How do I run an antibody discovery campaign without an animal and without a library?
We design the antibody instead of screening for it. You give us the antigen and the site you want to engage; we generate candidates against it, narrow them computationally, pre-validate the shortlist dynamically, then express and test the best ones. IgG, Fab or scFv, whichever your programme needs. Candidate sequences in six weeks, validated binders in nine.
- 6weekscandidate sequences
- 9weeksvalidated binders
What changes when the campaign is designed
A screening campaign gives you what binds. A designed campaign gives you what you asked for: the epitope, the format, the species cross-reactivity, the distance from a competitor's sequences. And when a design fails, we know which assumption failed, so the next round is informed rather than re-drawn.
Three ways in
From the target alone. Nothing in the pipeline sees a prior antibody. The right route when no antibody exists against your site, or when you want a design record independent of existing art.
Around an antibody you already know. The binding mode is proven, so the work is moving the sequence while keeping the interaction. The fastest route, and the one behind our work on patented epitopes.
From your own candidates. You bring sequences; we engineer affinity, expression, stability or specificity, and measure the result.
We tell you which route fits your target before you commit.
What you set, and selection cannot give you
- the epitope: a functional site, a conserved site, or one that avoids a competitor's claims
- the format: VHH, scFv, Fab, IgG, bispecific, minibinder
- developability: constraints applied while designing, not assessed afterwards
- species cross-reactivity: whether the binder also recognises the mouse or macaque orthologue, which decides whether your toxicology study is straightforward
- distance from prior art: identity to published and claimed sequences, computed per candidate
Species cross-reactivity is worth a programme on its own. If your binder does not recognise the cynomolgus orthologue, the preclinical route gets longer and more expensive. We flag it in week two, not month eighteen.
Formats
- IgG: effector function, long half-life, a conventional therapeutic route
- Fab: imaging, structural work, no effector function wanted
- scFv: CAR-T and other constructs where a single chain is required
- VHH / nanobody: difficult epitopes, small size, cheap expression. See nanobody discovery
- miniprotein binder: research reagents, very small scaffolds. See miniprotein binders
We will tell you which one serves your application before you commit, and we design to it rather than converting afterwards.
How it works
- Feasibility (days): We score your antigen and the site you want to engage, and return a compatibility verdict. Free.
- Prototyping (week 1): A short scoping run on your actual target: we design the workflow it needs and size the generative campaign, so the quote reflects your target rather than an average one.
- Generative design (weeks 2-4): Binders generated against the chosen site on a clinically precedented scaffold, under developability constraints
- Computational triage (weeks 4-5): Filters on sequence liabilities, structure and independent re-prediction narrow the output to 50 to 200 candidates worth looking at closely
- Molecular dynamics (weeks 5-6): The shortlist is pre-validated dynamically: we watch the complex hold, or fail to hold, before anything is expressed
- Wet-lab validation (weeks 7-9): Best candidates expressed and tested: binding, expression, behaviour in solution
You approve each stage before the next one starts.
What you receive
Ranked candidate sequences with the epitope each engages; a developability profile per candidate; identity to the relevant prior art; binding data from the bench; and the full design record. The sequences and the resulting intellectual property are yours.
Why clients come to us
Speed with a defined target. Six weeks to sequences against the site you named.
Control of the epitope. Functional site, conserved site, or a site that avoids existing claims.
Developability by design. Candidates are shaped inside the physicochemical space where clinical antibodies sit, which is narrower than the natural repertoires an immunisation campaign samples. Fewer liabilities to engineer out later, and a de-risked candidate at the point where it costs least to change.
One team from design to bench. The people who generate the candidates are the people who test them, so what the assay says goes back into the next design round.
Built for therapeutic programmes. We design therapeutic antibodies as well as research and diagnostic reagents, and the design record is written to follow the candidate into preclinical development.
Frequently asked questions
Which antibody discovery CRO is fastest from target to candidate?
Timelines depend on the target, not on the provider's brochure. Ours are six weeks to ranked sequences and nine to validated binders, and the free feasibility study tells you in days whether your particular site is tractable, before anyone commits to a schedule.
How do I choose an antibody discovery partner?
Ask four things: can you choose the epitope, or do you take what binds? Which numbers in the report are measured and which are predicted? Who owns the sequences? And what happens, contractually, if the campaign returns nothing. We answer those in the first call.
Can you work from a sequence alone, with no structure?
Yes. A structure helps and we will often predict one, but the antigen sequence is enough to start.
Do you do conventional antibodies, or only nanobodies?
Both. Single-domain antibodies are our native format, and conventional antibodies are ordinary work.
Who owns the sequences?
You do, along with the data and any intellectual property arising from the work.
What are the leading platforms for de novo antibody design?
They divide into companies that design against a target, companies that optimise antibodies you already have, and tool providers whose models you run yourself. We do the first two and validate in the laboratory. The question that separates providers quickly: which of your numbers are measured, and which are predicted?
Can AI design antibodies from scratch?
Yes, against a site you choose, and the field has moved from plausible designs to binders confirmed at the bench. We validate every candidate we deliver, which is how prediction becomes a molecule.
Which tools do you use?
Public models where they are the best available, our own where they are not, and an independent re-prediction step so no single model decides.
How is this different from phage display?
Display selects from diversity you generate; design builds against the site you choose. See nanobody discovery.
How long does it take?
Six weeks to ranked candidate sequences, nine weeks to validated binders.
Send us your antigen and the site you want to engage. The feasibility study is free and takes ten minutes.