EnzRetro: Enzymatic Retrosynthetic Planning With Site-Specific Reaction Edits Based on Sequence Generative Architecture

Abstract

Enzymatic biosynthesis has become increasingly crucial in green chemistry and biosynthesis. However, current computational tools struggle to effectively integrate enzyme identification with pathway synthesis due to the specificity of enzymes and their complex interactions with substrates. Here, we propose EnzRetro, a novel framework for enzymatic retrosynthesis that provides an end-to-end solution bridging retrosynthesis planning with enzymatic engineering. The core innovative concept of EnzRetro is site-specific reaction edits (SSREdits), a dynamic approach to representing structural transformations at specific enzyme active sites and forging a direct link between enzyme identification and reaction patterns. To enable the model to learn meaningful representations of SSREdits, we developed three pretraining tasks and then fine-tuned two specialized models: (1) the SSREdits generation model for pathway synthesis, which translates the target product into a sequence of reaction edits, and (2) the EC generation model for enzyme identification, which focuses on precise transformation sites within SSREdits, enabling generalization across diverse reactions. Extensive experiments demonstrate the superior performance of EnzRetro, with a promising 56.1% and 97.7% Top-1 accuracy on USPTO-50k dataset for retrosynthesis and ECREACT dataset for enzyme identification, respectively. Finally, EnzRetro combines the two-stage processes of retrosynthesis and enzyme identification into one-pot learning, enhancing both computational efficiency and interpretability, and bridging the gap between pathway synthesis and enzyme identification. The accuracy of EnzRetro has been validated by three enzymatic pathways. We developed a web platform for multi-step retrosynthesis planning that reconstructs multiple enzymatic pathways for putrescine biosynthesis with substantial diversity and significantly outperforms state-of-the-art baselines.

Keywords

biosynthesis / enzymatic retrosynthesis / enzyme identification / large language model

References

[1] K. M. Shebek, J. Strutz, L. J. Broadbelt, and K. E. J. Tyo, “Pickaxe: A Python Library for the Prediction of Novel Metabolic Reactions,” BMC Bioinformatics 24, no. 1 (2023): 106, https://doi.org/10.1186/s12859-023-05149-8.

[2] K. Sankaranarayanan and K. F. Jensen, “Computer-Assisted Multistep Chemoenzymatic Retrosynthesis Using a Chemical Synthesis Planner,” Chemical Science 14, no. 23 (2023): 6467–6475, https://doi.org/10.1039/D3SC01355C.

[3] Y. Chen, J. Gustafsson, A. Tafur Rangel, et al., “Reconstruction, Simulation and Analysis of Enzyme-Constrained Metabolic Models Using GECKO Toolbox 3.0,” Nature Protocols 19, no. 3 (2024): 629–667, https://doi.org/10.1038/s41596-023-00931-7.

[4] N. Haque, B. J. J. Abdullah, and N. H. A. Kasim, “Secretome: Pharmaceuticals for Cell-Free Regenerative Therapy,” Stem Cell Drugs-A New Generation of Biopharmaceuticals (Springer, 2018), 17–35.

[5] P. Berben, A. Bauer-Brandl, M. Brandl, et al., “Drug Permeability Profiling Using Cell-Free Permeation Tools: Overview and Applications,” European Journal of Pharmaceutical Sciences 119 (2018): 219–233.

[6] H. M. Abdallah, A. M. El-Halawany, A. Sirwi, et al., “Repurposing of Some Natural Product Isolates as Sars-Cov-2 Main Protease Inhibitors via In Vitro Cell Free and Cell-Based Antiviral Assessments and Molecular Modeling Approaches,” Pharmaceuticals 14, no. 3 (2021): 213.

[7] Y. Chen, A. Emileh, N. Brideau, and J. Britton, “Cell-Free Reactions in Continuous Manufacturing Systems,” Current Opinion in Green and Sustainable Chemistry 25 (2020): 100380.

[8] C. Yang, Y. Liu, W.-Q. Liu, C. Wu, and J. Li, “Designing Modular Cell-Free Systems for Tunable Biotransformation of L-Phenylalanine to Aromatic Compounds,” Frontiers in Bioengineering and Biotechnology 9 (2021): 730663.

