Predicting drug-target affinity through triple pre-activated random residual planet convolution coupled attention...

Area of Science:

• Computational chemistry and bioinformatics
• Artificial intelligence in drug discovery
• Molecular modeling and simulation

Background:

• Accurate drug-target affinity (DTA) prediction is crucial for efficient drug discovery.
• Traditional methods face challenges in scalability, accuracy, and interpretability.
• Improving DTA prediction enhances the identification of potential drug candidates.

Purpose of the Study:

• To develop a sophisticated approach for enhanced DTA prediction.
• To combine contact map representations with a novel deep learning network.
• To address limitations of existing methods in modeling drug-target interactions.

Main Methods:

• Utilized DTA, KIBA, and Davis datasets for input data.
• Employed Focal Vision Transformer with Gabor Filter for feature enhancement.
• Implemented Dual-Aggregation Transformer (DAT) for feature extraction.
• Developed the Triple Pre-Activated Random Residual Planet Convolution Attention Network (Tri-Pre-A2RP-2CAN) integrated with RCNN.
• Optimized the model using PACRTAMN architecture and Planet optimization for hyperparameter tuning.

Main Results:

• Achieved an outstanding prediction accuracy of 99.9%.
• Demonstrated superior performance compared to existing methods in modeling drug-target interactions.
• Successfully enhanced DTA prediction accuracy and molecular interaction analysis.

Conclusions:

• The proposed Tri-Pre-A2RP-2CAN approach offers a scalable and interpretable solution for drug discovery.
• This innovative method significantly optimizes drug discovery processes.
• The findings advance pharmaceutical research by improving the prediction of drug-target interactions.