Machine-Learning-Aided Engineering Hemoglobin as Carbene Transferase for Catalyzing Enantioselective Olefin...

Area of Science:

• Biochemistry
• Computational Biology
• Protein Engineering

Background:

• Vitreoscilla hemoglobin (VHb) engineering for biocatalysis is challenging.
• Predicting enzyme enantioselectivity is crucial for catalyst design.

Purpose of the Study:

• To develop a machine-learning-aided strategy for engineering VHb as a carbene transferase.
• To predict and enhance the enantioselectivity of VHb for olefin cyclopropanation.

Main Methods:

• Utilized a Natural Language Processing (NLP) model to create an enzyme enantioselectivity score predictor (EESP).
• Employed molecular docking to identify critical amino acid sites and site-saturated mutagenesis for library construction.
• Screened 160,000 virtual mutants using EESP and validated top candidates experimentally.
• Performed molecular dynamics simulations to analyze protein-substrate interactions.

Main Results:

• Identified VHb-WK (Y29W/P54K) as the top-performing mutant with high diastereoselectivity and enantioselectivity.
• VHb-WK efficiently catalyzed olefin cyclopropanation in aqueous conditions.
• Molecular dynamics revealed that specific residue interactions (R47, Q53, K84) in VHb-WK enhance enantioselectivity by restricting substrate access.

Conclusions:

• The integrated NLP model and enzyme modification strategy significantly reduces costs and workload in protein engineering.
• Engineered VHb, specifically VHb-WK, shows promise as a highly enantioselective carbene transferase.
• This approach offers a powerful platform for designing novel biocatalysts.