[9] A. R. Alcántara, P. Domínguez de María, J. A. Littlechild, M. Schürmann, R. A. Sheldon, and R. Wohlgemuth, “Biocatalysis as Key to Sustainable Industrial Chemistry,” ChemSusChem 15, no. 9 (2022): e202102709.

[10] B. J. Rasor, B. Vögeli, G. M. Landwehr, J. W. Bogart, A. S. Karim, and M. C. Jewett, “Toward Sustainable, Cell-Free Biomanufacturing,” Current Opinion in Biotechnology 69 (2021): 136–144.

[11] J. Jack, H. Fu, A. Leininger, T. K. Hyster, and Z. J. Ren, “Cell-Free Co₂ Valorization to C6 Pharmaceutical Precursors via a Novel Electro-Enzymatic Process,” ACS Sustainable Chemistry & Engineering 10, no. 13 (2022): 4114–4121.

[12] R. A. Sheldon and D. Brady, “Green Chemistry, Biocatalysis, and the Chemical Industry of the Future,” ChemSusChem 15, no. 9 (2022): e202102628.

[13] Y. Shee, H. Li, P. Zhang, et al., “Site-Specific Template Generative Approach for Retrosynthetic Planning,” Nature Communications 15, no. 1 (2024): 7818, https://doi.org/10.1038/s41467-024-52048-4.

[14] Y. Han, X. Xu, C.-Y. Hsieh, et al., “Retrosynthesis Prediction With an Iterative String Editing Model,” Nature Communications 15, no. 1 (2024): 6404, https://doi.org/10.1038/s41467-024-50617-1.

[15] A. V. Sadybekov and V. Katritch, “Computational Approaches Streamlining Drug Discovery,” Nature 616, no. 7958 (2023): 673–685, https://doi.org/10.1038/s41586-023-05905-z.

[16] Z. Zhong, J. Song, Z. Feng, et al., “Recent Advances in Deep Learning for Retrosynthesis,” WIREs Computational Molecular Science 14, no. 1 (2024): e1694, https://doi.org/10.1002/wcms.1694.

[17] C. W. Coley, L. Rogers, W. H. Green, and K. F. Jensen, “Computer-Assisted Retrosynthesis Based on Molecular Similarity,” ACS Central Science 3, no. 12 (2017): 1237–1245, https://doi.org/10.1021/acscentsci.7b00355.

[18] M. H. S. Segler and M. P. Waller, “Neural-Symbolic Machine Learning for Retrosynthesis and Reaction Prediction,” Chemistry – A European Journal 23, no. 25 (2017): 5966–5971, https://doi.org/10.1002/chem.201605499.

[19] H. Dai, C. Li, C. Coley, B. Dai, and L. Song, “Retrosynthesis Prediction With Conditional Graph Logic Network,” in Advances in Neural Information Processing Systems, ed. H. Wallach, H. Larochelle, A. Beygelzimer, F. dAlché-Buc, E. Fox, and R. Garnett, vol. 32 (Curran Associates, Inc., 2019).

[20] S. Chen and Y. Jung, “Deep Retrosynthetic Reaction Prediction Using Local Reactivity and Global Attention,” JACS Au 1, no. 10 (2021): 1612–1620, https://doi.org/10.1021/jacsau.1c00246.

[21] M. H. S. Segler and M. P. Waller, “Neural-Symbolic Machine Learning for Retrosynthesis and Reaction Prediction,” Chemistry – A European Journal 23, no. 25 (2017): 5966–5971, https://doi.org/10.1002/chem.201605499.

[22] H. Dai, C. Li, C. Coley, B. Dai, and L. Song, “Retrosynthesis prediction with conditional graph logic network,” in Advances in Neural Information Processing Systems, ed. H. Wallach, H. Larochelle, A. Beygelzimer, F. d Alché-Buc, E. Fox, and R. Garnett, vol. 32 (Curran Associates, Inc., 2019).

[23] S. Chen and Y. Jung, “Deep Retrosynthetic Reaction Prediction using Local Reactivity and Global Attention,” JACS Au 1, no. 10 (2021): 1612–1620, https://doi.org/10.1021/jacsau.1c00246.

[24] B. Liu, B. Ramsundar, P. Kawthekar, et al., “Retrosynthetic Reaction Prediction Using Neural Sequence-to-Sequence Models,” ACS Central Science 3, no. 10 (2017): 1103–1113, https://doi.org/10.1021/acscentsci.7b00303.

