Physical probes expose and alleviate chemical-environment collapse in molecular representations

2026-05-11

AI Summary

This work addresses the pervasive issue of “chemical environment collapse” in molecular representation learning, where subtle differences in local chemical environments are overlooked, rendering topologically equivalent atoms indistinguishable. To mitigate this, the authors construct a high-quality dataset combining experimental and computed ¹³C NMR spectra and propose CLAIM—a framework that integrates hierarchical chemical priors with cross-level contrastive learning. Without requiring explicit 3D molecular structures, CLAIM effectively aligns topological molecular inputs with atomic-level NMR observations. The method substantially improves accuracy in atomic-resolution spectral retrieval and ¹³C NMR chemical shift prediction, demonstrates robustness on flexible and tautomeric systems, and successfully transfers to downstream ADMET and fluorescence property prediction tasks. This study provides the first systematic characterization and effective alleviation of chemical environment collapse in molecular representation learning.

Abstract

Nuclear magnetic resonance (NMR) spectroscopy provides an experimental readout of local chemical environments, but its use in molecular representation learning has been constrained by heterogeneous data and incomplete atom-level assignments. Here we construct complementary high-fidelity experimental and computational 13C NMR resources, which reveal a recurrent form of representational collapse: atoms that are equivalent in molecular topology can remain experimentally distinct in their real chemical environments, whereas explicit 3D descriptions are further limited by static conformations in dynamic regimes. To alleviate this bottleneck, we develop CLAIM (Contrastive Learning for Atom-to-molecule Inference of Molecular NMR), a framework that aligns efficient topological molecular inputs with atom-resolved NMR observables. Through hierarchical chemical priors and cross-level contrastive learning, CLAIM restores lost chemical resolution and markedly improves atom-level molecule-spectrum retrieval. CLAIM remains robust in flexible and tautomeric systems for 13C NMR prediction, improves stereoisomer discrimination without explicit 3D modelling, and transfers to broader molecular property tasks including ADMET prediction and fluorescence estimation. These results establish physically grounded spectral alignment as an effective strategy for alleviating chemical-environment collapse and for guiding experimentally grounded molecular representation learning.

Problem

• chemical-environment collapse
• molecular representation learning
• NMR spectroscopy
• atom-level resolution
• dynamic molecular systems

Innovation

• chemical-environment collapse
• contrastive learning
• molecular representation learning
• NMR spectroscopy
• atom-resolved prediction