Computational Prediction of Enantioselectivity in Metal-Catalyzed Reactions

In this project, Professors Olaf Wiest and Paul Helquist at the University of Notre Dame are developing and applying methods for the prediction of which of several possible products of a chemical reaction is formed. Many reactions can form two different products, called enantiomers, that are mirror images of each other. Selectively making only one of them is an important challenge, especially in the pharmaceutical industry, because only one of these compounds will have the desired activity while the other one will be inactive or, in some cases, even toxic. Previously, this challenge was commonly addressed through a time-consuming trial-and-error approach, that even with modern automation is expensive and inefficient. The approach developed by Professors Wiest and Helquist together with collaborators at a large pharmaceutical company over the last decade uses fast computational methods to predict which catalysts will yield the desired enantiomer in a pure form before any experiments are run. This minimizes cost and waste while accelerating the manufacturing of high-value chemicals such as pharmaceuticals, leading to lower costs and other increasing economic competitiveness advantages. In the present grant period, the method is combined with recent development in artificial intelligence (AI) to make the method more broadly applicable and further accelerate it. The collaboration of the academic and private sector also provides hands-on training for the trainees involved in the project through internships at and regular interactions with a large pharmaceutical company, offering unique professional development opportunities for our undergraduate and graduate students and ensuring a competitive scientific workforce. Professors Wiest and Helquist, together with an industrial collaborator, have over the last decade developed the quantum guided molecular mechanics (Q2MM) method that provides for rapid fitting of transition state force fields (TSFFs) that allow the rapid and accurate prediction of the stereochemical outcome of enantioselective reactions, especially those that catalyzed by transition metal such as rhodium, palladium, iridium or ruthenium. The present grant will combine the Q2MM workflow with the machine learned intermolecular potential AIMNet2 to accelerate the generation of training data for Q2MM and exploit the recently developed QFUERZA and hybrid optimizer algorithms to automate the workflow, allowing more scientists to use this powerful approach. We will also continue to develop the CatVS method for the automated generation and conformational search of transition states of libraries of ligand/substrate combinations. All new developments will be made available to the scientific community free of charge at github.cim/q2mm. The method will be demonstrated for two synthetically important reactions, palladium-catalyzed conjugate additions to unsaturated lactams and the design of a new class of ruthenium piano stool ligands. In the latter case, the designed ligands will be synthesized and tested for the case of enantioselective hydrogen atom transfer to ketones. The combined computational and experimental approach, combined with the close collaboration with a large pharmaceutical company, provides a unique interdisciplinary training environment for undergraduate and graduate students. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.