[25] V. R. Somnath, C. Bunne, C. Coley, A. Krause, and R. Barzilay, “Learning Graph Models for Retrosynthesis Prediction,” Advances in Neural Information Processing Systems 34 (2021): 9405–9415.

[26] S. Zheng, J. Rao, Z. Zhang, J. Xu, and Y. Yang, “Predicting Retrosynthetic Reactions Using Self-Corrected Transformer Neural Networks,” Journal of Chemical Information and Modeling 60, no. 1 (2020): 47–55, https://doi.org/10.1021/acs.jcim.9b00949.

[27] S. Zheng, J. Rao, Z. Zhang, J. Xu, and Y. Yang, “Predicting Retrosynthetic Reactions Using Self-Corrected Transformer Neural Networks,” Journal of Chemical Information and Modeling 60, no. 1 (2019): 47–55.

[28] K. Lin, Y. Xu, J. Pei, and L. Lai, “Automatic Retrosynthetic Route Planning Using Template-Free Models,” Chemical Science 11, no. 12 (2020): 3355–3364, https://doi.org/10.1039/C9SC03666K.

[29] I. V. Tetko, P. Karpov, R. Van Deursen, and G. Godin, “State-of-the-Art Augmented NLP Transformer Models for Direct and Single-Step Retrosynthesis,” Nature Communications 11, no. 1 (2020): 5575, https://doi.org/10.1038/s41467-020-19266-y.

[30] E. Kim, D. Lee, Y. Kwon, M. S. Park, and Y.-S. Choi, “Valid, Plausible, and Diverse Retrosynthesis Using Tied Two-Way Transformers With Latent Variables,” Journal of Chemical Information and Modeling 61, no. 1 (2021): 123–133, https://doi.org/10.1021/acs.jcim.0c01074.

[31] S.-W. Seo, Y. Y. Song, J. Y. Yang, et al., “GTA: Graph Truncated Attention for Retrosynthesis,” Proceedings of the AAAI Conference on Artificial Intelligence 35, no. 1 (2021): 531–539, https://doi.org/10.1609/aaai.v35i1.16131.

[32] R. Sun, H. Dai, L. Li, S. Kearnes, and B. Dai, “Towards Understanding Retrosynthesis by Energy-Based Models,” in Advances in Neural Information Processing Systems, vol. 34 (Curran Associates, Inc., 2021), 10186–10194.

[33] Z. Tu and C. W. Coley, “Permutation Invariant Graph-to-Sequence Model for Template-Free Retrosynthesis and Reaction Prediction,” Journal of Chemical Information and Modeling 62, no. 15 (2022): 3503–3513, https://doi.org/10.1021/acs.jcim.2c00321.

[34] N. Yang, H. Wu, K. Zeng, Y. Li, and J. Yan, “Molecule Generation for Drug Design: A Graph Learning Perspective,” Fundamental Research 6, no. 1 (2024): 40–52.

[35] Y. Han, X. Xu, C.-Y. Hsieh, et al., “Retrosynthesis Prediction With an Iterative String Editing Model,” Nature Communications 15, no. 1 (2024): 6404, https://doi.org/10.1038/s41467-024-50617-1.

[36] C. Shi, M. Xu, H. Guo, M. Zhang, and J. Tang, “A Graph to Graphs Framework for Retrosynthesis Prediction,” in Proceedings of the 37th International Conference on Machine Learning, ed. H. D. III and A. Singh, vol. 119 (PMLR, 2008), 8818–8827.

[37] C. Yan, Q. Ding, P. Zhao, et al., “Retroxpert: Decompose Retrosynthesis Prediction Like a Chemist,” Advances in Neural Information Processing Systems 33 (2020): 11248–11258.

[38] V. R. Somnath, C. Bunne, C. W. Coley, A. Krause, and R. Barzilay, “Learning graph models for template-free retrosynthesis,” preprint, arXiv, June 12, 2020, https://doi.org/10.48550/arXiv.2006.07038.

[39] X. Wang, Y. Li, J. Qiu, et al., “RetroPrime: A Diverse, Plausible and Transformer-Based Method for Single-Step Retrosynthesis Predictions,” Chemical Engineering Journal 420 (2021): 129845, https://doi.org/10.1016/j.cej.2021.129845.

