Benchmarking explainable AI methods for toxicophore detection and toxicity prediction

2026/01/01 by Dina Khasanova, Igor V. Tetko · 1 voice
Computer Science · Materials Science · #Computational Drug Discovery Methods #Explainable Artificial Intelligence (XAI) #Machine Learning in Materials Science

paper · pdf · doi:10.1039/d5dd00576k

openalex publication_date 2026/01/01 · openalex created_date 2026/05/15 · openalex updated_date 2026/07/13

Abstract

High-accuracy models yield stable and chemically meaningful XAI explanations, with multiple methods consistently identifying functional groups and toxicophores across prediction tasks.

Citations

• Captum: A unified and generic model interpretability library for PyTorch
• Therapeutics Data Commons: Machine Learning Datasets and Tasks for Drug Discovery and Development
• Right for the Right Reasons: Training Differentiable Models by\n Constraining their Explanations
• Activation Functions in Deep Learning: A Comprehensive Survey and Benchmark
• Drug discovery with explainable artificial intelligence

Discussions

• How reliable is that explainable AI method for molecular property prediction? 🤔 Khasanova and Tetko from Helmholtz Munich investigate inconsistencies in #propertyprediction and the accuracy of expla [bsky, 2 points, 0 comments]

Related

• Combining Deep Learning and Explainable AI for Toxicity Prediction of Chemical Compounds
• Toxicity Prediction using Deep Learning
• Explaining Chemical Toxicity using Missing Features
• Accurate Clinical Toxicity Prediction using Multi-task Deep Neural Nets and Contrastive Molecular Explanations
• B-XAIC Dataset: Benchmarking Explainable AI for Graph Neural Networks Using Chemical Data