丁香园AV

Abstract

The optimization of catalytic, enantioselective reactions is challenging as it involves the empirical evaluation of several different reaction components (e.g., reactant, catalyst, solvent, etc.) to determine the best set of conditions. Our goal has been to develop data science-based tools that streamline this process by minimizing the number of experiments needed while maximizing the proportion that yields high levels of enantiomeric excess. In this talk, I will describe how we apply a diverse set of machine learning algorithms to aid in the identification of optimal reaction conditions and general catalyst systems. A significant portion of this talk will focus on our experimental efforts in evaluating these tools for developing enantioselective reactions. Continuous improvement of this workflow has directed us to develop bespoke machine learning algorithms for top-down mechanistic analysis. These techniques will also be covered and have been vetted in the complex organometallic space, serving as a useful complement to the traditional bottom-up approach. Finally, I will describe our latest efforts in developing novel catalyst structures and demonstrate their utility in various transformations.