AttenhERG: a reliable and interpretable graph neural network framework for predicting hERG channel blockers

Area of Science:

• Computational chemistry
• Pharmacology
• Drug discovery

Background:

• Drug-induced arrhythmias, particularly hERG channel dysfunction, are a major safety concern in pharmaceutical development.
• Current methods for predicting hERG toxicity are often expensive and time-consuming.
• Developing accurate computational models is essential for early-stage safety assessment.

Purpose of the Study:

• To introduce AttenhERG, a novel graph neural network framework for reliable and interpretable prediction of hERG channel blockers.
• To evaluate AttenhERG's performance against existing methods and assess its reliability through uncertainty estimation.
• To demonstrate the practical utility of AttenhERG in optimizing drug candidates for reduced hERG toxicity.

Main Methods:

• Development of AttenhERG, a graph neural network framework utilizing advanced machine learning techniques.
• Training and validation of the model on diverse datasets to predict hERG channel activity.
• Incorporation of uncertainty evaluation to quantify the model's predictive confidence.
• Application of AttenhERG in case studies for compound optimization.

Main Results:

• AttenhERG achieved an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.835, outperforming existing computational methods.
• The framework demonstrated reliable predictions across various datasets, confirmed by uncertainty analysis.
• Case studies showed AttenhERG's effectiveness in guiding the optimization of compounds, such as APH1A and NMT1 inhibitors, to mitigate hERG toxicity.

Conclusions:

• AttenhERG offers a significant advancement in predicting hERG channel blockers, enhancing accuracy and interpretability.
• The model's integrated uncertainty estimation provides a reliable measure of prediction confidence, crucial for drug safety.
• AttenhERG shows strong potential for rational drug design and efficient safety profiling in early drug development stages.