Functional hotspots identification via a hybrid NMR-computational approach facilitating directed evolution of large enzyme

Directed evolution has revolutionized enzyme engineering but remains challenging for large, multidomain proteins due to their complex dynamics and vast combinatorial sequence space. Traditional NMR spectroscopy, while powerful for identifying functional residues, is limited by the requirement for complete resonance assignments that become intractable in high molecular weight systems. Here, we present a hybrid NMR-computational framework that integrates chemical shift perturbation mapping with AlphaFold3-docked distance constraints, enabling the identification and assignment of ligand-proximal hotspots without the need for full resonance assignment. By analyzing side chain ¹ H- ¹³ C correlation spectra, probabilistic amino acid typing, and structural proximity filtering, we mapped functional hotspots in the ∼90 kDa Pyrococcus furiosus DNA polymerase. Targeted saturation mutagenesis of these sites resulted in variants exhibiting multi-fold increases in catalytic efficiency while sampling less than 0.6% of total residues. The identified hotspots revealed spatially distinct yet cooperative contributions to catalysis, as further supported by combinatorial mutagenesis and kinetic analyses. This hybrid strategy provides a practical pipeline for narrowing mutagenesis targets in large enzymes under catalytic conditions, overcoming the size, assignment, and screening limitations of conventional NMR-guided evolution, offering a powerful route toward efficient biocatalyst optimization by prioritizing high-probability target residues.