DTANet+: Dual Interaction and Kernel-Diverse Network for Drug-Target Affinity Prediction.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Drug-target binding affinity (DTA) prediction is crucial for drug design but limited by costly assays.
• Current computational methods often lack biochemical insights, hindering performance.

Purpose of the Study:

• To introduce DTANet+, a novel deep learning framework for enhanced DTA prediction.
• To integrate biochemical properties and binding site information into DTA prediction models.

Main Methods:

• Developed DTANet+, a deep learning framework incorporating Kernel-diverse Feature Extraction Block (KFEB), Cross-Scale Interaction Module (CSIM), Drug-Target Interaction Module (DTIM), and Multi-modal Fusion Module (MFM).
• Utilized KFEB and CSIM to extract features from functional groups and peptide chains.
• Employed DTIM to analyze drug-target binding sites and MFM for information integration.

Main Results:

• DTANet+ demonstrated superior performance on KIBA and Davis datasets.
• Achieved higher Concordance Index (CI) and lower Mean Squared Error (MSE) compared to existing methods.
• Validated the framework's effectiveness in DTA prediction.

Conclusions:

• DTANet+ effectively integrates biochemical knowledge for accurate DTA prediction.
• The framework shows significant potential for accelerating drug repositioning and enabling personalized medicine.
• Publicly available source code facilitates further research and application.