Reaction-aware hypergraph learning of chemical representations with LARK
- 1Shandong University
Abstract
Molecular pretraining offers a route to learning transferable chemical representations from unlabelled data. However, existing approaches pretrained on isolated molecular structures struggle to generalize to reaction-specific tasks because their pretraining objectives provide limited information about chemical transformations. A key challenge is to integrate this reaction information into chemical representations. Here we introduce LARK, a unified framework for learning atom-molecule-reaction knowledge from chemical transformations. Using approximately 0.6 million reactions encompassing 0.88 million unique molecular components, LARK combines bidirectional bond-electron reconstruction with structural and geometric supervision through a role-aware hierarchical hypergraph Transformer that links local atom representations to reaction context. The learned representations transfer to both reaction prediction and single-molecule property prediction, with strong performance across 42 tasks covering reaction outcomes, physicochemical properties, toxicity, metabolism and pharmacokinetics. Reactioncentre analyses show that LARK identifies where bonds break and form, while targeted interventions reveal that the learned atom representations contribute to reconstructing these changes. These findings indicate that LARK learns chemically informative representations of substructures involved in molecular transformations. Together, the results establish chemical transformations as a source of structured supervision for representation learning across molecular and reaction scales.
Related
- Relation-Aware Graph-Hypergraph Networks for Antibody Sequence-Structure Co-Design
- CLIVeg: Contrastive Image-Vegetation Alignment for Botany-Aware Earth Observation Representations
- Classification of strongly non-hyperbolic critical points of planar second-order chemical reaction systems
- HR-FRGS: A Novel Biomarker Discovery Protocol using Dual Layer Hypergraph Learning for NGS RNA Sequence Data
- Autocatalysis and Boundary Stability in Chemical Reaction Systems
BibTeX
@misc{qiao2026reactionaware,
title = {{Reaction-aware hypergraph learning of chemical representations with LARK}},
author = {J. Qiao and Y. Liu and J. Jin and D. Wang and Q. Zou and R. Su and L. Wei},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.15.751932},
url = {https://www.biorxiv.org/content/10.64898/2026.09.15.751932v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Reaction-aware hypergraph learning of chemical representations with LARK
AU - J. Qiao
AU - Y. Liu
AU - J. Jin
AU - D. Wang
AU - Q. Zou
AU - R. Su
AU - L. Wei
PY - 2026
DA - 2026/09/15
PB - bioRxiv
DO - 10.64898/2026.09.15.751932
UR - https://www.biorxiv.org/content/10.64898/2026.09.15.751932v1
AB - Molecular pretraining offers a route to learning transferable chemical representations from unlabelled data. However, existing approaches pretrained on isolated molecular structures struggle to generalize to reaction-specific tasks because their pretraining objectives provide limited information about chemical transformations. A key challenge is to integrate this reaction information into chemical representations. Here we introduce LARK, a unified framework for learning atom-molecule-reaction knowledge from chemical transformations. Using approximately 0.6 million reactions encompassing 0.88 million unique molecular components, LARK combines bidirectional bond-electron reconstruction with structural and geometric supervision through a role-aware hierarchical hypergraph Transformer that links local atom representations to reaction context. The learned representations transfer to both reaction prediction and single-molecule property prediction, with strong performance across 42 tasks covering reaction outcomes, physicochemical properties, toxicity, metabolism and pharmacokinetics. Reactioncentre analyses show that LARK identifies where bonds break and form, while targeted interventions reveal that the learned atom representations contribute to reconstructing these changes. These findings indicate that LARK learns chemically informative representations of substructures involved in molecular transformations. Together, the results establish chemical transformations as a source of structured supervision for representation learning across molecular and reaction scales.
ER -