Interpreting graph neural networks with Myerson values for cheminformatics approaches
Open Access
- 12 August 2026
- journal article
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 5 (8) , 3378-3389
Abstract
A new, scalable interpretability method for GNNs using Myerson values to provide faithful, node-wise attribution heatmaps. By incorporating graph structural constraints, molecular predictions are explained as individual atomic contributions.Keywords
All Related Versions
Funding Information
- Deutsche Forschungsgemeinschaft (KO 4689/5-1)
- Deutsche Forschungsgemeinschaft (SPP 2363 – Molecular Machine Learning)
- University of Oxford
This publication has 22 references indexed in Scilit:
- From Black Boxes to Actionable Insights: A Perspective on Explainable Artificial Intelligence for Scientific DiscoveryJournal of Chemical Information and Modeling, 2023
- Evaluating explainability for graph neural networksScientific Data, 2023
- Benchmarking Molecular Feature Attribution Methods with Activity CliffsJournal of Chemical Information and Modeling, 2022
- Using Domain-Specific Fingerprints Generated Through Neural Networks to Enhance Ligand-Based Virtual ScreeningJournal of Chemical Information and Modeling, 2021
- Interpretation of Compound Activity Predictions from Complex Machine Learning Models Using Local Approximations and Shapley ValuesJournal of Medicinal Chemistry, 2019
- Unmasking Clever Hans predictors and assessing what machines really learnNature Communications, 2019
- Explaining prediction models and individual predictions with feature contributionsKnowledge and Information Systems, 2013
- ZINC: A Free Tool to Discover Chemistry for BiologyJournal of Chemical Information and Modeling, 2012
- ZINC − A Free Database of Commercially Available Compounds for Virtual ScreeningJournal of Chemical Information and Modeling, 2004
- Prediction of Physicochemical Parameters by Atomic ContributionsJournal of Chemical Information and Computer Sciences, 1999