Predicting drug and target interaction with dilated reparameterize convolution.

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Predicting drug-target interactions (DTI) is crucial but challenging.
• Existing deep learning models struggle with large-scale sequence information and complex binding sites.

Purpose of the Study:

• To develop a novel deep learning framework for accurate DTI prediction.
• To address limitations in capturing large-scale sequence information and binding site interactions.

Main Methods:

• Proposed Rep-ConvDTI framework incorporating a large-kernel convolutional block with reparameterization.
• Introduced a gated attention mechanism for enhanced drug-target interaction characterization.
• Utilized three benchmark datasets for extensive experimental validation.

Main Results:

• Rep-ConvDTI achieved superior performance compared to state-of-the-art methods on benchmark datasets.
• Demonstrated effectiveness in capturing large-scale sequence information and binding site dynamics.
• Validated through model interpretability and drug screening experiments.

Conclusions:

• Rep-ConvDTI offers a powerful new approach for DTI prediction.
• The framework shows significant potential for accelerating drug discovery and screening.
• Rep-ConvDTI effectively models complex drug-target binding relationships.