About the Seminar

The outcomes of chemical reactions – yield, selectivity, and rate – are influenced by the shape and size of the molecules involved. To make reactions more efficient, greener, and faster, chemists often turn to catalysis. Catalysts are species that recognize and bind to reactants, and while their discovery is traditionally the result of trial-and-error, efforts are underway in our laboratory to design new catalysts computationally. These design efforts require us to convert molecular structures into digital fingerprints, unique representations that capture both two and three-dimensional characteristics of molecules in a way that can be processed computationally and used in statistical and machine learning workflows. We have developed a unique representation that captures the non-uniformity of molecular structures as well as pinpointing how their shape is distributed relative to a reaction site. This allows us to convert discrete molecules into continuous digital inputs, gain unprecedented insight into how molecular shape influences catalytic reaction outcomes, and predict the outcomes of new reactions. In addition to gaining new understanding, this new way of describing molecules sets the scene for the automated, and possibly fully autonomous, catalyst discovery that enables new chemical processes to be performed.