Work with the engine on your target
We take on a small number of collaborations each year. Each one starts with a target, a defined success criterion and a written agreement about who owns what, because ambiguity on any of those three is where collaborations fail.
Design campaign
One target, one campaign. We model the target, generate and score candidates, and deliver a ranked shortlist with the reasoning and scores behind each selection. Synthesis and assay can run with our collaborators or in your own labs.
Suits
A target you have struggled to raise a binder against, or a programme where you want a designed starting point rather than a selection campaign.
You receive
- Approved constraint specification
- Ranked candidate shortlist with per-candidate rationale
- Structural and developability scores
- Recommended assay plan and controls
Closed-loop collaboration
The full cycle, run repeatedly. Design, synthesis, binding characterization and post-assay analysis across multiple rounds, with the engine retrained on your campaign's data as it accumulates.
Suits
Programmes where a first shortlist is unlikely to be the answer and the value is in iteration speed and in what each round teaches.
You receive
- Everything in a design campaign, per round
- Kinetic and specificity data with full assay context
- Error analysis explaining why candidates failed
- Updated structure–activity hypotheses between rounds
Platform partnership
A multi-target, multi-year arrangement with agreed target reservation, dedicated capacity and the option to co-develop selected programmes. Structured with milestones rather than a single delivery.
Suits
Organizations adding a nucleic acid binder capability across a portfolio rather than testing it on one target.
You receive
- Reserved targets within agreed classes
- Dedicated design and analysis capacity
- Direct access to the scientific team
- Negotiated co-development and milestone terms
How an engagement actually runs
From first conversation to a result you can act on. Timelines depend on target tractability and assay availability, and we would rather set them with you than publish a number we cannot keep.
Technical call
You describe the target, what has already been tried and what a useful result would look like. We say plainly whether this is a good fit. Some targets are not.
Feasibility review
We assess structural data availability, assay tractability and counter-target risk, then write a scoped proposal with a success criterion defined before work starts.
Design cycles
Constraint specification approved, candidates generated, scored and shortlisted. Synthesis and measurement follow, then error analysis and the next cycle.
Handover or continuation
You receive the molecules, the data and the reasoning. Either the programme continues into optimization or we stop, having learned something specific.
The default split, stated up front
Partners own the sequences, molecules and campaign data generated against their targets. We retain the platform, the models, the methods and any general improvements to the engine. Target-specific inventions are the partner's; improvements to how the engine reasons are ours.
We put this in the agreement before work begins, including publication rights, confidentiality terms and what happens to data if the collaboration ends early. We are also willing to discuss other structures, including co-development with shared economics on selected programmes.
What we need from you
- Target identity, or enough detail to assess tractability under confidentiality
- Any structural data, published or internal, and its quality
- Prior binder attempts and why they fell short
- Counter-targets that must be avoided and the specificity margin required
- The assay you consider decisive, and who will run it
- The success criterion that would justify continuing
When we are the wrong choice
Saying this publicly saves everyone a quarter.
Intracellular targets without a delivery plan
We can design a binder for an intracellular protein. Getting it into the cytosol at a useful concentration is a different and largely unsolved problem, and affinity will not rescue it.
No discriminating assay
If candidates cannot be told apart experimentally, the loop has nothing to learn from and the engine degrades into expensive guessing.
A binder needed next month
Design cycles are faster than selection campaigns, not instant. Synthesis and characterization set the floor on how quickly anything real can come back.
Validation of a predetermined conclusion
If the programme requires the answer to be positive, our error analysis will be an inconvenience rather than an asset.
Tell us about your target
A thirty-minute technical call is usually enough to establish whether there is a programme here.