{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,18]],"date-time":"2026-09-18T06:45:39Z","timestamp":1789713939270,"version":"4.0.1"},"reference-count":37,"publisher":"American Chemical Society (ACS)","issue":"7","license":[{"start":{"date-parts":[[2023,3,22]],"date-time":"2023-03-22T00:00:00Z","timestamp":1679443200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2023,3,22]],"date-time":"2023-03-22T00:00:00Z","timestamp":1679443200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"},{"start":{"date-parts":[[2023,3,22]],"date-time":"2023-03-22T00:00:00Z","timestamp":1679443200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-045"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2023,4,10]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Retrosynthesis prediction, the task of identifying reactant molecules that can be used to synthesize product molecules, is a fundamental challenge in organic chemistry and related fields. To address this challenge, we propose a novel graph-to-graph transformation model, G2GT. The model is built on the standard transformer structure and utilizes graph encoders and decoders. Additionally, we demonstrate the effectiveness of self-training, a data augmentation technique that utilizes unlabeled molecular data, in improving the performance of the model. To further enhance diversity, we propose a weak ensemble method, inspired by reaction-type labels and ensemble learning. This method incorporates beam search, nucleus sampling, and top-k sampling to improve inference diversity. A simple ranking algorithm is employed to retrieve the final top-10 results. We achieved new state-of-the-art results on both the USPTO-50K data set, with a top-1 accuracy of 54%, and the larger more challenging USPTO-Full data set, with a top-1 accuracy of 49.3% and competitive top-10 results. Our model can also be generalized to all other graph-to-graph transformation tasks. Data and code are available at https:\/\/github.com\/Anonnoname\/G2GT_2<\/jats:p>","DOI":"10.1021\/acs.jcim.2c01302","type":"journal-article","created":{"date-parts":[[2023,3,22]],"date-time":"2023-03-22T08:19:51Z","timestamp":1679473191000},"page":"1894-1905","source":"Crossref","is-referenced-by-count":27,"title":["G2GT: Retrosynthesis\nPrediction with Graph-to-Graph\nAttention Neural Network and Self-Training"],"prefix":"10.1021","volume":"63","author":[{"given":"Zaiyun","family":"Lin","sequence":"first","affiliation":[{"name":"Stone Wise , Room 918, Eighth Floor, Building 1, No. 6 Danling Street , , ,","place":["Haidian District, Beijing, China, 100089"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8125-0367","authenticated-orcid":true,"given":"Shiqiu","family":"Yin","sequence":"additional","affiliation":[{"name":"Stone Wise , Room 918, Eighth Floor, Building 1, No. 6 Danling Street , , ,","place":["Haidian District, Beijing, China, 100089"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lei","family":"Shi","sequence":"additional","affiliation":[{"name":"Stone Wise , Room 918, Eighth Floor, Building 1, No. 6 Danling Street , , ,","place":["Haidian District, Beijing, China, 100089"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Wenbiao","family":"Zhou","sequence":"additional","affiliation":[{"name":"Stone Wise , Room 918, Eighth Floor, Building 1, No. 6 Danling Street , , ,","place":["Haidian District, Beijing, China, 100089"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2520-3923","authenticated-orcid":true,"given":"Yingsheng John","family":"Zhang","sequence":"additional","affiliation":[{"name":"Stone