A computationally designed panel of diverse and selective peroxygenases for terpene oxyfunctionalization

Enzyme engineering has a critical role in the transition to economical, low-energy and environmentally friendly chemical production. Current approaches relying on costly iterations of mutation and selection are limited to reactions with a straightforward experimental readout and struggle to address mutational epistasis. We focus on unspecific peroxygenases (UPOs), prized engineering targets due to their ability to oxyfunctionalize diverse organic molecules of industrial and environmental value. To address the lack of scalable screening for UPO functions, we applied enzyme-design calculations to focus experiments. Starting from an AlphaFold2 model of Mth UPO, the automated FuncLib algorithm generated 50 diverse active-site multipoint designs—all of which were functional. Screening against nine diverse terpenes revealed large improvements and new oxyfunctionalization products, resulting in molecules of high pharmaceutical and industrial value. We rationalized observed specificity changes using AI-based docking and molecular dynamics simulations, providing molecular insights that could generalize to engineering other UPOs. Thus, computational design and modeling can dramatically accelerate the urgently needed green transition of the chemical industry.