[40] M. Sacha, M. Błaż, P. Byrski, et al., “Molecule Edit Graph Attention Network: Modeling Chemical Reactions as Sequences of Graph Edits,” Journal of Chemical Information and Modeling 61, no. 7 (2021): 3273–3284, https://doi.org/10.1021/acs.jcim.1c00537.

[41] W. Zhong, Z. Yang, and C. Y.-C. Chen, “Retrosynthesis Prediction Using an End-to-End Graph Generative Architecture for Molecular Graph Editing,” Nature Communications 14, no. 1 (2023): 3009, https://doi.org/10.1038/s41467-023-38851-5.

[42] Z. Chen, O. R. Ayinde, J. R. Fuchs, H. Sun, and X. Ning, “G2Retro as a Two-Step Graph Generative Models for Retrosynthesis Prediction,” Communications Chemistry 6, no. 1 (2023): 1–19, https://doi.org/10.1038/s42004-023-00897-3.

[43] V. R. Somnath, C. Bunne, C. Coley, A. Krause, and R. Barzilay, “Learning Graph Models for Retrosynthesis Prediction,” Advances in Neural Information Processing Systems 34 (2021): 9405–9415.

[44] Y. Wang, C. Pang, Y. Wang, et al., “Retrosynthesis Prediction With an Interpretable Deep-Learning Framework Based on Molecular Assembly Tasks,” Nature Communications 14, no. 1 (2023): 6155, https://doi.org/10.1038/s41467-023-41698-5.

[45] M. Sacha, M. Błaż, P. Byrski, et al., “Molecule Edit Graph Attention Network: Modeling Chemical Reactions as Sequences of Graph Edits,” Journal of Chemical Information and Modeling 61, no. 7 (2021): 3273–3284, https://doi.org/10.1021/acs.jcim.1c00537.

[46] P. Carbonell, J. Wong, N. Swainston, et al., “Selenzyme: Enzyme Selection Tool for Pathway Design,” Bioinformatics (Oxford, England) 34, no. 12 (2018): 2153–2154, https://doi.org/10.1093/bioinformatics/bty065.

[47] H. Lu, D. J. Diaz, N. J. Czarnecki, et al., “Machine Learning-Aided Engineering of Hydrolases for PET Depolymerization,” Nature 604, no. 7907 (2022): 662–667, https://doi.org/10.1038/s41586-022-04599-z.

[48] F. Li, L. Yuan, H. Lu, et al., “Deep Learning-Based k_cat prediction enables improved enzyme-constrained model reconstruction,” Nature Catalysis 5, no. 8 (2022): 662–672, https://doi.org/10.1038/s41929-022-00798-z.

[49] G. B. Kim, J. Y. Kim, J. A. Lee, C. J. Norsigian, B. O. Palsson, and S. Y. Lee, “Functional Annotation of Enzyme-Encoding Genes Using Deep Learning With Transformer Layers,” Nature Communications 14, no. 1 (2023): 7370, https://doi.org/10.1038/s41467-023-43216-z.

[50] S. Zheng, T. Zeng, C. Li, et al., “Deep Learning Driven Biosynthetic Pathways Navigation for Natural Products With BioNavi-NP,” Nature Communications 13, no. 1 (2022): 3342, https://doi.org/10.1038/s41467-022-30970-9.

[51] W. Finnigan, L. J. Hepworth, S. L. Flitsch, and N. J. Turner, “RetroBioCat as a Computer-Aided Synthesis Planning Tool for Biocatalytic Reactions and Cascades,” Nature Catalysis 4, no. 2 (2021): 98–104, https://doi.org/10.1038/s41929-020-00556-z.

[52] D. Kreutter, P. Schwaller, and J.-L. Reymond, “Predicting Enzymatic Reactions With a Molecular Transformer,” Chemical Science 12, no. 25 (2021): 8648–8659, https://doi.org/10.1039/D1SC02362D.

[53] D. Probst, M. Manica, Y. G. Nana Teukam, A. Castrogiovanni, F. Paratore, and T. Laino, “Biocatalysed Synthesis Planning Using Data-Driven Learning,” Nature Communications 13, no. 1 (2022): 964, https://doi.org/10.1038/s41467-022-28536-w.

