AI-accelerated reactant-state enzyme discovery reveals SABP2 as a broad-spectrum, high-yield ester synthetase for green industrial biocatalysis

The development of general and efficient biocatalysts for carboxylic ester synthesis in aqueous media is a persistent challenge. We report the computational identification of salicylic acid-binding protein 2 (SABP2) as a proficient ester synthetase through an AI-augmented reactant-state enzyme discovery and design theory (REDD) and three-dimensional (3D) QM/MM free-energy simulations. REDD combines AlphaFold3-predicted structures with physics-based interaction descriptors—total hydrogen-bond energy (tHBE) and nucleophilic attack distance (NAD)—to rank catalytic potential. Strong correlations between these descriptors and experimental hydrolysis rates (ρ = 0.78 for tHBE, −0.60 for NAD) validated the prediction. 3D free-energy simulations revealed that transesterification (ΔG ‡ = 20.5 kcal·mol⁻¹) is kinetically favored over hydrolysis (ΔG ‡ = 21.6 kcal·mol⁻¹), explaining the synthetic bias. In an aqueous buffer system, SABP2 catalyzed the transesterification of 52 alkyl esters with conversions reaching up to 84% at a substrate concentration of 20 mM, outperforming conventional commercial lipases. This work not only presents a scalable computational platform for enzyme discovery but also identifies SABP2 as a sustainable, water-compatible biocatalyst for esterification reactions.