MolAnchor method for explaining compound predictions based on substructures

In medicinal chemistry, the impact of machine learning remains limited if predictions are not understood, which often precludes experimental follow-up. Therefore, chemically intuitive approaches that aid in model understanding and interpretation at the molecular level of detail are sought after. Whi...

Full description

Saved in:
Bibliographic Details
Main Authors: Alec Lamens, Jürgen Bajorath
Format: Article
Language:English
Published: Elsevier 2024-12-01
Series:European Journal of Medicinal Chemistry Reports
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S277241742400102X
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850137947779629056
author Alec Lamens
Jürgen Bajorath
author_facet Alec Lamens
Jürgen Bajorath
author_sort Alec Lamens
collection DOAJ
description In medicinal chemistry, the impact of machine learning remains limited if predictions are not understood, which often precludes experimental follow-up. Therefore, chemically intuitive approaches that aid in model understanding and interpretation at the molecular level of detail are sought after. While feature attribution methods quantifying feature importance for model decisions are widely used in many areas, they must typically be combined with visualization techniques, if possible, to render the results accessible from a chemical viewpoint. On the other hand, there are approaches such as counterfactuals that yield closely related chemical structures with different prediction outcomes, providing direct access to structural features that critically influence model decisions. Herein, we introduce another approach designed to rationalize chemical predictions based on molecular structure. Therefore, we adapt principles underlying the anchor concept from explainable artificial intelligence (XAI) and alter them for molecular machine learning. The resulting method, termed MolAnchor, systematically identifies substructures in test compounds that determine property predictions, thus ensuring chemical interpretability. The MolAnchor methodology is made freely available to the medicinal chemistry community as a part of our study.
format Article
id doaj-art-e94d5ded8f4b4e7c950f25fb23152576
institution OA Journals
issn 2772-4174
language English
publishDate 2024-12-01
publisher Elsevier
record_format Article
series European Journal of Medicinal Chemistry Reports
spelling doaj-art-e94d5ded8f4b4e7c950f25fb231525762025-08-20T02:30:42ZengElsevierEuropean Journal of Medicinal Chemistry Reports2772-41742024-12-011210023010.1016/j.ejmcr.2024.100230MolAnchor method for explaining compound predictions based on substructuresAlec Lamens0Jürgen Bajorath1Department of Life Science Informatics and Data Science, B-IT, LIMES Program Unit Chemical Biology and Medicinal Chemistry, Rheinische Friedrich-Wilhelms-Universität, Friedrich-Hirzebruch-Allee 5/6, D-53115, Bonn, Germany; Lamarr Institute for Machine Learning and Artificial Intelligence, Rheinische Friedrich-Wilhelms-Universität Bonn, Friedrich-Hirzebruch-Allee 5/6, D-53115, Bonn, GermanyDepartment of Life Science Informatics and Data Science, B-IT, LIMES Program Unit Chemical Biology and Medicinal Chemistry, Rheinische Friedrich-Wilhelms-Universität, Friedrich-Hirzebruch-Allee 5/6, D-53115, Bonn, Germany; Lamarr Institute for Machine Learning and Artificial Intelligence, Rheinische Friedrich-Wilhelms-Universität Bonn, Friedrich-Hirzebruch-Allee 5/6, D-53115, Bonn, Germany; Corresponding author. Department of Life Science Informatics and Data Science, B-IT, LIMES Program Unit Chemical Biology and Medicinal Chemistry, Rheinische Friedrich-Wilhelms-Universität, Friedrich-Hirzebruch-Allee 5/6, D-53115 Bonn, Germany.In medicinal chemistry, the impact of machine learning remains limited if predictions are not understood, which often precludes experimental follow-up. Therefore, chemically intuitive approaches that aid in model understanding and interpretation at the molecular level of detail are sought after. While feature attribution methods quantifying feature importance for model decisions are widely used in many areas, they must typically be combined with visualization techniques, if possible, to render the results accessible from a chemical viewpoint. On the other hand, there are approaches such as counterfactuals that yield closely related chemical structures with different prediction outcomes, providing direct access to structural features that critically influence model decisions. Herein, we introduce another approach designed to rationalize chemical predictions based on molecular structure. Therefore, we adapt principles underlying the anchor concept from explainable artificial intelligence (XAI) and alter them for molecular machine learning. The resulting method, termed MolAnchor, systematically identifies substructures in test compounds that determine property predictions, thus ensuring chemical interpretability. The MolAnchor methodology is made freely available to the medicinal chemistry community as a part of our study.http://www.sciencedirect.com/science/article/pii/S277241742400102XMolecular machine learningModel interpretationExplainable artificial intelligenceAnchorsMolecular fragments
spellingShingle Alec Lamens
Jürgen Bajorath
MolAnchor method for explaining compound predictions based on substructures
European Journal of Medicinal Chemistry Reports
Molecular machine learning
Model interpretation
Explainable artificial intelligence
Anchors
Molecular fragments
title MolAnchor method for explaining compound predictions based on substructures
title_full MolAnchor method for explaining compound predictions based on substructures
title_fullStr MolAnchor method for explaining compound predictions based on substructures
title_full_unstemmed MolAnchor method for explaining compound predictions based on substructures
title_short MolAnchor method for explaining compound predictions based on substructures
title_sort molanchor method for explaining compound predictions based on substructures
topic Molecular machine learning
Model interpretation
Explainable artificial intelligence
Anchors
Molecular fragments
url http://www.sciencedirect.com/science/article/pii/S277241742400102X
work_keys_str_mv AT aleclamens molanchormethodforexplainingcompoundpredictionsbasedonsubstructures
AT jurgenbajorath molanchormethodforexplainingcompoundpredictionsbasedonsubstructures