Selective functionalization of hindered meta-C–H bond of o-alkylaryl ketones promoted by automation and deep learning

Qiu, Jia; Xie, Jiancong; Su, Shimin; Gao, Yadong; Han, Meng; Yang, Yuedong; Liao, Kuangbiao

doi:10.1016/j.chempr.2022.08.015

Cited by 19 publications

(15 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…HTE and graph neural networks have previously been used for identifying substrates suitable for C-H activation. [37] This present study extends this original approach by (i) utilizing HTE and graph neural networks for drug molecules, (ii) introducing a literature search strategy that enables the selection of a structurally diverse set of substrates and ideal plate reaction screening conditions, and (iii) introducing a flexible geometric deep learning approach that considers the influence of steric and electronic effects of the substrates and allows the prediction of reaction outcome, yield, and regioselectivity.…”

Section: Discussionmentioning

confidence: 98%

“…For the binary reaction outcome as observed for reaction yield prediction, the same trend can be perceived, i.e., GTNNs slightly outperformed the ECFP4NN and GNN models and that QM augmentation and 3D information did not affect the performance of the models (Table 1). Figure 5E shows three drugs (1,25,29) and three fragments (37,38,45) that were predicted by GTNN3D to yield successful reaction outcomes for unseen substrates. The main reaction products of these six substrates were isolated with reaction yields ranging from 5% to 90% (see SI11).…”

Section: Reaction Yield and Reaction Outcomementioning

confidence: 99%

“…[36] The combination of graph machine learning with HTE enabled the optimization of reaction conditions for C-H activation of organic substrates. [37] However, this approach has been limited to small molecular structures, making the utilization of such models challenging for applications involving structurally more intricate drug-like molecules. [38] Importantly, the influence of steric and electronic effects on the model performance for C-H activation reactions and their application to regioselectivity for molecules with multiple aromatic ring systems remains unexplored.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning

Nippa

Atz

Hohler

et al. 2022

Preprint

View full text Add to dashboard Cite

Late-stage functionalization (LSF) is an economical approach to optimize the properties of drug candidates. However, the chemical complexity of drug molecules often makes late-stage diversification challenging. To address this problem, an LSF platform based on geometric deep learning and high-throughput reaction screening was developed. Considering borylation as a critical step in LSF, the computational model correctly predicted the reactivity of 81% of novel substrates, while reaction yields for diverse reaction conditions were predicted with a mean absolute error margin of 4–5%. The regioselectivity of the major products was accurately captured in up to 90% of the cases studied. When applied to 23 diverse commercial drug molecules, the platform successfully identified numerous opportunities for structural diversification. The influence of steric and quantum mechanical information on model performance was quantified and a new comprehensive simple user-friendly reaction format (SURF) is introduced which proved to be a key enabler for seamlessly integrating deep learning and high-throughput experimentation (HTE) for LSF.

show abstract

Section: Discussionmentioning

confidence: 98%

Section: Reaction Yield and Reaction Outcomementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning

Nippa

Atz

Hohler

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…For the binary reaction outcome as observed for reaction yield prediction, a similar trend can be perceived, i.e., GTNNs slightly outperformed (90.9 -91.8% AUC) the ECFP4NN (89.3% AUC) and GNN (87.5-89.1% AUC) models and that QM augmentation and 3D information did not affect the performance of the models (Table 1). Figure 5E shows three drugs (1, 25, 29) and three fragments (37,38,45) that were predicted by GTNN3DQM to yield successful reaction outcomes for unseen substrates. The main reaction products of these six substrates were isolated with reaction yields ranging from 5% to 90% (see SI11).…”

Section: Reaction Yield and Reaction Outcomementioning

confidence: 99%

“…The LSF informer library containing 23 structurally diverse approved drugs (1, 14-36) complemented with 12 fragments (37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48) and five idealized substrates (49)(50)(51)(52)(53) yielded a data set covering essential chemical motives relevant in drug discovery. A functional group analysis revealed that 33 (82.5%) of the 40 most abundant functional groups extracted from the 1174 drug molecules are covered by the LSF informer library.…”

Section: Regioselectivitymentioning

confidence: 99%

Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning

Nippa

Atz

Hohler

et al. 2022

Preprint

View full text Add to dashboard Cite

Late-stage functionalization (LSF) is an economical approach to optimize the properties of drug candidates. However, the chemical complexity of drug molecules often makes late-stage diversification challenging. To address this problem, an LSF platform based on geometric deep learning and high-throughput reaction screening was developed. Considering borylation as a critical step in LSF, the computational model predicted reaction yields for diverse reaction conditions with a mean absolute error margin of 4–5%, while the reactivity of novel reactions with known and unknown substrates were classified with a balanced accuracy of 92% and 67%, respectively. The regioselectivity of the major products was accurately captured in up to 90% of the cases studied. When applied to 23 diverse commercial drug molecules, the platform successfully identified numerous opportunities for structural diversification. The influence of steric and quantum mechanical information on model performance was quantified and a new comprehensive simple user-friendly reaction format (SURF) is introduced which proved to be a key enabler for seamlessly integrating deep learning and high-throughput experimentation (HTE) for LSF.

show abstract

High‐Throughput Experimentation and Machine Learning‐Assisted Optimization of Iridium‐Catalyzed Cross‐Dimerization of Sulfoxonium Ylides

Xu,

Gao,

et al. 2023

Angewandte Chemie

Self Cite

View full text Add to dashboard Cite

A novel and convenient approach that combines high‐throughput experimentation (HTE) with machine learning (ML) technologies to achieve the first selective cross‐dimerization of sulfoxonium ylides via iridium catalysis is presented. A variety of valuable amide‐, ketone‐, ester‐, and N‐heterocycle‐substituted unsymmetrical E‐alkenes are synthesized in good yields with high stereoselectivities. This mild method avoids the use of diazo compounds and is characterized by simple operation, high step‐economy, and excellent chemoselectivity and functional group compatibility. The combined experimental and computational studies identify an amide‐sulfoxonium ylide as a carbene precursor. Furthermore, a comprehensive exploration of the reaction space is also performed (600 reactions) and a machine learning model for reaction yield prediction has been constructed.

show abstract

Selective functionalization of hindered meta-C–H bond of o-alkylaryl ketones promoted by automation and deep learning

Cited by 19 publications

References 48 publications

Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning

Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning

Enabling late-stage drug diversification by high-throughput experimentation with geometric deep learning

High‐Throughput Experimentation and Machine Learning‐Assisted Optimization of Iridium‐Catalyzed Cross‐Dimerization of Sulfoxonium Ylides

Contact Info

Product

Resources

About