Synthetic reaction data from transformation rules for graph-to-sequence pretraining in biodegradation pathway prediction

Irsyaad Rijwan; Liam Brydon; Aidan Jared; Jörg Simon Wicker

cs.LGOct 14, 2026

Synthetic reaction data from transformation rules for graph-to-sequence pretraining in biodegradation pathway prediction

Authors: Irsyaad Rijwan, Liam Brydon, Aidan Jared, Jörg Simon Wicker

Organizations: School of Computer Science, University of Auckland, Auckland, New Zealand. · enviPath Limited, Auckland, New Zealand.

Abstract

The scarcity of observed biodegradation reactions constrains the training of product and pathway predictors. We present enviGraph, a framework that generates and curates synthetic reaction data, uses these data to pretrain a graph-to-sequence predictor, and fine-tunes the predictor on observed environmental reactions for product and recursive pathway prediction. Applying inherited environmental transformation rules to structures from PubChem, NORMAN SusDat and EPA DSSTox, followed by curation, produces 776,851 canonical parent-product pairs for pretraining. We compare this dataset with three United States Patent and Trademark Office (USPTO) reaction datasets: USPTO-50K, USPTO-MIT and USPTO-Stereo. We evaluate enviGraph and the sequence-based enviFormer using ten folds split by pathway on the EAWAG Biocatalysis/Biodegradation Database (BBD), regulatory pesticide biodegradation data (Soil) and activated-sludge biodegradation data (Sludge). For both predictors, synthetic pretraining gives the highest BBD mean precision-recall area under the curve (PR-AUC) among the evaluated pretraining datasets for single-generation and multi-generation prediction. enviGraph achieves BBD single-generation PR-AUC of $0.202 \pm 0.050$ with synthetic pretraining, compared with $0.169 \pm 0.065$ for the retrospectively selected strongest USPTO condition, USPTO-Stereo (mean $\pm$ sample standard deviation across folds). Its exact BBD top-1 and top-5 accuracies with synthetic pretraining are 0.355 and 0.594. The strongest USPTO conditions give higher PR-AUC means than synthetic pretraining in most Soil and Sludge comparisons. enviGraph has higher Soil multi-generation PR-AUC means than enviFormer under each of the four main pretraining datasets, and with synthetic pretraining it also leads enviFormer on BBD single-generation PR-AUC and exact top-1 and top-5 accuracy.

Figures & tables

Figure 1: The integrated enviGraph framework.

Dataset	Compounds	Reactions	Rules	Pathways
BBD	1,399	1,480	290	219
Soil	2,609	2,447	80	317
Sludge	1,070	521	0	184

Table 1: Object counts in the local environmental packages before fold construction.

PSD construction population	Count
Cleaned source records	1,168,683
Unique seed structures	1,090,572
Represented SMIRKS entries, before / after curation	134 / 120
Top-five rule share, before / after curation	56.14% / 23.51%

Table 2: Construction and curation of the PSD synthetic corpus.

		BBD		Soil		Sludge
Predictor	Pretraining	Single PR-AUC $\uparrow$	Multi PR-AUC $\uparrow$	Single PR-AUC $\uparrow$	Multi PR-AUC $\uparrow$	Single PR-AUC $\uparrow$	Multi PR-AUC $\uparrow$
enviGraph	PSD-Curated	0.202 $\pm$ 0.050	0.110 $\pm$ 0.029	0.154 $\pm$ 0.037	0.119 $\pm$ 0.035	0.017 $\pm$ 0.017	0.010 $\pm$ 0.004
enviFormer	PSD-Curated	0.175 $\pm$ 0.054	0.112 $\pm$ 0.037	0.130 $\pm$ 0.028	0.088 $\pm$ 0.027	0.044 $\pm$ 0.045	0.018 $\pm$ 0.021
enviGraph	USPTO-50K	0.107 $\pm$ 0.027	0.052 $\pm$ 0.017	0.103 $\pm$ 0.024	0.098 $\pm$ 0.046	0.043 $\pm$ 0.033	0.019 $\pm$ 0.014
enviFormer	USPTO-50K	0.111 $\pm$ 0.034	0.051 $\pm$ 0.024	0.106 $\pm$ 0.019	0.071 $\pm$ 0.022	0.060 $\pm$ 0.060	0.014 $\pm$ 0.010
enviGraph	USPTO-MIT	0.148 $\pm$ 0.048	0.068 $\pm$ 0.025	0.149 $\pm$ 0.038	0.126 $\pm$ 0.057	0.112 $\pm$ 0.049	0.037 $\pm$ 0.015

Table 3: Fine-tuned single-generation and multi-generation PR-AUC.

Figure 2: PR-AUC differences between synthetic and USPTO pretraining.

		BBD		Soil		Sludge
Predictor	Pretraining	Top-1 $\uparrow$	Top-5 $\uparrow$	Top-1 $\uparrow$	Top-5 $\uparrow$	Top-1 $\uparrow$	Top-5 $\uparrow$
enviGraph	PSD-Curated	0.355 $\pm$ 0.052	0.594 $\pm$ 0.077	0.316 $\pm$ 0.046	0.563 $\pm$ 0.070	0.063 $\pm$ 0.035	0.263 $\pm$ 0.083
enviFormer	PSD-Curated	0.295 $\pm$ 0.057	0.530 $\pm$ 0.056	0.268 $\pm$ 0.050	0.525 $\pm$ 0.063	0.108 $\pm$ 0.087	0.344 $\pm$ 0.109
enviGraph	USPTO-50K	0.246 $\pm$ 0.032	0.461 $\pm$ 0.053	0.248 $\pm$ 0.052	0.509 $\pm$ 0.045	0.125 $\pm$ 0.055	0.379 $\pm$ 0.139
enviFormer	USPTO-50K	0.229 $\pm$ 0.047	0.396 $\pm$ 0.069	0.243 $\pm$ 0.042	0.475 $\pm$ 0.058	0.161 $\pm$ 0.099	0.351 $\pm$ 0.150
enviGraph	USPTO-MIT	0.282 $\pm$ 0.063	0.482 $\pm$ 0.059	0.308 $\pm$ 0.052	0.580 $\pm$ 0.070	0.252 $\pm$ 0.052	0.492 $\pm$ 0.114

Table 4: Fine-tuned exact top-1 and top-5 accuracy.

