AI-guided enzyme optimization using cell-free and plant expression systems
An integrated design-build-test-learn platform that accelerates recombinant protein engineering through AI-designed mutation libraries, automated cell-free expression, high-throughput LC-MS screening, and plant-based validation. The platform enables rapid, parallel evaluation of hundreds to thousands of enzyme variants and iterative optimization for applications such as herbicide tolerance.
Overview
This integrated design-build-test-learn platform accelerates recombinant protein engineering by combining AI-driven mutation design, automated cell-free expression, and high-throughput mass spectrometry analysis. It addresses the bottleneck of moving from designed protein sequences to screened, validated candidates by enabling hundreds to thousands of sequence variants to be evaluated in parallel—without the time and cost of conventional cell-based cloning. The workflow is particularly suited for engineering enzymes with improved activity, stability, or novel functions, such as herbicide tolerance.
Technical specifications
The platform integrates computational and experimental modules:
- Design: Structure-guided deep learning (e.g., ProteinMPNN), protein language models (e.g., ESM-2), and evolutionary analyses (e.g., EVmutation) identify mutation hotspots and generate focused, sequence-defined variant libraries of ~1,000 single amino acid mutations.
- Build: Low-cost gene synthesis and automated liquid handling produce variant libraries for expression in microbial or plant-derived cell-free systems.
- Test: Soluble expression and enzyme function are screened by LC-MS to quantify substrate-to-product conversion, providing direct functional readouts.
- Learn: Protein embeddings, zero-shot fitness predictions, and regression models identify sequence–function relationships and predict improved combinatorial variants for subsequent design rounds.
- Validate: Optimized sequences are transferred into plant-compatible vectors and validated by transient expression in Nicotiana benthamiana, followed by functional assays such as herbicide challenge.
The platform is demonstrated with two herbicide-tolerance enzymes: EPSPS (target-site resistance) and glyphosate N-acetyltransferase (metabolic detoxification).
Technology readiness level
The platform is at an early stage of development. The proposed demonstration project will validate the full design–build–test–learn cycle for EPSPS and GAT, including library generation, cell-free screening, AI-guided iterative optimization, and transient plant expression. Successful completion will move the platform from concept to validated prototype ready for further scaling and broader enzyme engineering applications.
About University of Wisconsin, Madison
The University of Wisconsin–Madison is the state’s flagship public research university, a comprehensive institution with a large research enterprise and broad disciplinary breadth. Industry engages through shared instrumentation, pilot‑scale testbeds, and co-located core facilities that support prototyping and scale-up. An affiliated research and technology park hosts startups and corporate R&D, while integration with a major hospital system enables clinical translation; a statewide extension network connects campus advances to companies and communities across Wisconsin. Research is sustained by competitive federal funding from agencies such as NIH, NSF, DOE, USDA, and DoD, alongside state and industry partnerships. Dedicated commercialization support—through an affiliated foundation and campus offices—provides IP management, licensing, startup mentoring, and flexible, industry-friendly agreements.
delivered to your inbox.