CDI-DTI: A Strong Cross-Domain Interpretable Drug-Target Interaction Prediction Framework Based on Multi-Strategy...

Area of Science:

• Bioinformatics
• Computational Chemistry
• Drug Discovery

Background:

• Accurate drug-target interaction (DTI) prediction is crucial for efficient drug discovery.
• Existing DTI prediction methods struggle with cross-domain generalization, cold-start problems, and lack interpretability.

Purpose of the Study:

• To introduce CDI-DTI, a novel cross-domain interpretable framework for DTI prediction.
• To address limitations in current DTI prediction methodologies, including generalization and cold-start scenarios.

Main Methods:

• Integrated multimodal features (textual, structural, functional) using a multistrategy fusion approach.
• Employed a multisource cross-attention mechanism for early feature alignment and fusion.
• Utilized a bidirectional cross-attention layer for intramodal interaction and Gram Loss with a deep orthogonal fusion module for late fusion.

Main Results:

• CDI-DTI demonstrated superior performance compared to existing methods on benchmark datasets.
• Achieved significant improvements in cross-domain and cold-start DTI prediction tasks.
• Maintained high interpretability, facilitating practical applications.

Conclusions:

• CDI-DTI offers a robust and interpretable solution for drug-target interaction prediction.
• The framework effectively overcomes cross-domain and cold-start challenges in DTI prediction.
• Presents a valuable tool for advancing drug discovery and development.