Chemical Reaction Prediction

See all 17 papers

Apr 15, 2026

Computational framework for multistep metabolic pathway design

In silico tools are important for generating novel hypotheses and exploring alternatives in de novo metabolic pathway design. However, while many computational frameworks have been proposed for retrobiosynthesis, few successful examples of algorithm-guided xenobiotic biochemical retrosynthesis have been reported in the literature. Deep learning has improved the quality of synthesis and retrosynthesis in organic chemistry applications. Inspired by this progress, we explored combining deep learning of biochemical transformations with the traditional retrobiosynthetic workflow to improve in silico synthetic metabolic pathway designs. To develop our computational biosynthetic pathway design framework, we assembled metabolic reaction and enzymatic template data from public databases. A data augmentation procedure, adapted from literature, was carried out to enrich the assembled reaction dataset with artificial metabolic reactions generated by enzymatic reaction templates. Two neural network-based pathway ranking models were trained as binary classifiers to distinguish assembled reactions from artificial counterparts; each model output a scalar quantifying the plausibility of a 1-step or 2-step pathway. Combining these two models with enzymatic templates, we built a multistep retrobiosynthesis pipeline and validated it by reproducing some natural and non-natural pathways computationally.

Peter Zhiping Zhang, Jeffrey D. Varner

Robert Frederick Smith School of Chemical and Biomolecular Engineering Cornell University Ithaca, NY 14850

Retrosynthesis

Ask PDF

Oct 8, 2026

Elucidating the Space of Enzymatic Reaction: A Unified Benchmark and Pretrained Model

Existing reaction models primarily learn molecular transformations, whereas enzy- matic reactions depend jointly on molecular structure and catalytic function. We formulate this problem as learning an enzymatic reaction space linking reactants, products, and Enzyme Commission (EC) annotations. To characterize this space, we introduce VenusRX-Bench, a unified benchmark for forward reaction prediction, single-step retrosynthesis, and EC-number prediction. VenusRX-Bench integrates reactions from multiple biochemical databases with standardized curation, leakage- controlled splits, and consistent evaluation. Benchmarking representative chemical and enzymatic models reveals a clear chemical-to-enzymatic domain gap, driven by limited domain data, catalytic-context dependency, and the difficulty of modeling large biomolecular structures. To bridge this gap, we develop VenusRX, a unified T5-style sequence-to-sequence model for enzymatic reactions. VenusRX jointly learns forward prediction, ret- rosynthesis, and reaction reconstruction, with two-stage training on millions of template-expanded reactions followed by real biochemical reactions. In addition, optional EC conditioning incorporates catalytic context, while Molecule Library- Constrained Decoding improves the generation of complex biomolecules. Across benchmark tasks and challenging generalization splits, VenusRX achieves the best or competitive performance on most evaluated settings over representative chem- ical and enzymatic baselines. Moreover, EC information consistently improves reaction prediction, while learned reaction representations support accurate EC prediction, revealing a bidirectional relationship between reaction structure and catalytic function. Together, VenusRX-Bench and VenusRX provide a unified framework for elucidating and modeling enzymatic reaction space

Yutong Hu, Tianming Huang, Yanbo Zhao +6

Shanghai Jiao Tong University · Shanghai Artificial Intelligence Laboratory · The University of Sydney

Sequence-to-Sequence Models Chemical Reaction Prediction Sequence Modeling

Ask PDF

Sep 29, 2026

RetroGEF: Dynamic Graph Edit Flow for Single-Step Retrosynthesis

Retrosynthesis enables the discovery of viable synthetic routes to target molecules. It plays a central role in modern drug discovery and materials design. Retrosynthesis involves molecular graph transformations that can change both connectivity and graph size. These transformations may introduce reactant components absent from the target while revising the product-derived structure. To model these transformations, we propose RetroGEF, a flow-based generative model for single-step retrosynthesis. Starting from the target molecule, it constructs possible reactants by adding atoms and changing bonds in the molecular graph. RetroGEF models molecular transformations and changes in graph size within the same generative process, rather than relying on a fixed-size graph canvas. It learns this process directly from product--reactant pairs without requiring a prescribed edit order. Experiments on representative retrosynthesis benchmarks demonstrate that RetroGEF achieves state-of-the-art performance.

Xiaozhuang Song, Xuemin Chen, Xinjian Zhao +2

The Chinese University of Hong Kong, Shenzhen · Shanghai AI Laboratory

One-Step Generative Modeling Molecular Generation Retrosynthesis

Ask PDF

Abstract

Chemical Reaction Prediction

Explore similar work

Computational framework for multistep metabolic pathway design

Elucidating the Space of Enzymatic Reaction: A Unified Benchmark and Pretrained Model

RetroGEF: Dynamic Graph Edit Flow for Single-Step Retrosynthesis