MAGC-DTI: modality-shared space and adaptive gated interactive cross-attention for drug-target interaction prediction

Area of Science:

• Bioinformatics
• Computational Chemistry
• Drug Discovery

Background:

• Accurate drug-target interaction (DTI) prediction is vital for drug repurposing and development.
• Current deep learning methods face challenges in analyzing hierarchical protein patterns and bidirectional drug-protein information exchange.

Purpose of the Study:

• To develop an advanced end-to-end framework, MAGC-DTI, for accurate DTI prediction.
• To address limitations in existing models by integrating bidirectional information exchange and hierarchical pattern capture.

Main Methods:

• Proposed MAGC-DTI framework with three innovations: multi-scale attention aggregation (MSAA) for protein patterns, adaptive gated interactive cross attention (AGICA) for cross-modal interaction, and multi-path residual classifier (MPRC) for fusion.
• Evaluated on six benchmark datasets against seven state-of-the-art baselines.

Main Results:

• MAGC-DTI demonstrated favorable performance compared to existing DTI prediction methods.
• Achieved competitive results in challenging cold-start and cross-domain prediction scenarios.
• The model offers interpretable insights via attention visualization, confirming biological relevance.

Conclusions:

• MAGC-DTI effectively predicts drug-target interactions by integrating advanced deep learning techniques.
• The framework shows promise for accelerating drug discovery and repurposing.
• Interpretability of the model aids in understanding biological relationships.