Wise , Room 918, Eighth Floor, Building 1, No. 6 Danling Street , , ,","place":["Haidian District, Beijing, China, 100089"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"316","published-online":{"date-parts":[[2023,3,22]]},"reference":[{"key":"2026081804435025800_cit1","doi-asserted-by":"publisher","first-page":"408","DOI":"10.1126\/science.3838594","article-title":"Computer-assisted analysis in organic\nsynthesis","volume":"228","author":"Corey","year":"1985","journal-title":"Science"},{"key":"2026081804435025800_cit2","doi-asserted-by":"publisher","first-page":"593","DOI":"10.1021\/ci800228y","article-title":"Route designer:\na retrosynthetic analysis tool utilizing\nautomated retrosynthetic rule generation","volume":"49","author":"Law","year":"2009","journal-title":"J.\nChem. Inf. Model."},{"key":"2026081804435025800_cit3","doi-asserted-by":"publisher","first-page":"434","DOI":"10.1021\/acscentsci.7b00064","article-title":"Prediction\nof organic reaction outcomes using machine learning","volume":"3","author":"Coley","year":"2017","journal-title":"ACS Cent. Sci."},{"key":"2026081804435025800_cit4","unstructured":"Seidl, P.; Renz, P.; Dyubankova, N.; Neves, P.; Verhoeven, J.; Segler, M.; Wegner, J. K.; Hochreiter, S.; Klambauer, G.\n          Modern hopfield networks for few-and\nzero-shot reaction template prediction. arXiv\nPreprint, arXiv:2104.03279, 2021.10.48550\/arXiv.2104.03279."},{"key":"2026081804435025800_cit5","doi-asserted-by":"publisher","first-page":"560","DOI":"10.1002\/wcms.1140","article-title":"Automatic\nreaction mapping and reaction\ncenter detection","volume":"3","author":"Chen","year":"2013","journal-title":"WIREs Comput. Mol. Sci."},{"key":"2026081804435025800_cit6","doi-asserted-by":"crossref","unstructured":"Liu, X.; Li, P.; Song, S.\n          Decomposing Retrosynthesis into Reactive Center Prediction\nand Molecule\nGeneration. bioRxiv Preprint, 2020. 10.1101\/677849.","DOI":"10.1101\/677849"},{"key":"2026081804435025800_cit7","unstructured":"Somnath, V. R.; Bunne, C.; Coley, C.; Krause, A.; Barzilay, R.\n          Learning graph models for retrosynthesis\nprediction. In 35th Conference on Neural Information\nProcessing Systems (NeurIPS 2021), 2021, Vol. 34."},{"key":"2026081804435025800_cit8","doi-asserted-by":"publisher","first-page":"129845","DOI":"10.1016\/j.cej.2021.129845","article-title":"RetroPrime: ADiverse, plausible and\nTransformer-based method for Single-Step retrosynthesis predictions","volume":"420","author":"Wang","year":"2021","journal-title":"Chem. Eng. J."},{"key":"2026081804435025800_cit9","first-page":"8818","volume-title":"A Graph to Graphs Framework for Retrosynthesis Prediction","author":"Shi","year":"2020"},{"key":"2026081804435025800_cit10","first-page":"11248","article-title":"Retroxpert: Decompose retrosynthesis\nprediction like a chemist","volume":"33","author":"Yan","year":"2020","journal-title":"Adv. Neural Inf.\nProcess. Syst."},{"key":"2026081804435025800_cit11","unstructured":"Gao, Z.; Tan, C.; Wu, L.; Li, S. Z.\n          SemiRetro: Semi-template framework boosts deep retrosynthesis\nprediction. arXiv Preprint, arXiv:2202.08205, 2022. 10.48550\/arXiv.2202.08205."},{"key":"2026081804435025800_cit12","doi-asserted-by":"publisher","first-page":"1612","DOI":"10.1021\/jacsau.1c00246","article-title":"Deep retrosynthetic reaction prediction using local reactivity and\nglobal attention","volume":"1","author":"Chen","year":"2021","journal-title":"JACS Au"},{"key":"2026081804435025800_cit13","doi-asserted-by":"publisher","first-page":"1103","DOI":"10.1021\/acscentsci.7b00303","article-title":"Retrosynthetic reaction\nprediction using neural sequence-to-sequence\nmodels","volume":"3","author":"Liu","year":"2017","journal-title":"ACS Cent. Sci."