Computational design of serine hydrolases

Area of Science:

• Biochemistry
• Computational Biology
• Protein Engineering

Background:

• Designing enzymes for complex, multi-step reactions is a significant challenge in biochemistry.
• Serine hydrolases serve as a valuable model system for studying enzyme catalysis.

Purpose of the Study:

• To develop a de novo approach for designing enzymes with complex active sites capable of multistep reactions.
• To leverage artificial intelligence and computational methods for enzyme design.

Main Methods:

• Utilized RFdiffusion, a generative AI model, for designing enzyme active sites.
• Employed an ensemble generation method to assess active site preorganization throughout the reaction coordinate.
• Performed experimental characterization, including catalytic efficiency measurements and X-ray crystallography.

Main Results:

• Achieved catalytic efficiencies (kcat/Km) up to 2.2 × 10^5 M^-1 s^-1.
• Obtained crystal structures with high fidelity to the design models (Cα RMSD <1 Å).
• Identified novel catalysts with five distinct folds, different from natural serine hydrolases.

Conclusions:

• The de novo design approach successfully created functional enzymes for complex transformations.
• Demonstrated the importance of structural compatibility across the reaction coordinate for enzyme function.
• Provided insights into the geometric principles of catalysis and a framework for future enzyme design.