Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors
© The Royal Society of Chemistry 2021. Accurate and rapid evaluation of whether substrates can undergo the desired the transformation is crucial and challenging for both human knowledge and computer predictions. Despite the potential of machine learning in predicting chemical reactivity such as sele...
Main Authors: | , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Royal Society of Chemistry (RSC)
2021
|
Online Access: | https://hdl.handle.net/1721.1/133475 |
_version_ | 1811072912610820096 |
---|---|
author | guan, yanfei Coley, Connor W Wu, Haoyang Duminda, Ranasinghe Heid, Esther Struble, Thomas James Pattanaik, Lagnajit Green, William H Jensen, Klavs F |
author_facet | guan, yanfei Coley, Connor W Wu, Haoyang Duminda, Ranasinghe Heid, Esther Struble, Thomas James Pattanaik, Lagnajit Green, William H Jensen, Klavs F |
author_sort | guan, yanfei |
collection | MIT |
description | © The Royal Society of Chemistry 2021. Accurate and rapid evaluation of whether substrates can undergo the desired the transformation is crucial and challenging for both human knowledge and computer predictions. Despite the potential of machine learning in predicting chemical reactivity such as selectivity, popular feature engineering and learning methods are either time-consuming or data-hungry. We introduce a new method that combines machine-learned reaction representation with selected quantum mechanical descriptors to predict regio-selectivity in general substitution reactions. We construct a reactivity descriptor database based onab initiocalculations of 130k organic molecules, and train a multi-task constrained model to calculate demanded descriptors on-the-fly. The proposed platform enhances the inter/extra-polated performance for regio-selectivity predictions and enables learning from small datasets with just hundreds of examples. Furthermore, the proposed protocol is demonstrated to be generally applicable to a diverse range of chemical spaces. For three general types of substitution reactions (aromatic C-H functionalization, aromatic C-X substitution, and other substitution reactions) curated from a commercial database, the fusion model achieves 89.7%, 96.7%, and 97.2% top-1 accuracy in predicting the major outcome, respectively, each using 5000 training reactions. Using predicted descriptors, the fusion model is end-to-end, and requires approximately only 70 ms per reaction to predict the selectivity from reaction SMILES strings. |
first_indexed | 2024-09-23T09:19:48Z |
format | Article |
id | mit-1721.1/133475 |
institution | Massachusetts Institute of Technology |
language | English |
last_indexed | 2024-09-23T09:19:48Z |
publishDate | 2021 |
publisher | Royal Society of Chemistry (RSC) |
record_format | dspace |
spelling | mit-1721.1/1334752021-10-28T04:45:58Z Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors guan, yanfei Coley, Connor W Wu, Haoyang Duminda, Ranasinghe Heid, Esther Struble, Thomas James Pattanaik, Lagnajit Green, William H Jensen, Klavs F © The Royal Society of Chemistry 2021. Accurate and rapid evaluation of whether substrates can undergo the desired the transformation is crucial and challenging for both human knowledge and computer predictions. Despite the potential of machine learning in predicting chemical reactivity such as selectivity, popular feature engineering and learning methods are either time-consuming or data-hungry. We introduce a new method that combines machine-learned reaction representation with selected quantum mechanical descriptors to predict regio-selectivity in general substitution reactions. We construct a reactivity descriptor database based onab initiocalculations of 130k organic molecules, and train a multi-task constrained model to calculate demanded descriptors on-the-fly. The proposed platform enhances the inter/extra-polated performance for regio-selectivity predictions and enables learning from small datasets with just hundreds of examples. Furthermore, the proposed protocol is demonstrated to be generally applicable to a diverse range of chemical spaces. For three general types of substitution reactions (aromatic C-H functionalization, aromatic C-X substitution, and other substitution reactions) curated from a commercial database, the fusion model achieves 89.7%, 96.7%, and 97.2% top-1 accuracy in predicting the major outcome, respectively, each using 5000 training reactions. Using predicted descriptors, the fusion model is end-to-end, and requires approximately only 70 ms per reaction to predict the selectivity from reaction SMILES strings. 2021-10-27T19:53:02Z 2021-10-27T19:53:02Z 2021 2021-06-09T16:42:56Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/133475 en 10.1039/d0sc04823b Chemical Science Creative Commons Attribution 3.0 unported license https://creativecommons.org/licenses/by/3.0/ application/pdf Royal Society of Chemistry (RSC) Royal Society of Chemistry (RSC) |
spellingShingle | guan, yanfei Coley, Connor W Wu, Haoyang Duminda, Ranasinghe Heid, Esther Struble, Thomas James Pattanaik, Lagnajit Green, William H Jensen, Klavs F Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors |
title | Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors |
title_full | Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors |
title_fullStr | Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors |
title_full_unstemmed | Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors |
title_short | Regio-selectivity prediction with a machine-learned reaction representation and on-the-fly quantum mechanical descriptors |
title_sort | regio selectivity prediction with a machine learned reaction representation and on the fly quantum mechanical descriptors |
url | https://hdl.handle.net/1721.1/133475 |
work_keys_str_mv | AT guanyanfei regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT coleyconnorw regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT wuhaoyang regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT dumindaranasinghe regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT heidesther regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT strublethomasjames regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT pattanaiklagnajit regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT greenwilliamh regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors AT jensenklavsf regioselectivitypredictionwithamachinelearnedreactionrepresentationandontheflyquantummechanicaldescriptors |