Data-driven catalyst optimization for stereodivergent asymmetric synthesis by iridium/boron hybrid catalysis

Abstract

Asymmetric catalysis enabling divergent control of multiple stereocenters remains challenging in synthetic organic chemistry. Although machine-learning-based optimization of molecular catalysis is a rapidly growing research field, the use of such approaches for catalyst design to achieve stereodivergent asymmetric synthesis producing multiple reaction outcomes, such as constitutional selectivity, diastereoselectivity, and enantioselectivity, is unprecedented. Here, we report the straightforward identification of asymmetric two-component iridium/boron hybrid catalyst systems for α- C -allylation of carboxylic acids. Structural optimization of the chiral ligands for iridium catalysts was driven by molecular-field-based regression analysis with a dataset containing overall 32 molecular structures. The catalyst systems enabled selective access to all the possible isomers of chiral carboxylic acids bearing contiguous stereocenters. This chemoselective and stereodivergent, asymmetric catalysis is applicable to late-stage structural modifications of drugs and their derivatives. Practical regression-based, data-driven, chiral catalyst design and optimization method 3-Dimensional, quantitative structure-selectivity relationships and molecular field analysis Control/improvement of four reaction outcomes with regression analysis Stereodivergent asymmetric synthesis of α-allyl carboxylic acids Chen et al. report that data-driven catalyst design facilitates stereodivergent asymmetric synthesis, which remains challenging in organic synthesis. This research indicates that digital transformations (DXs) of organic synthesis (e.g., the use of machine-leaning and materials-informatics techniques) enable analysis and control of complicated reactions, opening new avenues in molecular catalysis.