Generative artificial intelligence for enzyme design and biocatalysis

Area of Science:

• Biochemistry
• Computational Biology
• Artificial Intelligence

Background:

• Enzyme design historically faced low success rates with computational methods.
• Recent innovations in generative artificial intelligence (AI) have dramatically advanced protein design.
• Generative AI models are now frequently employed for designing proficient enzymes.

Purpose of the Study:

• To provide a comprehensive overview and classification of generative AI models for enzyme design.
• To highlight experimentally validated models relevant to real-world applications.
• To outline the limitations of current generative AI models in enzyme design.

Main Methods:

• Review and classification of existing generative AI models for enzyme design.
• Analysis of experimental validation data for AI-designed enzymes.
• Assessment of model limitations and potential for industrial applications.

Main Results:

• Generative AI models have overcome previous limitations, achieving frequent success in enzyme design.
• Several AI models show promise for real-world applications, with experimental validation.
• Key limitations of current models are identified, paving the way for future improvements.

Conclusions:

• Generative AI models have reached maturity for creating and optimizing enzymes for industrial use.
• Wider adoption of AI models with experimental feedback loops will accelerate biocatalyst development.
• Community assessment of AI models will guide the next generation of enzyme design tools.