MutexaGPT converts plain-English enzyme requests into physics-based mutation designs

MutexaGPT accepts a scientist’s plain-English, intuition-driven request about an enzyme change, uses large language model agents to elicit missing information, builds physics-based models, and converts the results into actionable mutation design proposals—an arrangement meant to raise the quality of enzyme engineering work rather than remove people from it. Assistant professor of chemistry Zhongyue (John) Yang developed the system at Vanderbilt University with former doctoral student Qianzhen Shao, now a postdoc at Stanford, and other team members. A Vanderbilt University report carried by Phys.org on October 31, 2026, ties the tool to a paper in Nature Computational Science.

Yang has said he shares broader concerns that artificial intelligence could replace humans in the workforce. In that account he framed MutexaGPT as a way to translate imaginative human intuition into a workable plan.

Enzymes are specialized proteins that speed up chemical reactions and can be viewed as nature’s catalysts. Engineered enzymes synthesize new medications, preserve nutrient-rich foods, and help recycle polymers to reduce plastic waste. Each enzyme consists of hundreds of amino acids, and changing even a single amino acid—a mutation—can produce a substantial change in function.

Yang described the major challenge as the lack of a reliable way to predict mutations that alter genetic code without heavy trial and error in industry. In the same October 31, 2026, Vanderbilt University account carried by Phys.org, he estimated that the enzyme engineering market yields hundreds of billions of dollars in revenue each year, and that the trial-and-error pattern brings large volumes of expensive testing and resource costs. MutexaGPT is intended to let researchers explore, predict, and design enzyme mutations through natural language while connecting those questions to computational tools grounded in chemistry and physics.

Among the uses named are cold-active enzymes that keep high catalytic competence at lower temperatures, which matters for cold-temperature processes such as lactose-free milk and food preservation that retains heat-sensitive nutrients. Another practical application is laundry detergents that break down complex stains without hot water. The report also notes designs with greater catalytic efficiency and selectivity as a route to pharmaceutical manufacturing that uses less energy and fewer harsh reaction conditions.

The Vanderbilt University account carried by Phys.org on October 31, 2026, supplies the market figure and the application list for this report.