How do I humanise an antibody or a nanobody without losing affinity?
By treating it as a design problem rather than a sequence substitution. We move the framework towards human germlines while preserving the residues that hold the binding site in shape, deliver a ranked panel rather than a single variant, and measure affinity, expression and stability against the parent molecule. You receive humanised candidates with the numbers that prove they still work.
Where humanisation goes wrong, and what we do about it
A naive graft moves the loops onto a human framework and loses affinity, because a handful of framework residues carry the geometry of the binding site. Recovering that affinity afterwards, one back-mutation at a time, is what turns a two-week job into a six-month one.
We identify those positions first, numbering the molecule by IMGT and mapping the residues that support the loops and the interface, then we humanise around them. Each position kept non-human is a deliberate choice with a recorded reason, which is also what your regulatory file will ask for.
A panel, not a bet
Tolerance to any given substitution is specific to the clone: the same change that is neutral in one antibody abolishes binding in another. A single humanised sequence is therefore a guess. We generate and rank a panel, take the best forward, and let the measurements decide.
Nanobodies need a different hand
A VHH has no light chain, and four conserved framework-2 residues occupy the surface a light chain would cover. They hold solubility and binding together, and they are where most VHH humanisation campaigns quietly fail. We know which positions can move, which cannot, and how to buy humanness elsewhere in the sequence, including by designing the loops themselves, which is where the residual T-cell epitopes actually sit.
That last point is the capability most providers do not have: we design, so we can engineer the loops, not only clean up the framework.
What you receive
- Ranked humanised sequences, with identity to human germline genes.
- The measured comparison against the parent: affinity, expression, thermal stability, aggregation behaviour.
- The list of preserved positions, each with its rationale. That is the design record for your file.
- A documented immunogenicity risk assessment of the kind a regulatory submission expects.
- Full ownership of the sequences, the data and the resulting intellectual property.
How it works
- Assessment: We read the molecule: numbering, the positions that carry the binding site, the sequence liabilities. Free.
- Design: The framework moves towards human germlines under those constraints; a ranked panel is generated.
- Measurement: Best candidates expressed and compared with the parent: binding, expression, stability, aggregation.
Why clients come to us
We preserve what makes it bind. The positions that carry the binding geometry are identified before anything is mutated, not recovered afterwards.
We measure. Every delivered candidate carries before-and-after numbers from the bench.
We can engineer the loops, not just the framework. Design capability, not a sequence tool.
Developability by design. Humanised variants are kept inside the physicochemical space where clinical antibodies sit, so humanness is not bought at the cost of aggregation or expression.
Built for therapeutic programmes. Humanisation is a step on the way to a therapeutic antibody, and the documented record we deliver is written for the file your preclinical development will rest on.
VHH is our native format. Single-domain antibodies are what we build, so framework-2 is familiar ground rather than a surprise.
Frequently asked questions
Which companies offer AI-based antibody humanization?
Several, and most deliver a sequence. We deliver a ranked panel with measured affinity, expression and stability against the parent, and we can redesign the loops as well as the framework.
How do I humanise a nanobody?
Number the molecule, identify the framework residues supporting the loops and the four conserved framework-2 positions, humanise around them, and measure. Those four positions are where campaigns fail: they hold solubility and binding, and they do not move for free.
Does humanisation reduce affinity?
A naive graft often does. Designing around the positions that carry the binding geometry is how we keep it, and the measurement against the parent is how you know we did.
What is the difference between chimeric, humanised and fully human?
How much of the sequence is human: constant regions only, framework too, or the whole molecule. Which one your programme needs depends on the route to the clinic. Tell us and we will advise.
Can you humanise a molecule you did not design?
Yes. Send us the sequence of your antibody or nanobody; the assessment is free.
Send us your sequence. We will tell you which positions we would preserve, and what we would measure, at no cost.