Learning long- and short-term dependencies for improving drug-target binding affinity prediction using transformer...

Area of Science:

• Computational biology
• Drug discovery
• Bioinformatics

Background:

• Accurate drug-target affinity (DTA) prediction is vital for drug discovery.
• Existing deep learning methods struggle to capture global protein features.

Purpose of the Study:

• To develop a novel model, ETransDTA, for enhanced DTA prediction.
• To simultaneously extract global and local features of target proteins.

Main Methods:

• ETransDTA combines convolutional layers and transformers for feature extraction.
• A graph pooling mechanism is integrated into topology adaptive graph convolutional network (TAGCN) for chemical compound representation.
• The model utilizes queries, keys, and values from stacked convolutional neural networks (CNNs) for integrated context.

Main Results:

• ETransDTA outperformed baseline methods on Davis and KIBA datasets.
• Achieved a mean square error (MSE) of 0.125 on the KIBA dataset, a 0.6% improvement.
• Demonstrated enhanced integration of local and global protein context for improved DTA prediction accuracy.

Conclusions:

• ETransDTA offers a significant advancement in DTA prediction accuracy.
• The model's ability to capture both global and local features improves drug discovery pipelines.
• ETransDTA provides a more robust tool for identifying potential drug candidates.