A Machine Learning-Guided\nApproach to Navigate the\nSubstrate Activity Scope of Galactose Oxidase: Application in the\nConversion of Pharmaceutically Relevant Bulky Secondary Alcohols

Biocatalysis is increasingly being adopted in industry\nfor producing\nimportant chemicals in a selective and efficient manner. Engineering\nan enzyme can often confer it with an altered chemical scope, making\nit accessible to nontraditional and desirable chemistry. Identifying\nenzymes with the desired substrate specificity and activity, however,\nremains time-consuming and costly. Galactose oxidase (GOase) is a\ncopper-dependent enzyme that converts alcohols to their corresponding\ncarbonyls, an important transformation in industrial synthesis. Here,\nwe present a machine learning aided protocol to develop a catalytic\nactivity prediction model (<i>R</i><sup>2</sup> ∼\n0.7–0.9) for GOase based on a focused data set of engineered\nGOase variants with activity toward bulky benzylic secondary alcohols.\nThe trained GOase activity prediction models (with no additional training)\nalso partially retained their predictive power when applied to another\nmember of the oxidase family, an aryl-alcohol oxidase. Inspired by\nthe fragment-based optimization methods used in drug discovery, we\ndeveloped an active-site structure-aware substrate library for select\nGOase variants. Experimental validation of a subset of the constructed\nsubstrate library against select variants indicates that the trained\nmodels provide reasonable prediction (<i>R</i><sup>2</sup> = 0.61) of GOase activity, enabling the identification of the best\nGOase variant from the select variant subset for each identified substrate.\nThis ability to identify optimal GOase variants from the selected\nvariants for the synthesis of industrially important chemicals was\ndemonstrated for dyclonine, an FDA-approved drug. Our machine learning-guided\napproach enables rapid navigation of the substrate-activity scope\nof GOase, thereby reducing the burden of extensive experimental screening\nand streamlining the deployment of biocatalysis in industrial synthesis.