PCA‐Based Database Mining Enables the Discovery of Bacterial Carbene Transferases for Stereodivergent Cyclopropanation

Abstract

ABSTRACT Protein engineering is a practical approach to providing enzymes with an “abiotic” catalytic activity. However, it remains difficult to explore the full diversity of natural sequence space through the engineering of a single specific protein. As an alternative to these protein engineering approaches, we here demonstrate a database mining approach using a principal component analysis (PCA)‐based clustering method to facilitate the identification of promising enzyme candidates. As a proof of concept, we applied this method to the cyclopropanation of styrene, and the sequence space of bacterial globins in the database was extensively investigated. By screening 275 globins from 171 different organisms, we successfully discovered enzymes capable of catalyzing stereodivergent carbene transfer reactions. Furthermore, statistical analyses of sequence data allowed us to detect characteristic structural properties of these globins, which determine the unique stereoselectivity of cyclopropanation. While these bioinformatics tools have primarily been applied to predict enzymes’ natural biological functions, this study demonstrates their applicability to exploring enzyme candidates for abiotic reactions unrelated to their native biological activity. Given the increasing interest in biocatalytic applications beyond natural reactivity, this PCA‐based mining approach provides a promising direction for expanding the functional diversity of biocatalysts.