DICCA-DTA: Diffusion and Contextualized Capsule Attention guided Factorized Cross-Pooling for Drug-Target Affinity...

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Drug-Target Affinity (DTA) prediction is vital for drug discovery.
• Current deep learning methods face challenges in molecular graph representation and interaction modeling.
• Accurate DTA prediction requires effective feature extraction and interaction prioritization.

Purpose of the Study:

• To introduce the DICCA-DTA framework for improved DTA prediction.
• To address limitations in molecular information propagation and interaction modeling.
• To enhance the accuracy and interpretability of drug-target interaction predictions.

Main Methods:

• Utilized a Diffused Isomorphic Network (DIN) for comprehensive drug feature extraction.
• Employed a Contextualized Capsule Attention Network (CCAN) for protein sequence characteristic modeling.
• Implemented an attention-guided Factorized Cross-Pooling (FCP) mechanism for refined interaction modeling and explainable attention maps.

Main Results:

• DICCA-DTA demonstrated superior performance across Davis, KIBA, Metz, and BindingDB datasets.
• The framework accurately models complex drug-protein binding site interactions.
• Explainable attention maps provided transparent insights into critical interactions.

Conclusions:

• The DICCA-DTA framework significantly advances DTA prediction accuracy and interpretability.
• It offers a robust approach for identifying key drug-protein affinities.
• The framework has the potential to accelerate drug discovery and repurposing efforts.