arXiv:2611.21276 · v1
Classical structural priors improve neural molecular representations in molecule language models
摘要
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, without adding parameters or requiring fingerprints at inference. In MolCA and Atomas, SAGA improves molecule-text retrieval, shifts non-exact retrieval toward closer structural analogues, and increases average scaffold-split property-prediction performance, while molecular description and text-to-molecule generation are improved or preserved in single-run ablations. These results show that established chemoinformatics priors can make molecule-language representations more chemically faithful while retaining their multimodal utility.
作者备注:40 pages. This is the originally submitted version. The final version, revised after peer review, is published in J. Chem. Inf. Model. (2026): https://doi.org/10.1021/acs.jcim.6c02225