chemistry
• Bioinformatics
• Artificial intelligence in drug discovery

Background:

• Current AI methods for drug-target interaction (DTI) prediction struggle with molecular diversity and noise in similarity measures.
• Limitations in generalizing to broader chemical spaces hinder efficient drug development.

Purpose of the Study:

• To develop an advanced AI framework, Adaptive Iterative Graph Optimization (AIGO)-DTI, for more accurate and reliable drug-target interaction prediction.
• To overcome limitations of existing methods by integrating atomic cluster information and enhancing molecular feature representation.

Main Methods:

• The AIGO-DTI framework integrates atomic cluster information and enhances molecular features using functional group prompts and graph encoders.
• DTI association network construction is optimized, transforming graph structure optimization into a node similarity learning problem.
• Multihead similarity metric functions are employed for iterative network structure updates to refine DTI information quality.

Main Results:

• AIGO-DTI demonstrated superior performance on public and label reversal datasets.
• Experimental validation, including molecular docking and case studies, confirmed the framework's effectiveness and reliability.
• The method successfully constructs comprehensive and reliable DTI association network information.

Conclusions:

• The proposed AIGO-DTI framework offers novel graphing and optimization strategies for DTI prediction.
• This contributes to accelerating efficient drug development and reducing drug target discovery costs.
• The study highlights the potential of advanced AI in overcoming challenges in predicting drug-target interactions.