Tailoring industrial enzymes for thermostability and activity evolution by the machine learning-based iCASE strategy

Area of Science:

• Biochemistry
• Molecular Biology
• Enzyme Engineering

Background:

• Enzyme evolution faces a critical stability-activity trade-off, limiting the development of highly active and stable enzymes.
• Existing methods struggle to efficiently engineer enzymes with enhanced performance characteristics.

Purpose of the Study:

• To develop a novel strategy for enzyme engineering that overcomes the stability-activity trade-off.
• To establish a predictive model for enzyme function and fitness using machine learning.

Main Methods:

• Developed the isothermal compressibility-assisted dynamic squeezing index perturbation engineering (iCASE) strategy.
• Constructed hierarchical modular networks for enzymes of varying complexity.
• Utilized structure-based supervised machine learning to create a dynamic response predictive model.

Main Results:

• The iCASE strategy successfully constructed hierarchical modular networks for diverse enzymes.
• Molecular mechanism analysis revealed a structural response mechanism driving adaptive evolution.
• The predictive model demonstrated robust performance in predicting enzyme function, fitness, and epistasis across datasets.
• The universality of iCASE was validated across four enzyme types with different structures and catalytic functions.

Conclusions:

• The iCASE strategy offers a powerful approach to engineer enzymes with improved stability and activity.
• Machine learning-based prediction of enzyme fitness provides valuable guidance for future enzyme evolution studies.
• This work advances the understanding of enzyme adaptive evolution and engineering principles.