Predicting target-ligand interactions with graph convolutional networks for interpretable pharmaceutical discovery.

Area of Science:

• Computational Chemistry
• Bioinformatics
• Drug Discovery

Background:

• Drug discovery is complex, costly, and has low returns.
• Identifying effective therapeutic candidates is a major challenge.

Purpose of the Study:

• To propose PLA-Net, a deep learning model for predicting protein-ligand interactions.
• To improve the efficiency and accuracy of identifying potential drug candidates.

Main Methods:

• Developed a two-module deep graph convolutional network (PLA-Net).
• Incorporated adversarial data augmentations to enhance model interpretability and performance.
• Combined ligand and target chemical information for binding capability prediction.

Main Results:

• PLA-Net achieved 86.52% mean average precision for 102 target proteins.
• Demonstrated significant performance increase using joint ligand-target information and adversarial augmentations.
• Successfully predicted pharmacologically relevant molecules from large datasets.

Conclusions:

• PLA-Net effectively predicts protein-ligand interactions, improving drug discovery efficiency.
• Adversarial augmentations enhance model interpretability and predictive power.
• The model shows promise for identifying novel therapeutic candidates and drug repurposing opportunities.