Machine learning-aided engineering of a cytochrome P450 for optimal bioconversion of lignin fragments.

Using machine learning, molecular dynamics simulations, and density functional theory calculations we gain insight into the selectivity patterns of substrate activation by the cytochromes P450. In nature, the reactions catalyzed by the P450s lead to the biodegradation of xenobiotics, but recent work has shown that fungi utilize P450s for the activation of lignin fragments, such as monomer and dimer units. These fragments often are the building blocks of valuable materials, including drug molecules and fragrances, hence a highly selective biocatalyst that can produce these compounds in good yield with high selectivity would be an important step in biotechnology. In this work a detailed computational study is reported on two reaction channels of two P450 isozymes, namely the -deethylation of guaethol by CYP255A and the -demethylation aromatic hydroxylation of -anisic acid by CYP199A4. The studies show that the second-coordination sphere plays a major role in substrate binding and positioning, heme access, and in the selectivity patterns. Moreover, the local environment affects the kinetics of the reaction through lowering or raising barrier heights. Furthermore, we predict a site-selective mutation for highly specific reaction channels for CYP199A4.