},{"key":"2026081804435025800_cit14","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1038\/s41467-020-19266-y","article-title":"State-of-the-art augmented\nNLP transformer models for direct and single-step retrosynthesis","volume":"11","author":"Tetko","year":"2020","journal-title":"Nat. Commun."},{"key":"2026081804435025800_cit15","doi-asserted-by":"publisher","first-page":"47","DOI":"10.1021\/acs.jcim.9b00949","article-title":"Predicting retrosynthetic\nreactions using self-corrected transformer neural networks","volume":"60","author":"Zheng","year":"2020","journal-title":"J. Chem. Inf. Model."},{"key":"2026081804435025800_cit16","doi-asserted-by":"publisher","first-page":"3273","DOI":"10.1021\/acs.jcim.1c00537","article-title":"Molecule edit graph attention network: modeling chemical\nreactions as sequences of graph edits","volume":"61","author":"Sacha","year":"2021","journal-title":"J. Chem.\nInf. Model."},{"key":"2026081804435025800_cit17","unstructured":"Chen, B.; Shen, T.; Jaakkola, T. S.; Barzilay, R.\n          Learning to make generalizable\nand diverse predictions for retrosynthesis. arXiv Preprint, arXiv:1910.09688, 2019. 10.48550\/arXiv.1910.09688"},{"key":"2026081804435025800_cit18","doi-asserted-by":"publisher","first-page":"31","DOI":"10.1021\/ci00057a005","article-title":"SMILES, a\nchemical language and information system. 1. Introduction to methodology\nand encoding rules","volume":"28","author":"Weininger","year":"1988","journal-title":"J. Chem. Inf. Comput. Sci."},{"key":"2026081804435025800_cit19","first-page":"na","article-title":"Attention\nis all you need","volume":"30","author":"Vaswani","year":"2017","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2026081804435025800_cit20","first-page":"na","article-title":"Do Transformers Really Perform Badly\nfor Graph Representation","volume":"34","author":"Ying","year":"2021","journal-title":"Adv. Neural Inf. Process.\nSyst"},{"key":"2026081804435025800_cit21","unstructured":"Dwivedi, V. P.; Bresson, X.\n          A generalization\nof transformer networks to graphs. arXiv Preprint, arXiv:2012.09699, 2020. 10.48550\/arXiv.2012.09699."},{"key":"2026081804435025800_cit22","unstructured":"Dwivedi, V. P.; Joshi, C. K.; Laurent, T.; Bengio, Y.; Bresson, X.\n          Benchmarking graph\nneural networks. arXiv Preprint,\narXiv:2003.00982, 2020. 10.48550\/arXiv.2003.00982."},{"key":"2026081804435025800_cit23","first-page":"na","article-title":"Rethinking graph transformers with\nspectral attention","volume":"34","author":"Kreuzer","year":"2021","journal-title":"Adv. Neural Inf. Process.\nSyst."},{"key":"2026081804435025800_cit24","unstructured":"Velic\u030ckovic\u0301, P.; Cucurull, G.; Casanova, A.; Romero, A.; Lio, P.; Bengio, Y.\n          Graph attention\nnetworks. arXiv Preprint, arXiv:1710.10903, 2017. 10.48550\/arXiv.1710.10903."},{"key":"2026081804435025800_cit25","first-page":"12559","article-title":"Self-supervised graph transformer\non large-scale molecular data","volume":"33","author":"Rong","year":"2020","journal-title":"Adv. Neural Inf.\nProcess. Syst."},{"key":"2026081804435025800_cit26","unstructured":"Hu, W.; Liu, B.; Gomes, J.; Zitnik, M.; Liang, P.; Pande, V.; Leskovec, J.\n          Strategies for pre-training graph neural networks. arXiv Preprint, arXiv:1905.12265, 2019. 10.48550\/arXiv.1905.12265."