hERGAT: predicting hERG blockers using graph attention mechanism through atom- and molecule-level interaction...

Area of Science:

• Computational chemistry and cheminformatics
• Pharmacology and toxicology
• Artificial intelligence in drug discovery

Background:

• The human ether-a-go-go-related gene (hERG) channel is vital for cardiac electrical activity; its blockers can induce cardiotoxicity.
• Accurate prediction of hERG channel blockers is essential for drug development safety.
• Existing in silico models often struggle with high performance and interpretability.

Purpose of the Study:

• To develop an interpretable and high-performing in silico model for predicting hERG channel blockers.
• To leverage graph neural networks and attention mechanisms for analyzing atomic and molecular interactions.
• To enhance early-stage drug safety assessment and reduce cardiotoxic risks.

Main Methods:

• Proposed hERGAT, a graph neural network model incorporating graph attention mechanisms (GAT) and gated recurrent units (GRU).
• Analyzed atomic-level interactions using GAT to integrate information from neighboring and distant atoms.
• Incorporated molecule-level attention mechanisms to identify critical substructures and integrated physicochemical properties.

Main Results:

• The hERGAT model achieved high predictive performance with an Area Under the Receiver Operating Characteristic curve of 0.907 and an Area Under the Precision-Recall curve of 0.904.
• Attention mechanisms successfully highlighted molecular substructures crucial for hERG activity prediction, confirmed by literature review.
• Clustering analysis and correlation heatmaps validated the model's consideration of distant atomic interactions.

Conclusions:

• hERGAT offers a reliable and interpretable framework for predicting hERG channel blockers.
• The model's ability to capture complex atomic and molecular interactions improves early cardiotoxicity assessment.
• hERGAT demonstrates significant potential for optimizing drug safety during the early stages of drug development.