Comparing quantitative prediction methods for the discovery of small-molecule chiral catalysts

Abstract

Listen

Translate article

Advances in density functional theory (DFT) mean that it is now possible to study catalytic reactions with sufficient accuracy that the results compare favourably with experiment. These high-level calculations have been applied to understand and predict variations in catalytic performance from one catalyst to another, but can require substantial computational resources. By contrast, multivariate linear regression (MLR) methods are rapidly becoming versatile, statistical tools for predicting and understanding the roles of catalysts and substrates and act as a useful complement to complex transition state calculations, with a substantially lower computational cost. Herein, we compare these approaches, DFT calculations and data analysis techniques, and discuss their ability to provide meaningful predictions of catalyst performance. Examples of applications are selected to demonstrate the advantages and limitations of both tools. Several ongoing challenges in the predictions of reaction outcomes are also highlighted. Multivariate linear regression methods have become useful predictive tools that can complement potentially computationally expensive and complex transition state calculations. This Review compares these methods, highlights the advantages of each and identifies challenges for the future.