Optimizing enzyme thermostability by combining multiple mutations using protein language model.

Area of Science:

• Protein Engineering
• Biotechnology
• Computational Biology

Background:

• Enzyme thermostability is crucial for industrial applications and protein science.
• Current methods for enhancing enzyme thermostability, like rational design and random mutagenesis, struggle with complex epistatic interactions in combinatorial mutants.
• Optimizing enzymes often requires lengthy, iterative design processes.

Purpose of the Study:

• To develop an AI-aided strategy for efficiently engineering enzyme thermostability by recombining beneficial single-point mutations.
• To address the challenge of epistasis in high-order combinatorial mutants.
• To create a faster and more effective framework for enzyme design.

Main Methods:

• Utilized creatinase thermostability data, including single to quadruple mutants.
• Employed a temperature-guided protein language model (Pro-PRIME) to learn epistatic features.
• Designed and generated combinatorial mutants using the AI model.

Main Results:

• Achieved 100% success rate in obtaining 50 combinatorial mutants with superior thermostability after two design rounds.
• Developed a mutant (13M4) with 13 mutations, near-wild-type activity, a 10.19°C increase in melting temperature, and a ~655-fold longer half-life at 58°C.
• Successfully modeled high-order epistasis, including sign and synergistic epistasis, and elucidated mechanisms using a dynamics cross-correlation matrix method.

Conclusions:

• The AI-aided strategy efficiently facilitates the recombination of beneficial mutations for enzyme thermostability.
• The Pro-PRIME model effectively captures complex epistatic effects in high-order mutants.
• This framework offers a powerful tool for protein-directed evolution and the design of thermostable enzymes.