De novo enzyme design

Function is conserved.Structure is free.

Evolution is a greedy, path-dependent search over a vanishing fraction of functional protein space. Zymics designs the part it never reached — enzymes that perform natural chemistry through structures natural selection could never have stumbled into.

objective-driven generative designclosed-loop design–build–test–learnbeachhead: cellulosomesde novo folds, natural function

The thesis

Nature optimized for survival, not for the task.

Natural enzymes are local optima. They were shaped by constraints that have nothing to do with the chemistry they perform — codon usage, folding kinetics, ancestral scaffolds, neutral drift. We decouple function from evolutionary history and ask a different question: what is the global optimum for this reaction if the catalytic machinery could be redesigned from scratch?

01

Evolution hill-climbs without foresight

Selection can only take mutational steps that are viable right now. Whole fold classes and active-site geometries that solve a chemistry better were never reachable, because no continuous viable path led there.

02

We do objective-driven generation

We generate backbones with generative models conditioned on a designed objective — binding geometry to the substrate, transition-state stabilization, processivity — and can cross fitness valleys that evolution cannot.

03

Function, not lineage, is the target

Fitness here is the task itself, not reproductive success. That reframing is the whole company: we treat four billion years of evolution as a baseline to beat, not a library to copy.

The search

Evolution settles on the nearest peak. We cross the valley.

Natural selection takes only viable steps, so it converges on a local optimum and stops. Objective-driven generation isn't bound to a continuous viable path — it can descend through a fitness valley to reach a higher structure evolution could never have arrived at.

Natural evolution
hill-climbs to the nearest local optimum
Guided design
crosses the valley to the global optimum

The engine

A design–build–test–learn loop for molecular machines

The generative layers are increasingly commodity. Our moat is above them: a function-conditioned objective layer that scores catalytic competence, and a closed wet-lab loop whose proprietary assay data retrains the models. Models commoditize; functional data does not.

Generative denoising over backbone spaceσ → 0

We generate structure, not noise about it.

Backbones are sampled by diffusion and flow-matching models conditioned on a functional objective — a directed search over a learned manifold of designable structures, not an enumeration of sequence space. Inverse-folding then assigns sequences that fold to the design and present its catalytic geometry.

The generative layers are increasingly commodity. What compounds is everything wrapped around them — the objective that defines a good catalyst, and the wet lab that tells us when we're right.

  1. 01

    Objective

    Encode the task as a scorable target: substrate-binding geometry, transition-state preorganization, processivity, foldability and expressibility.

  2. 02

    Backbone generation

    Diffusion / flow-matching over backbone space, conditioned on the objective — searching a learned manifold of designable structures, not enumerating sequence space.

  3. 03

    Sequence design

    Inverse-folding models assign sequences that fold to the generated backbone and present the catalytic geometry.

  4. 04

    Structure & physics

    Structure prediction filters for fold confidence; MD / FEP / QM-MM stress-test the active site. In-silico scores are proxies — never the final word.

  5. 05

    Build & assay

    Express the survivors and measure real turnover in the wet lab. The assay is the ground truth the models are missing.

  6. 06

    Learn

    Functional data flows back into the objective layer and the generative models. The loop is the asset.

↺ feedbackWet-lab turnover retrains the objective layer and the generative models. The loop is the asset.
Moat 01

Function-conditioned objective layer for catalytic competence

Moat 02

Proprietary closed-loop wet-lab assay data

Moat 03

Substrate-targeting for hard, insoluble substrates

The beachhead

Cellulosomes: a sharp wedge for a general engine

Cellulosomes are the bacterial machinery for breaking down crystalline cellulose. They are the right first target — and a demo, not the destination.

01

A market that needs better catalysts

Biofuels, textiles, paper and waste valorization all hinge on the cost of turning cellulose into sugar. Better enzymes move the economics directly.

02

A naturally modular architecture

Cohesin–dockerin scaffolds snap catalytic modules together combinatorially — an architecture practically built for redesign.

03

A clean, quantifiable readout

Cellulose → glucose turnover is directly measurable. A hard benchmark de-risks the platform claim with a number, not a narrative.

Type I cohesin–dockerin complexPDB 1OHZA real cellulosomal building block, shown for reference. Our de novo designs will replace illustrative structures here as they clear the wet lab.

Cellulosomes are the demonstration. The asset is a general design-build-test-learn engine for natural function through unnatural structure — next targets include plastic-degrading enzymes, carbon-fixation machinery, and novel biosynthetic catalysts.

How we hold ourselves honest

We publish the method before the marketing

De novo enzymes have historically been slow catalysts; the field's job is climbing rate enhancement by orders of magnitude. We don't claim that's solved. We claim a loop that climbs it — and we report what the loop actually measures.

In-silico scores are proxies

MD, FEP and QM-MM approximate catalysis; they don't equal k_cat. Every computational claim is paired with wet-lab turnover before we stand behind it.

We track the designability gap

Many designs don't fold or express. We measure fold and expression yield as first-class objectives, not afterthoughts.

Benchmarks, not adjectives

Cellulose-to-glucose turnover is our public yardstick. Numbers replace this scaffold as designs clear assay — we won't pre-announce them.

Roadmap

  1. Now

    Engine + beachhead

    Closed-loop pipeline running against cellulosome redesign.

  2. Next

    First validated designs

    De novo catalytic modules with measured turnover vs. natural baselines.

  3. Later

    Platform expansion

    Generalize the loop to new natural functions and harder substrates.

Designing enzymes evolution forgot to invent.

For updates on the platform and our first validated designs, leave your email. For partnerships, research collaboration or investment, get in touch directly.

Product updates

Follow the platform

Occasional updates on the engine and our first validated designs. No noise.

Get in touch

Partnerships, research & investment