MutexaGPT: an intuition-to-design translator for physics-based enzyme engineering.

MutexaGPT enables translation of enzyme engineering intuition into physics-based models, producing actionable design proposals for two protein engineering tasks.

• Why it matters: Systematic methods for converting qualitative enzyme design intuition into quantitative principles are lacking, limiting the efficiency of enzyme engineering efforts.
• What they did: MutexaGPT is an open-access, multi-agent large language model platform that takes plain-English requests, constructs physics-based models, runs high-throughput molecular workflows, and generates mutation libraries, demonstrated in two protein engineering cases.
• The result: MutexaGPT successfully facilitated enzyme modifications for bulkier substrates and improved activity at lower temperatures, establishing it as a tool that integrates human creativity with molecular modeling to democratize enzyme design.