GraphATT-DTA: Attention-Based Novel Representation of Interaction to Predict Drug-Target Binding Affinity

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Drug-target binding affinity (DTA) prediction is crucial for efficient drug discovery.
• Current deep learning methods often overlook specific drug substructure and protein subsequence interactions.
• Understanding these localized interactions is key to accurate affinity prediction.

Purpose of the Study:

• To develop a novel deep learning model, GraphATT-DTA, for improved DTA prediction.
• To incorporate an attention mechanism to identify and analyze critical interaction regions between drugs and proteins.
• To enhance the interpretability of DTA prediction models.

Main Methods:

• GraphATT-DTA model construction utilizing graph-based representations.
• Implementation of an attention mechanism to capture local-to-global interactions between drug substructures and protein subsequences.
• Training and evaluation on the Davis dataset and human kinase dataset, with external validation using BindingDB.

Main Results:

• GraphATT-DTA demonstrated superior performance in DTA prediction compared to existing state-of-the-art models.
• The attention mechanism provided enhanced interpretability by highlighting key interaction regions.
• Consistent results were observed across both internal and external validation datasets.

Conclusions:

• GraphATT-DTA offers a significant advancement in DTA prediction accuracy and interpretability.
• The model's ability to focus on specific interaction sites improves prediction reliability.
• This approach holds promise for accelerating the drug discovery pipeline.