Designing chiral catalysts via machine learning using computational screening data
February 7, 2022
Toward digital transformation of organic syntheses
A research group of RIKEN CSRS and Kanazawa University successfully performed in silico chiral catalyst design to improve enantioselectivity by combining transition-state calculations and machine learning.
Because biological activity of molecules is highly dependent on their stereochemistry, acceleration of the discovery of chiral catalysts enabling precise control of product stereoselectivity are important to develop drug molecules. Although utilization of artificial intelligence (AI) has been attracting tremendous attention in recent years for this purpose, however, experimental data are generally needed to construct AI. If catalysts can be designed using computational screening data collected by quantum chemical calculations, various applications can be expected such as the efficient development of reaction systems that involve expensive and synthetically difficult catalysts.
In this study, the collaborative group successfully improved product enantioselectivity in a model asymmetric catalytic reaction system through data-driven in silico catalyst design using only 30 training samples obtained through transition-state calculations.
This finding will contribute to enhance digital transformation (DX) of organic syntheses.
- Original article
- Bulletin of the Chemical Society of Japan doi:10.1246/bcsj.20210349
- M. Mukai, K. Nagao, S. Yamaguchi, H. Ohmiya,
- "Molecular Field Analysis Using Computational-Screening Data in Asymmetric N-Heterocyclic Carbene-Copper Catalysis toward Data-driven in silico Catalyst Optimization".
- Shigeru Yamaguchi
- Visiting Scientist
- Advanced Catalysis Research Group