HyperAttentionDTI: improving drug-protein interaction prediction by sequence-based deep learning with attention...

Area of Science:

• Computational biology
• Drug discovery and development
• Bioinformatics

Background:

• Identifying drug-target interactions (DTIs) is essential for drug repurposing and discovery.
• In silico methods accelerate DTI identification, reducing time and cost.
• Existing sequence-based methods often overlook complex atomic-level interactions.

Purpose of the Study:

• To propose an end-to-end bio-inspired model, HyperAttentionDTI, for accurate DTI prediction.
• To enhance DTI prediction by modeling complex non-covalent interactions between atoms and amino acids.

Main Methods:

• Utilized deep convolutional neural networks (CNNs) to learn feature matrices for drugs and proteins.
• Applied an attention mechanism to the feature matrices to model atomic-level interactions.
• Assigned an attention vector to each atom or amino acid for refined interaction modeling.

Main Results:

• HyperAttentionDTI demonstrated significantly improved performance on benchmark datasets compared to state-of-the-art methods.
• The model achieved superior accuracy in predicting drug-target interactions.
• A case study on human Gamma-aminobutyric acid receptors validated the model's predictive power.

Conclusions:

• HyperAttentionDTI offers a powerful tool for accurate in silico DTI prediction.
• The model's ability to capture complex atomic interactions enhances its utility in drug discovery.
• This approach contributes to accelerating the development of new therapeutics.