Data-Driven Protein Engineering for Improving Catalytic Activity and Selectivity.

Area of Science:

• Biocatalysis
• Protein Engineering
• Computational Biology

Background:

• Traditional protein engineering methods like directed evolution and rational design face challenges in screening vast protein mutation spaces.
• Optimizing enzyme substrate scope, catalytic activity, and selectivity is crucial for biocatalysis applications.
• Machine learning (ML) offers a powerful alternative to approximate protein fitness landscapes and identify catalytic patterns.

Purpose of the Study:

• To review machine learning models for assessing enzyme-substrate-catalysis relationships.
• To highlight ML's role in guiding data-driven protein engineering campaigns.
• To prospect future developments in ML for enzyme optimization.

Main Methods:

• Review of existing machine learning models applied to enzyme engineering.
• Analysis of ML's capability in predicting enzyme performance based on limited experimental data.
• Exploration of ML for understanding enzyme-substrate-catalysis interactions.

Main Results:

• Machine learning models can effectively approximate protein fitness landscapes.
• ML facilitates the identification of catalytic patterns, guiding protein engineering efforts.
• Data-driven approaches using ML can improve enzyme activity and selectivity.

Conclusions:

• Machine learning provides a novel and efficient strategy for protein engineering.
• ML-guided enzyme optimization is essential for advancing biocatalysis.
• Future developments in ML will offer enhanced tools for designing enzymes with desired properties.