September 13, 2019
Toward automatic and efficient catalytic reaction development process using AI
RIKEN CSRS has succeeded in designing a molecule that shows improved enantioselectivity in asymmetric catalysis through data analysis using intermediate structures in an enantio-determining step.
Artificial intelligence and data science are expected to automate and accelerate the development of catalytic reactions that are currently conducted through reseachers’ trial and error. However, data science techniques can predict only reaction outcomes of molecules that lie within the range of the data used to generate the statistical models (training data). It is therefore not easy to perform data-driven design of molecules that exhibit superior perfomance in comparison to those in the training dataset.
RIKEN CSRS researchers discovered that, in asymmetric catalysis, data analysis using intermediate structures in an enantio-determining step enables extraction and visualization of structural information for design of molecules that show improved enantioselectivity. Based on the visualized structural information, the researchers designed substrate and catalyst molecules. They confirmed experimentally that the designed substrate exhibited higher enantioselectivity than those in the training data.
Bulletin of the Chemical Society of Japan doi:10.1246/bcsj.20190132
S. Yamaguchi, M. Sodeoka,
"Molecular Field Analysis Using Intermediates in Enantio-Determining Steps Can Extract Information for Data-Driven Molecular Design in Asymmetric Catalysis".
Shigeru Yamaguchi; Special Postdoctoral Researcher
Mikiko Sodeoka; Group Director
Catalysis and Integrated Research Group