Structure‐Aware Machine Learning for Polymers: A Hierarchical Graph Network for Predicting Properties From Statistical Ensembles
Macromolecular Rapid Communications · 2026-06 · 47 (12)
DOI: 10.1002/marc.202500671 ↗Kurz gefasst
Fachartikel in der Zeitschrift Macromolecular Rapid Communications, online erschienen am 06.01.2026. Verfasst von Julian Kimmig und 7 weiteren Autor:innen. Beteiligte Einrichtung: Laboratory of Organic and Macromolecular Chemistry (IOMC), Friedrich Schiller University Jena, Jena, Germany. Hauptthemen laut MEDLINE: Polymers, Machine Learning. Die Arbeit wurde bisher 5-mal zitiert (Crossref) und nennt 57 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Thüringer Aufbaubank, Joachim Herz Stiftung.
Abstract (Originalsprache)
Machine learning applications in polymer science are often inefficient due to molecular representations that neglect the inherent hierarchical and statistical nature of macromolecules. This work introduces a structure‐aware graph convolutional network (GCN) framework that addresses this limitation by treating polymer samples as statistical ensembles. The approach utilizes a hierarchical graph representation where nodes correspond to monomer units and explicitly integrates molecular mass distribution (MMD) data to account for sample dispersity. A key innovation is an ensemble‐based training strategy using topologically realistic graphs generated on‐demand via an optimized kinetic Monte Carlo simulation. The model's efficacy was validated on a broad range of tasks. On synthetic data, it achieved more than 98% accuracy in classifying complex polymer architectures. When applied to a large experimental dataset, the model predicts glass transition temperatures ( T g ) with high accuracy (R 2 = 0.89 ± 0.01). Crucially, a fine‐tuning experiment demonstrated that the model could successfully learn the physically / chemically grounded relationship between T g and molar mass by integrating MMD information. This work establishes a robust and physically realistic paradigm for polymer informatics, enabling more accurate property predictions and paving the way for accelerated in silico material design.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Laboratory of Organic and Macromolecular Chemistry (IOMC), Friedrich Schiller University Jena, Jena, Germany.
- Jena Center for Soft Matter (JCSM), Friedrich Schiller University Jena, Jena, Germany.
- Linkdlab GmbH, Jena, Germany.
- Faculty of Engineering Science, University of Bayreuth, Bayreuth, Germany.
- Bavarian Center For Battery Technology (BayBatt), University of Bayreuth, Bayreuth, Germany.
- Helmholtz Institute for Polymers in Energy Applications Jena (HIPOLE Jena), Jena, Germany.
ORCID-Profile: Julian Kimmig, Yannik Köster, Timo Koswig, Ulrich S Schubert
Förderung
- Thüringer Aufbaubank
- Joachim Herz Stiftung
Zitieren
APA
Kimmig, J., Köster, Y., Koswig, T., Raviswamy, P., Ganti, S. V. S., Zechel, S., Kuenneth, C., & Schubert, U. S. (2026). Structure‐Aware Machine Learning for Polymers: A Hierarchical Graph Network for Predicting Properties From Statistical Ensembles. Macromolecular Rapid Communications, 47(12). https://doi.org/10.1002/marc.202500671
Vancouver
Kimmig J, Köster Y, Koswig T, Raviswamy P, Ganti SVS, Zechel S, et al. Structure‐Aware Machine Learning for Polymers: A Hierarchical Graph Network for Predicting Properties From Statistical Ensembles. Macromolecular Rapid Communications. 2026;47(12). doi:10.1002/marc.202500671
BibTeX
@article{Kimmig2026,
author = {Kimmig, Julian and Köster, Yannik and Koswig, Timo and Raviswamy, Punith and Ganti, Subhash V S and Zechel, Stefan and Kuenneth, Christopher and Schubert, Ulrich S},
title = {Structure‐Aware Machine Learning for Polymers: A Hierarchical Graph Network for Predicting Properties From Statistical Ensembles},
journal = {Macromolecular Rapid Communications},
year = {2026},
volume = {47},
number = {12},
doi = {10.1002/marc.202500671}
}
EUVIMED