Benchmarking explainable AI methods for toxicophore detection and toxicity prediction

Puntos clave

• This research aims to evaluate the performance of explainable AI (XAI) methods in detecting toxicophores and predicting toxicity accurately.
• Utilized high-accuracy models for toxicophore detection.
• Assessed multiple XAI methods for stability and chemical relevance.
• Identified functional groups across various prediction tasks.
• High-accuracy models demonstrated stable XAI explanations.
• Multiple XAI methods consistently identified toxicophores across prediction tasks.
• Functional groups were accurately detected in toxicity prediction.