AI-Designed Serine Hydrolases Achieve High Catalytic Efficiency

A major breakthrough in enzyme design has been achieved using AI-driven computational modeling to create highly efficient serine hydrolases. Researchers combined RFdiffusion, a generative AI tool, with an active site preorganization assessment method to design enzymes from minimal active site descriptions. The resulting enzymes exhibited catalytic efficiencies (kcat/Km) up to 2.2 × 10⁵ M⁻¹ s⁻¹, with crystal structures closely matching design models (Cα RMSDs < 1 Å). Notably, the study identified five distinct protein folds that differ from natural serine hydrolases, demonstrating the potential to expand the structural diversity of engineered enzymes. This de novo enzyme design approach not only provides insights into the geometric principles of catalysis but also establishes a roadmap for designing custom biocatalysts capable of multistep transformations, opening new frontiers in synthetic biology and industrial biotechnology.