Classical Structural Priors Improve Neural Molecular Representations in Molecule Language Models
Journal of Chemical Information and Modeling · 16.11.2026
DOI: 10.1021/acs.jcim.6c02225 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Chemical Information and Modeling, online erschienen am 16.11.2026. Verfasst von Donghyeok Lee und 3 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Gefördert von: Ministry of Health and Welfare, Ministry of Science and ICT, South Korea.
Abstract (Originalsprache)
Classical molecular similarity is central to chemical information systems, yet molecule-language models trained on paired structure–text data do not necessarily organize molecular embeddings according to graded structural relationships. Here, we introduce Structure-Aware Guided Alignment (SAGA), a training-only regularizer that uses Centered Kernel Alignment to align neural molecular representations with fingerprint-based Tanimoto similarity and requires no additional parameters or fingerprints at inference. In MolCA and Atomas, SAGA improves molecule–text retrieval, retrieves closer structural analogues, and raises average performance in scaffold-split property prediction, without clear degradation of molecular description or text-to-molecule generation. These results, together with control experiments that attribute the gains to the chemical content of the prior, show that established chemoinformatics priors can make molecule-language representations more chemically faithful while retaining their multimodal utility.
Förderung
- Ministry of Health and Welfare
- Ministry of Science and ICT, South Korea
Zitieren
APA
Lee, D., Park, S., Chang, M., & Yoon, S. (2026). Classical Structural Priors Improve Neural Molecular Representations in Molecule Language Models. Journal of Chemical Information and Modeling. https://doi.org/10.1021/acs.jcim.6c02225
Vancouver
Lee D, Park S, Chang M, Yoon S. Classical Structural Priors Improve Neural Molecular Representations in Molecule Language Models. Journal of Chemical Information and Modeling. 2026. doi:10.1021/acs.jcim.6c02225
BibTeX
@article{Lee2026,
author = {Lee, Donghyeok and Park, Sangkwon and Chang, Munyoung and Yoon, Sungroh},
title = {Classical Structural Priors Improve Neural Molecular Representations in Molecule Language Models},
journal = {Journal of Chemical Information and Modeling},
year = {2026},
doi = {10.1021/acs.jcim.6c02225}
}
EUVIMED