AI-guided enzyme design, built for small, high-information experiments.
We design 24–48 variant enzyme campaigns that test clear hypotheses, compress experimental search, and help teams learn faster from every assay.
Our work starts with focused enzyme rescue for companies building high-value molecules — but our deeper mission is broader: to build the computational engine for protein invention.
Computational design. Compact libraries. Real experimental feedback. Ownable enzyme IP.
Designed for information gain, not just predicted winners.
24–48 Variant Campaigns
Most enzyme engineering still depends on large experimental screens or slow manual intuition. We take a different approach.
We build compact variant libraries designed to answer specific questions:
Can this enzyme become more active?
Does stability need to be improved?
Is expression the real bottleneck?
Can we improve selectivity without losing catalytic efficiency?
Can we expand substrate scope or reduce unwanted inhibition?
Each campaign is built around a hypothesis, not a random search.
Fewer variants, sharper hypotheses, faster learning.
What We Do
We rescue bottleneck enzymes.
Many teams already have enzymes that almost work. They may show activity in a clean assay but fail under process conditions. They may express poorly, lose stability, make side products, stall at high substrate concentration, or create pathway bottlenecks.
We help teams diagnose the likely failure mode and design a small, ranked variant set for testing.
Our campaigns can target: activity, stability, soluble expression, selectivity, substrate scope, product inhibition, solvent tolerance, temperature tolerance, pH tolerance, and process robustness.
We are a computational protein design company.
Computation is not a support layer for us. It is the heart of the company.
We combine protein structure, evolutionary context, substrate information, process constraints, and prior assay data to design enzyme variants with high expected learning value.
Every campaign is treated as a model-building opportunity.
Each campaign turns sparse experimental data into better priors for the next round of enzyme design.
Our Research Thesis
Enzyme engineering should become more predictable.
The future of enzyme design will not be built by brute-force screening alone.
It will come from better representations of how mutations change function under real constraints: activity, folding, expression, dynamics, selectivity, stability, inhibition, and process compatibility.
We are building systems that learn from small experiments and turn low-data enzyme engineering into a repeatable discipline.
How it works
-
We start with the enzyme, target reaction, assay data, and process constraints.
-
We identify likely causes of failure: activity, folding, stability, inhibition, expression, selectivity, or condition mismatch.
-
We generate a compact, ranked library designed to test the most important hypotheses.
-
The results tell us what mattered, what failed, and where to search next.
Better enzyme design starts with better questions.