Descriptor-Based Pre-Training Improves Reaction Property Prediction
4 affiliations
- 1Department of Chemical Engineering
- 2Massachusetts Institute of Technology
- 3Modeling and Informatics
- 4Merck & Co., Inc.
Abstract
Pre-training graph neural networks on large unlabeled datasets has substantially improved molecular property prediction, yet pre-training approaches for reaction property prediction remain underexplored. We introduce CheMeleon-Rxn, a pre-trained reaction graph neural network that adapts the CheMeleon framework to condensed graphs of reaction (CGR) and demonstrates that descriptor-regression is an effective pre-training strategy for low-data reaction property prediction. An O(10M)-parameter D-MPNN encoder was pre-trained on ∼2 million reactions by regressing onto dense reaction descriptors pooled from classical molecular descriptors, then fine-tuned on small reaction property prediction datasets. Across regression tasks spanning gasphase activation energies, reaction enthalpies, experimental yields, and rate coefficients, CheMeleon-Rxn is best or statistically tied-for-best on seven of eight tasks. It outperforms the same encoder trained from scratch, as well as classical-descriptor, fingerprint, and pre-trained-fingerprint baselines. We further tested the pre-training strategy across various graph neural network architectures and found that its benefit holds for other edge-aware backbones such as GINE, and that it is robust to the choice of descriptor set, pooling operation, and other pre-training hyperparameters. We visualized and analyzed the CheMeleon-Rxn embeddings, which reveal chemically coherent structure in the learned reaction representation space. The CheMeleon-Rxn code has been released publicly (https://github.com/MSDLLCpapers/chemeleon-rxn) to encourage adoption and extension by the community. The model weights will be released upon publication.
Related
- Transfer Protocol Matters More Than Source-Task Selection in Low-Data Molecular Property Prediction: An Audit of QM9 Pretraining on Computed Oxidation Potentials and Redox Reaction Energies
- Transformers for molecular property prediction: domain adaptation efficiently improves performance
- AqSol-Blend: A CPU-Trained Descriptor Ensemble for Aqueous Solubility Prediction
- Accept or abstain? A pre-specified test of reliability gates calibrated on held-out chemistry for molecular property and reaction-yield models
- Transfer learning on universal interatomic potential embeddings improves generalization in structure-property defect models
BibTeX
@misc{zalte2026descriptorbased,
title = {{Descriptor-Based Pre-Training Improves Reaction Property Prediction}},
author = {Akshat Shirish Zalte and Cheng Fang and Yiting Zheng and Essam Metwally and Alan Cheng and Alec Glisman and Hao-Wei Pang},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15008692/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008692/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Descriptor-Based Pre-Training Improves Reaction Property Prediction
AU - Akshat Shirish Zalte
AU - Cheng Fang
AU - Yiting Zheng
AU - Essam Metwally
AU - Alan Cheng
AU - Alec Glisman
AU - Hao-Wei Pang
PY - 2026
DA - 2026/09/11
PB - ChemRxiv
DO - 10.26434/chemrxiv.15008692/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008692/v1
AB - Pre-training graph neural networks on large unlabeled datasets has substantially improved molecular property prediction, yet pre-training approaches for reaction property prediction remain underexplored. We introduce CheMeleon-Rxn, a pre-trained reaction graph neural network that adapts the CheMeleon framework to condensed graphs of reaction (CGR) and demonstrates that descriptor-regression is an effective pre-training strategy for low-data reaction property prediction. An O(10M)-parameter D-MPNN encoder was pre-trained on ∼2 million reactions by regressing onto dense reaction descriptors pooled from classical molecular descriptors, then fine-tuned on small reaction property prediction datasets. Across regression tasks spanning gasphase activation energies, reaction enthalpies, experimental yields, and rate coefficients, CheMeleon-Rxn is best or statistically tied-for-best on seven of eight tasks. It outperforms the same encoder trained from scratch, as well as classical-descriptor, fingerprint, and pre-trained-fingerprint baselines. We further tested the pre-training strategy across various graph neural network architectures and found that its benefit holds for other edge-aware backbones such as GINE, and that it is robust to the choice of descriptor set, pooling operation, and other pre-training hyperparameters. We visualized and analyzed the CheMeleon-Rxn embeddings, which reveal chemically coherent structure in the learned reaction representation space. The CheMeleon-Rxn code has been released publicly (https://github.com/MSDLLCpapers/chemeleon-rxn) to encourage adoption and extension by the community. The model weights will be released upon publication.
ER -