De Novo Design of Miniature and Efficient Metallo-Ketoreductases

The rational design of enzymes capable of catalyzing abiotic transformations represents a frontier in synthetic biocatalysis. While computational strategies have successfully produced enzymes for acid-base chemistry, examples of de novo designed oxidoreductases are still rare. Here, we report a deep learning-guided workflow for designing metallo-ketoreductases from theoretical active sites, enabling asymmetric reduction of ketones via an abiotic hydride-transfer mechanism. The resulting miniature enzyme contains only 130 residues, while exhibiting high catalytic performance under whole-cell conditions, achieving k_cat/k_uncat up to 1.4 × 10⁶, turnover numbers (TON) up to 19,000, enantiomeric excess (e.e.) values of up to 98%, broad substrate scope, and regioselective reduction of diketones. Notably, the designed scaffold shows exceptional thermal stability toward 90 °C treatment, outperforming natural promiscuous reductases, and exhibits tolerance to various organic solvents. This work demonstrates the power of de novo enzyme design to access non-natural catalytic functions, offering a scalable and sustainable route to engineer tailor-made biocatalysts for asymmetric synthesis.