Chemists at Vanderbilt University have built an AI tool that turns a researcher’s plain-English hunch into a workable plan for redesigning an enzyme. The tool, called MutexaGPT, takes intuition-driven requests and converts them into design proposals grounded in chemistry and physics, according to the team behind it.
Enzymes drive countless chemical reactions in daily life. Engineered versions help synthesise new medications, preserve nutrient-rich foods, and recycle polymers to cut plastic waste. Designing one usually means long rounds of trial and error, since even a single amino acid change can shift how an enzyme behaves.
MutexaGPT works by asking follow-up questions to fill gaps in a request, then building physics-based models and returning concrete mutation proposals. Assistant professor of chemistry Zhongyue Yang led the work with former doctoral student Qianzhen Shao, now a postdoc at Stanford. The team says the aim is to elevate what scientists do, not replace them.
One use the researchers point to is cold-active enzyme technology: engineering enzymes that keep working well at lower temperatures. The study appears in Nature Computational Science. The name Mutexa is short for “Alexa for Mutant”, a nod to the original idea of a voice-style assistant for designing enzyme mutants.
Source: Phys.org
