Modeling Reaction Feasibility to Identify New Opportunities for Sigmatropic Rearrangements

Abstract

Defining the scope and limitations of synthetic methods enables bolder, lower-risk synthetic design. Computation can facilitate this if able to anticipate reaction success or failure. Herein, we combine a machine-learned interatomic potential with the TS-tools package for high-throughput prediction of transition state geometries and barriers for hypothetical reactions. We focus on sigmatropic rearrangements, which are of synthetic interest, enable non-obvious retrosynthetic disconnections, and are explicable by electronic structure theory. Benchmarking against reference theory and experimental kinetics establishes quantitative accuracy, and retrospective case studies illustrate utility in synthesis planning. Prospectively, we identify and experimentally realize a methylenation/Claisen sequence toward the natural product (±)-norbrachycoumarin and a UR-144 degradant. A coarse-grained kinetic study across ten substrates, seven installing all-carbon quaternary centers β to a ketone, reveals good agreement between computed barriers and experimental outcomes. Collectively, this work demonstrates how computational reaction feasibility modeling uncovers opportunities to apply existing methodologies to valuable chemical spaces.