Catalysis Meets Machine Learning: A Guide to Data-Driven Discovery and Design

Abstract

Machine learning (ML) has rapidly become an indispensable tool across the chemical sciences, offering powerful methods to extract patterns from data and make accurate predictions in complex, multidimensional systems. In organometallic catalysis, its potential is particularly evident: while transition-metal catalysed reactions are at the core of modern synthesis, their design and optimization remain challenging due to the vastness of chemical space, the scarcity of standardized data, and the intricate interplay of steric, electronic, and mechanistic factors. This contribution aims to provide chemists with both a conceptual and practical entrypoint into the field, beginning with a concise introduction to the principles of ML and its most widely used algorithms. It then surveys recent advances by structuring the discussion according to applications: optimization of reaction conditions, mechanistic elucidation, ligand classification and design, stereocontrol, and the discovery of novel catalysts. By combining methodological insights with case studies, we highlight how ML can reduce experimental workload, enhance mechanistic understanding, and guide rational catalyst development, while also outlining current limitations and future opportunities at the interface of data science and catalysis.