Beyond Predefined Ligand Libraries: A Genetic Algorithm Approach for De Novo Discovery of Catalysts for the Suzuki Coupling Reactions

This study introduces a novel approach for the unrestricted de novo design of transition metal catalysts, leveraging the power of genetic algorithms (GAs) and density functional theory (DFT) calculations. By focusing on the Suzuki reaction, known for its significance in forming carbon-carbon bonds, we demonstrate the effectiveness of fragment-based and graph-based genetic algorithms in identifying novel ligands for palladium-based catalytic systems. Our research highlights the capability of these algorithms to generate ligands with desired thermodynamic properties, moving beyond the restriction of enumerated chemical libraries. Limitations in the applicability of machine learning models are overcome by calculating thermodynamic properties from first principle. The inclusion of synthetic accessibility scores further refines the search, steering it towards more practically feasible ligands. Through the examination of both palladium and alternative transition metal catalysts like copper and silver, our findings reveal the algorithms' ability to uncover unique catalyst structures within the target energy range, offering insights into the electronic and steric effects necessary for effective catalysis. This work not only proves the potential of genetic algorithms in the cost-effective and scalable discovery of new catalysts but also sets the stage for future exploration beyond predefined chemical spaces, enhancing the toolkit available for catalyst design.