Vanderbilt's MutexaGPT turns a chemist's plain English hunch into real enzyme designs
Enzymes quietly power medicine making, food preservation and plastic recycling, but tuning them usually means expensive trial and error. As Vanderbilt University reports, chemist Zhongyue (John) Yang and his team built MutexaGPT, described in Nature Computational Science, to change that. A scientist types an idea in plain English, and large language model agents ask for missing details, build physics based models and return concrete mutation proposals. One use is cold active enzymes, which could help make lactose free milk or detergents that clean in cold water. Yang says the goal is to amplify human intuition rather than replace researchers.