Generative protein language models for designing substrate-promiscuous enzymes
Caltech researchers developed a generative AI platform using the GenSLM protein language model to design high-performance, substrate-promiscuous tryptophan synthase (TrpB) enzymes. This technology accelerates biocatalyst discovery by creating stable, active, and autonomous enzymes that outperform natural and laboratory-evolved variants, bypassing time-intensive traditional directed evolution methods for industrial chemical and pharmaceutical applications.
Overview
Researchers have leveraged the GenSLM protein language model to innovate the field of biocatalysis by generating novel variants of the tryptophan synthase beta-subunit (TrpB). Traditionally, identifying functional starting points for enzyme engineering is a major bottleneck that requires time-intensive laboratory directed evolution. This AI-driven approach bypasses such constraints by generating enzymes that are inherently stable, catalytically active, and capable of functioning autonomously without their natural partner subunit, TrpA. By accessing sequence space far beyond known natural enzymes, this platform provides a powerful solution for discovering high-performance biocatalysts with enhanced properties.
Technical specifications
- Employs the GenSLM protein language model for generative enzyme design.
- Produces novel TrpB variants that express efficiently in Escherichia coli.
- Enzymes operate independently, eliminating the requirement for a partner TrpA subunit.
- Demonstrates significant substrate promiscuity, accepting non-canonical substrates that are typically inaccessible to natural counterparts.
- Models have successfully created variants that match or outperform both natural enzymes and those optimized through traditional laboratory evolution.
- The design pipeline includes computational filtering to identify high-quality candidates prior to experimental validation.
Technology readiness level
The technology is at level 4. The research has successfully demonstrated the generation and experimental validation of stable, active TrpB enzymes in laboratory conditions (E. coli expression), with results published in a peer-reviewed journal. The research team has also made the code and datasets publicly available for further implementation and validation.
About California Institute of Technology
Caltech is a private, STEM-focused research university in Pasadena, notable for its small size, graduate intensity, and tightly integrated academic–research culture. Companies engage through shared user facilities for micro/nanofabrication and advanced instrumentation, prototyping shops, and contract mechanisms managed by experienced corporate‑partnership staff. An affiliated NASA laboratory, the Jet Propulsion Laboratory, and the surrounding Southern California aerospace and deep‑tech cluster provide unique pathways to mission‑driven engineering, rigorous testing, and specialized suppliers. Research here is backed by competitive federal funding from agencies such as NSF, NIH, DOE, NASA, and the Department of Defense. A dedicated technology transfer office streamlines IP protection, licensing, and startup formation, connecting innovators with investors and regional accelerators.
delivered to your inbox.