Can you fix a clone that binds beautifully but aggregates?
The problem this solves
Most discovery campaigns end with a clone that is excellent at the one thing it was selected for. Developability is everything else: whether it stays monomeric at 100 mg/mL, whether it can be manufactured at a sane titre, whether it carries deamidation or oxidation hot spots that will drift during storage, and whether it sticks to things it should not stick to.
These properties are cheap to fix at the lead-engineering stage and ruinously expensive to discover in late preclinical. The awkward part is that they trade against each other, and against affinity. Optimising them one at a time by rational mutation tends to produce a molecule that is better on one axis and worse on two others.
How we do it
We treat developability as a selection problem rather than a design problem. Using STEM™ we expand clone diversity around your parent, then select variants with good biophysical behaviour under conditions that expose the liability you care about — so improvements on several axes are found together instead of being traded off sequentially.
What that looks like in practice, from our published anti-TNFα work: the engineered antibody reached 2.8× the potency of infliximab; its self-association, measured by AC-SINS, moved from 19.33 to −1.67; and transient expression rose from about 50 to 70–80 µg/mL. Potency, stickiness and manufacturability all moved the right way in the same campaign.
Those are the figures from the public version of that work. We quote published numbers on the website and reserve internal or unpublished data for scoped conversations under CDA — if a number appears here you can check it against the source.
Liabilities we work on
- Aggregation and self-association — measured, not predicted; AC-SINS and SEC as the readouts
- High-concentration viscosity — the property that decides whether subcutaneous dosing is possible
- Sequence liabilities — deamidation, isomerisation, oxidation, unpaired cysteines, glycosylation sites in CDRs
- Polyreactivity — non-specific binding that shows up later as fast clearance
- Expression titre — a clone that cannot be made is not a lead, whatever its KD
- Thermostability — often improved alongside the above rather than as a separate campaign
- Engineered variants with sequences, ranked on the axes you specified
- Before/after biophysical data on the same assay, so the improvement is comparable rather than asserted
- Affinity for every variant, so you can see whether binding was preserved
- Expression titre for the panel
When to run this
The best moment is immediately after lead selection and before any tox or CMC spend. The second-best moment is when a specific liability has already blocked you — a viscosity ceiling, an aggregation result that will not pass, a titre too low to be economic. In that case bring us the data you already have; the campaign is usually narrower and faster when the target is one named property.
Tell us what your lead is missing
Send the sequence and the property that is blocking you. A PhD scientist reads it and replies within two business days. No NDA needed to start.
Talk to a scientist →Frequently asked questions
Will fixing developability cost me affinity?
Not necessarily. Because we generate diversity and select rather than mutate one residue at a time, variants that improve biophysics while retaining or improving affinity are found within the same campaign. Every variant we return comes with its affinity measured, so the trade is visible rather than hidden.
Can you work from a humanized clone, or does it have to be the parent?
Either. Developability engineering is routinely run after humanization, and it can also be combined with humanization in one continuous pipeline rather than as a second project.
Do you run the biophysical assays or do we?
We run the assays that drive selection and report them. If your organisation has a specific in-house panel that a molecule must pass, tell us at scoping and we will match the readouts so the numbers are comparable to yours.
What if the liability turns out not to be fixable?
Then we say so. Some liabilities are inseparable from the paratope, and in those cases the honest recommendation is to go back to the panel rather than keep engineering one clone. We would rather tell you that early than bill for a campaign that cannot succeed.
Who owns the engineered sequences?
You do, with no downstream royalties in standard scope.