Scholarlink

<p>(<i>R</i>)-( +)-1-Boc-3-aminopyrrolidine is an indispensable chiral intermediate for central nervous system drugs and Janus kinase inhibitors, yet its traditional chemical synthesis is plagued by harsh conditions, heavy pollution, and inadequate stereocontrol. Biocatalysis with transaminases (TAs) offers an attractive green alternative, but discovering optimal enzymes remains a major bottleneck. Here, we report a synergistic integration of machine learning (ML) with experimental validation to break this bottleneck. By ML-enabled evaluation of seven phylogenetically diverse TAs, we rapidly identified TA-6 from <i>Arthrobacter</i> sp. as an outstanding (<i>R</i>)-selective biocatalyst for the asymmetric amination of prochiral N-Boc-3-pyrrolidinone. Notably, using a whole-cell catalyst loading of only 10&#xa0;g/L (DCW), TA-6 converted 100&#xa0;g/L of ketone substrate on a 20-L preparative scale, delivering the desired (<i>R</i>)-enantiomer in 99.5% enantiomeric excess. This work not only provides a practical, scalable, and sustainable biocatalytic route to a pharmaceutically essential chiral amine, but also demonstrates that machine learning enables a paradigm shift from trial-and-error to data-driven precision in asymmetric synthesis, dramatically accelerating enzyme discovery.</p> Graphical Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From Computational Screening to Preparative Scale: A Transaminase Identified by Machine Learning for (R)-1-Boc-3-aminopyrrolidine

  • 期刊论文

  • 发表日期：

    2026年8月3日

查看全文
  • MiMi Duan,
  • Xiafeng Lu,
  • Yan Wu,
  • Peibing Huang,
  • Wen Zheng,
  • Sainan Zhang,
  • Kai Liu,
  • Changlong Jia,
  • Feng Du,
  • Min Cao,
  • Muqing Ma,
  • Zhibo Luo

摘要

(R)-( +)-1-Boc-3-aminopyrrolidine is an indispensable chiral intermediate for central nervous system drugs and Janus kinase inhibitors, yet its traditional chemical synthesis is plagued by harsh conditions, heavy pollution, and inadequate stereocontrol. Biocatalysis with transaminases (TAs) offers an attractive green alternative, but discovering optimal enzymes remains a major bottleneck. Here, we report a synergistic integration of machine learning (ML) with experimental validation to break this bottleneck. By ML-enabled evaluation of seven phylogenetically diverse TAs, we rapidly identified TA-6 from Arthrobacter sp. as an outstanding (R)-selective biocatalyst for the asymmetric amination of prochiral N-Boc-3-pyrrolidinone. Notably, using a whole-cell catalyst loading of only 10 g/L (DCW), TA-6 converted 100 g/L of ketone substrate on a 20-L preparative scale, delivering the desired (R)-enantiomer in 99.5% enantiomeric excess. This work not only provides a practical, scalable, and sustainable biocatalytic route to a pharmaceutically essential chiral amine, but also demonstrates that machine learning enables a paradigm shift from trial-and-error to data-driven precision in asymmetric synthesis, dramatically accelerating enzyme discovery.

Graphical Abstract