What is antibody developability, and how do you fix it before it costs you a year?

A molecule that does not bind fails in a week. A molecule that binds beautifully and aggregates at formulation fails after a year of investment. That asymmetry is the whole argument: the value of a developability assessment is not in confirming the good candidates, it is in removing the bad ones early, before they have consumed a campaign.

The effect is measurable on molecules that already reached the clinic. Across 239 clinical antibodies, those whose CDR face carries the least surface charge turn out to have fourteen times fewer developability flags than those carrying the most: 2 molecules out of 47 in the lightest fifth, against 28 out of 47 in the heaviest. This is a property of the published set, read backwards, not a prediction we are making about your molecule. It is why we look at charge distribution on the structure before anything else.

Most profiling is done on sequence alone, because sequence is what everyone has. Sequence tells you about obvious motifs. It does not tell you which hydrophobic patch is actually exposed, which charge pair sits on the interface, or how the loops pack together. Those are structural questions, and they are where the late failures hide.

We build the structure, then measure on it: exposed hydrophobicity, charge distribution and patch geometry, the surface available for self-association, the conformational stability of the loops. Each candidate gets a profile across the properties that matter, and a ranked list of the specific positions that carry the risk.

That list is the useful part. "This molecule is risky" is not actionable. "These four residues carry the exposed hydrophobic patch, and here are the substitutions that remove it without touching the binding site" is.

Properties trade against each other. Push thermostability and you often pay in expression; remove an aggregation-prone patch and you can lose affinity. A provider who reports that every property improved at once has either not measured carefully or not pushed hard.

We state the trade explicitly: which property we chose to protect, which one we accepted to spend, and why. The molecule you receive is a documented arbitration, not a black box.

1. Assessment: We profile your molecule from its structure and return the ranked liabilities. Free.
2. Optimisation: Variants designed to remove the liabilities under an explicit constraint on binding.
3. Measurement: Best candidates expressed and compared with the parent at the bench.

We say what cannot work, early. Our profiling is built to eliminate, not to reassure. That is an uncomfortable deliverable and a cheap one, compared with the alternative.

We profile from structure, not only sequence. The liabilities that survive sequence filters are exactly the ones that fail late.

We design, so we can fix what we find. An assessment that ends in a report leaves you with a problem. We hand back the re-engineered molecule and the numbers that show it worked.

You own everything. Sequences, data and intellectual property are yours.