The development of predictive tools to assess enzyme mutant performance and physical organic approaches to enzyme mechanistic interrogation are crucial to the field of biocatalysis. While many indispensable tools exist to address qualitative aspects of biocatalytic reaction design, they often require extensive experimental data sets or a priori knowledge of reaction mechanism. However, quantitative prediction of enzyme performance is lacking. Herein, we present a workflow that merges both computational and experimental data to produce statistical models that predict the performance of new substrates and enzyme mutants while also providing insight into reaction mechanism. As a validating case study, this platform was applied to investigate a non-native enantioselective photoenzymatic radical cyclization. Statistical models enabled interrogation of the reaction mechanism, and the predictive capabilities of these same models led to the quantitative prediction of the enantioselectivities of new substrates with several enzyme mutants. This platform was constructed for application to any biocatalytic system wherein mechanistic interrogation, prediction of reaction performance with new substrates, or quantitative performance of enzyme mutants would be desirable. Overall, this proof of concept study provides a new tool to complement existing protein engineering and reaction design strategies.
experimental and computational information
computational files and code