Machine learning-guided dual optimization of the substrate specificity and thermostability of yeast alcohol dehydrogenase

Abstract Background Yeast alcohol dehydrogenase I (YADH) catalyzes the reversible reduction of aldehydes to primary alcohols. Thermostable YADH variants with broad substrate specificities at elevated temperatures are desirable for industrial biocatalysis. In this study, we aimed to alter the substrate specificity of YADH to efficiently catalyze the reduction of the non-native substrate, trans -2-nonenal, to trans -2-nonenol, while simultaneously enhancing thermostability. Residues around the substrate-binding site were identified by docking simulation and subjected to saturation and random mutagenesis. Measured activity and thermostability were used to train a machine-learning model to predict YADH variants with improved activity and thermostability. Results The machine learning-guided approach identified variants with improved properties. Four variants showed increased specific activity toward trans -2-nonenal compared with the wild type, and one variant also exhibited improved thermostability. The best-performing variant showed a more than twofold increase in specific activity along with increased thermal resistance. Variants that lost activity toward the native substrate acetaldehyde while gaining higher activity toward trans -2-nonenal were also obtained, indicating altered substrate specificity. These results demonstrate efficient identification of beneficial mutations from a limited dataset. Conclusions Overall, this study presents a machine learning–guided framework that integrates docking and multi-model evaluation for simultaneously optimizing substrate specificity and thermostability in YADH.