},{"key":"2026081804435025800_cit27","doi-asserted-by":"crossref","first-page":"127","DOI":"10.1021\/bk-2014-1164.ch008","article-title":"The Making of Reaxys\ue5f8Towards\nUnobstructed Access to Relevant Chemistry Information","volume-title":"The Future of the History of Chemical Information","author":"Lawson","year":"2014"},{"key":"2026081804435025800_cit28","doi-asserted-by":"publisher","first-page":"583","DOI":"10.1038\/s41586-021-03819-2","article-title":"Highly accurate protein\nstructure prediction with AlphaFold","volume":"596","author":"Jumper","year":"2021","journal-title":"Nature"},{"key":"2026081804435025800_cit29","unstructured":"Holtzman, A.; Buys, J.; Du, L.; Forbes, M.; Choi, Y.\n          Curious case of\nneural text degeneration. arXiv Preprint, arXiv:1904.09751, 2019 .10.48550\/arXiv.1904.09751."},{"key":"2026081804435025800_cit30","unstructured":"Smith, S. L.; Kindermans, P.-J.; Ying, C.; Le, Q. V.\n          Don\u2019t decay the learning\nrate, increase the batch size. arXiv Preprint, arXiv:1711.00489, 2017. 10.48550\/arXiv.1711.00489."},{"key":"2026081804435025800_cit31","unstructured":"Lowe, D. M.\n          \n          Extraction\nof chemical structures and reactions from the literature. Ph.D. Thesis, University of Cambridge, 2012."},{"key":"2026081804435025800_cit32","first-page":"na","article-title":"Retrosynthesis prediction\nwith conditional graph logic network","volume":"32","author":"Dai","year":"2019","journal-title":"Adv. Neural\nInf. Process. Syst."},{"key":"2026081804435025800_cit33","unstructured":"Reaxys. www.reaxys.com, Reaxys is a registered\ntrademark of RELX Intellectual Properties SA used under license."},{"key":"2026081804435025800_cit34","doi-asserted-by":"publisher","first-page":"5966","DOI":"10.1002\/chem.201605499","article-title":"Neural-symbolic machine learning\nfor retrosynthesis\nand reaction prediction","volume":"23","author":"Segler","year":"2017","journal-title":"Chem.\ue5f8Eur. J."},{"key":"2026081804435025800_cit35","unstructured":"Sun, R.; Dai, H.; Li, L.; Kearnes, S.; Dai, B.\n          Energy-based View of Retrosynthesis. arXiv Preprint, arXiv:2007.13437, 2020. 10.48550\/arXiv.2007.13437."},{"key":"2026081804435025800_cit36","first-page":"817","volume-title":"A Transformer\nModel for Retrosynthesis","author":"Karpov","year":"2019"},{"key":"2026081804435025800_cit37","volume-title":"Compendium\nof Chemical Terminology","author":"Nic","year":"1997"}],"container-title":["Journal of Chemical Information\nand Modeling"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pubs.acs.org\/doi\/pdf\/10.1021\/acs.jcim.2c01302","content-type":"application\/pdf","content-version":"vor","intended-application":"unspecified"},{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article-pdf\/63\/7\/1894\/8909396\/ci2c01302.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article-pdf\/63\/7\/1894\/8909396\/ci2c01302.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,18]],"date-time":"2026-08-18T09:00:48Z","timestamp":1787043648000},"score":1,"resource":{"primary":{"URL":"https:\/\/pubs.acs.org\/jcisd8\/article\/63\/7\/1894\/850337\/G2GT-Retrosynthesis-Prediction-with-Graph-to-Graph"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,3,22]]},"references-count":37,"journal-issue":{"issue":"7","published-online":{"date-parts":[[2023,3,22]]},"published-print":{"date-parts":[[2023,4,10]]}},"URL":"https:\/\/doi.org\/10.1021\/acs.jcim.2c01302","relation":{},"ISSN":["1549-9596","1549-960X"],"issn-type":[{"value":"1549-9596","type":"print"},{"value":"1549-960X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,3,22]]}}}