[54] W. Qian, X. Wang, Y. Kang, P. Pan, T. Hou, and C.-Y. Hsieh, “A General Model for Predicting Enzyme Functions Based on Enzymatic Reactions,” Journal of Cheminformatics 16, no. 1 (2024): 38.

[55] P. Schwaller, T. Gaudin, D. Lányi, C. Bekas, and T. Laino, ““Found in Translation”: Predicting Outcomes of Complex Organic Chemistry Reactions Using Neural Sequence-to-Sequence Models,” Chemical Science 9, no. 28 (2018): 6091–6098, https://doi.org/10.1039/C8SC02339E.

[56] C. W. Coley, W. Jin, L. Rogers, et al., “A Graph-Convolutional Neural Network Model for the Prediction of Chemical Reactivity,” Chemical Science 10, no. 2 (2019): 370–377, https://doi.org/10.1039/C8SC04228D.

[57] P. Schwaller, T. Laino, T. Gaudin, et al., “Molecular Transformer: A Model for Uncertainty-Calibrated Chemical Reaction Prediction,” 5, no. 9, 1572–1583, https://doi.org/10.1021/acscentsci.9b00576.

[58] J. Lu and Y. Zhang, “Unified Deep Learning Model for Multitask Reaction Predictions With Explanation,” Journal of Chemical Information and Modeling 62, no. 6 (2022): 1376–1387, https://doi.org/10.1021/acs.jcim.1c01467.

[59] P. D. Karp, R. Billington, R. Caspi, et al., “The BioCyc Collection of Microbial Genomes and Metabolic Pathways,” Briefings in Bioinformatics 20, no. 4 (2019): 1085–1093, https://doi.org/10.1093/bib/bbx085.

[60] E. O'Reilly, C. Iglesias, and N. J. Turner, “Monoamine Oxidase–ω-Transaminase Cascade for the Deracemisation and Dealkylation of Amines,” ChemCatChem 6, no. 4 (2014): 992–995.

[61] A. S. Karim and M. C. Jewett, “A Cell-Free Framework for Rapid Biosynthetic Pathway Prototyping and Enzyme Discovery,” Metabolic Engineering 36 (2016): 116–126, https://doi.org/10.1016/j.ymben.2016.03.002.

[62] X. Chen, C. Zhang, R. Zou, K. Zhou, G. Stephanopoulos, and H. P. Too, “Statistical Experimental Design Guided Optimization of a One-Pot Biphasic Multienzyme Total Synthesis of Amorpha-4,11-diene,” Plos One 8, no. 11 (2013): 1–10, https://doi.org/10.1371/journal.pone.0079650.

[63] Y. Cao, T. Zhang, X. Zhao, X. Jia, and B. Li, “MooSeeker: A Metabolic Pathway Design Tool Based on Multi-Objective Optimization Algorithm,” IEEE/ACM Transactions on Computational Biology and Bioinformatics 20, no. 6 (2023): 3609–3622, https://doi.org/10.1109/TCBB.2023.3307363.

[64] A. Sekowska, P. Bertin, and A. Danchin, “Characterization of Polyamine Synthesis Pathway in Bacillus subtilis 168,” Molecular Microbiology 29, no. 3 (1998): 851–858, https://doi.org/10.1046/j.1365-2958.1998.00979.x.

[65] S. Garde, P. K. Chodisetti, and M. Reddy, “Peptidoglycan: Structure, Synthesis, and Regulation,” EcoSal Plus 9, no. 2 (2021): eESP-0002-2021, https://doi.org/10.1128/ecosalplus.ESP-0010-2020.

[66] D. R. Morris and A. B. Pardee, “Multiple Pathways of Putrescine Biosynthesis in Escherichia coli,” Journal of Biological Chemistry 241, no. 13 (1966): 3129–3135.

[67] S. Kim, J. Chen, T. Cheng, et al., “Pubchem 2025 Update,” Nucleic Acids Research 53, no. D1 (2024): D1516–D1525, https://doi.org/10.1093/nar/gkae1059.

[68] C. Raffel, N. Shazeer, A. Roberts, et al., “Exploring the Limits of Transfer Learning With a Unified Text-to-Text Transformer,” Journal of Machine Learning Research 21, no. 140 (2020): 1–67.

RIGHTS & PERMISSIONS

2026 The Author(s). Exploration published by Henan University and John Wiley & Sons Australia, Ltd.