Generative discovery of chemical reaction networks
Abstract
Chemical kinetics determines molecular stability, reaction selectivity and the evolution of reacting mixtures, but discovering the underlying reaction networks remains challenging. Here we introduce ReactNexus, a generative framework for kinetic-guided network exploration starting solely from reactant graphs. Generative models propose transition states and predict the full reaction trajectories connecting reactants and products, while calculations using a machine-learned interatomic potential verify connectivity and evaluate electronic barriers. On an in-distribution benchmark of 10,000 generated candidates, 80.8% connect to the input reactant after density functional theory (DFT) optimization and intrinsic reaction coordinate verification. The resulting networks uncover lower-barrier cooperative pathways and competition between catalyst release and catalyst-bound branches. In the examined carbocation cascades, coverage of known product precursors saturates after limited expansion and distinct entries converge onto shared intermediates, indicating a compact accessible reaction space. Network-derived candidate structures further assist mass-spectral interpretation of a complex nucleophilic aromatic substitution reaction. Together, these results establish a generative approach to reaction-network discovery and mechanistic analysis without predefined products or reaction templates.
Related
- Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery
- Large-deviations theory for growing chemical reaction networks
- Classification of strongly non-hyperbolic critical points of planar second-order chemical reaction systems
- Autocatalysis and Boundary Stability in Chemical Reaction Systems
- Scalable Enzyme Reaction Retrieval Enables Metagenomic Biocatalyst Discovery
BibTeX
@misc{han2026generative,
title = {{Generative discovery of chemical reaction networks}},
author = {Yunhong Han and Gaopeng Ren and Yan Liu and Ronghua Zhang and Huangjiang Huang and Ruiqi Zhang and Zhaoxi Yu and Chuncheng Liu and Zhao Li and Xiaohan Lin and Jiaxu Wang and Yuxuan Qin and Meijie Wang and Guoshun Mei and Doudou Wang and Taoyong Cui and Haojun Jia and Chenru Duan and Qiyuan Zhao},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.38236},
eprint = {2610.38236},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.38236},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - Generative discovery of chemical reaction networks
AU - Yunhong Han
AU - Gaopeng Ren
AU - Yan Liu
AU - Ronghua Zhang
AU - Huangjiang Huang
AU - Ruiqi Zhang
AU - Zhaoxi Yu
AU - Chuncheng Liu
AU - Zhao Li
AU - Xiaohan Lin
AU - Jiaxu Wang
AU - Yuxuan Qin
AU - Meijie Wang
AU - Guoshun Mei
AU - Doudou Wang
AU - Taoyong Cui
AU - Haojun Jia
AU - Chenru Duan
AU - Qiyuan Zhao
PY - 2026
DA - 2026/10/28
PB - arXiv
DO - 10.48550/arxiv.2610.38236
UR - https://arxiv.org/abs/2610.38236
AB - Chemical kinetics determines molecular stability, reaction selectivity and the evolution of reacting mixtures, but discovering the underlying reaction networks remains challenging. Here we introduce ReactNexus, a generative framework for kinetic-guided network exploration starting solely from reactant graphs. Generative models propose transition states and predict the full reaction trajectories connecting reactants and products, while calculations using a machine-learned interatomic potential verify connectivity and evaluate electronic barriers. On an in-distribution benchmark of 10,000 generated candidates, 80.8% connect to the input reactant after density functional theory (DFT) optimization and intrinsic reaction coordinate verification. The resulting networks uncover lower-barrier cooperative pathways and competition between catalyst release and catalyst-bound branches. In the examined carbocation cascades, coverage of known product precursors saturates after limited expansion and distinct entries converge onto shared intermediates, indicating a compact accessible reaction space. Network-derived candidate structures further assist mass-spectral interpretation of a complex nucleophilic aromatic substitution reaction. Together, these results establish a generative approach to reaction-network discovery and mechanistic analysis without predefined products or reaction templates.
ER -