Designed by Nature. Engineered by Evozyne.
The Evozyne platform combines the exquisite complexity of Nature with machine learning to create proteins with extraordinary impact.
Step 1: Learn from Nature
We begin our process by applying proprietary statistical methods on the entire evolutionary history of a given protein to discover Nature’s design rules for its structure and function.
Step 1: Learn from Nature
Protein sequences contain the information for specifying their essential properties – folding, function, and capacity to adapt. Understanding that information has been one of the grand challenges in molecular biology, and is now enabled by vast amounts of sequence data. We take a statistical, rather than a direct physical, approach to learn the design rules for proteins from their deep evolutionary history.
Step 2: Design synthetic proteins
Using statistical and machine learning algorithms, we compress the vast library of sequence data to a simpler, searchable design space to create novel synthetic proteins with desired functional properties.
Step 2: Design synthetic proteins
Our algorithms produce generative models relating sequence to function. This enables design without direct use of structure-based information and access a space of solutions vastly larger than possible through traditional directed evolution. The models also reveal adaptive regions of a protein that code for existing function, creating an opportunity for enhanced or novel functions.
Step 3: Build prototypes
We make synthetic proteins, at a scale of thousands per iteration, using custom gene synthesis and expression technologies.
Step 3: Build prototypes
Using large-scale, parallel gene assembly techniques, we efficiently make libraries of synthetic genes that code for designed proteins. Proteins are then expressed in various host strains or cell-free systems, customized as necessary for each target. We accomplish these steps through extensive miniaturization, process optimization, and automation.
Step 4: Test for performance
Using novel technologies, we make custom assays to experimentally test large libraries of designed proteins against target specifications.
Step 4: Test for performance
We develop proprietary functional assays to make multidimensional measurements of protein activities. The strategies include in vivo assays in various host strains and in vitro plate-based or microfluidic assays, all capable of a throughput of thousands to millions of measurements per design cycle. The measurements offer a quantitative evaluation of the performance of the proteins against multiple desired target specifications.
Step 5: Iterate and redesign
Experimental data from each design-build-test cycle are used to retrain computational models within a virtuous feedback loop, advancing designed proteins towards the desired target goals.
Step 5: Iterate and redesign
In a process that hints at natural evolution, the measurements from each design round are used to update computational models and generate new sequences with improved performance parameters. Through iterative cycles of design-build-test, the models converge on a sequence space manifold that contains designs that meet desired performance objectives.
Step 6: Deliver optimized solutions
Our workflow ultimately yields libraries of novel protein sequences with optimal performance specifications.
Step 6: Deliver optimized solutions
Our platform yields a library of solutions on the order of thousands, all of which meet the desired design criteria: multi-dimensional performance enhancement, new protein functionality, and novel designs for new intellecutual property. Our solutions can readily be integrated into products and processes or be further optimized for downstream activities to make the solutions viable at scale.
Drive extraordinary impact
Our solutions are incorporated into real-world applications across industries to deliver meaningful value.