Most drug discovery in our space asks you to compromise early. Simplify the molecule, narrow the design space, commit to a lead before the data is in. We run it the other way around: widen the design space, build the real molecules, test them in their final format, and let biological function decide which one moves forward.
Our platforms and processes exist to make that possible. An antibody design that lets you keep your binders and a real IgG, and a process built to test many candidates at once instead of betting on a few.
Solving the problem
A monoclonal antibody has two identical arms, so its heavy and light chains pair correctly on their own. A multispecific has different arms, so the chains can pair the wrong way, and every arm you add multiplies the number of potential problem combinations. Get it wrong and you make a mess of mispaired, non-functional molecules. With trispecifics, the complexity increases sharply.
Most platforms get around this by giving up the natural antibody shape (fusing fragments or building unnatural scaffolds) or forcing every arm onto a shared light chain. Both work, and both cost you something. The first gives up a real antibody’s half-life and manufacturability. The second limits can limit which binders you are allowed to use.
We put the engineering where it doesn’t impact the biology. The changes happen in the constant domains, the structural scaffold, so the right chains pair on their own, maintaining an IgG-like structure. The binding regions — the pieces that recognize your target — are left alone.
We will use the B-Body platform as an example. A similar approach applies when using the T-Body platform, allowing for a direct funtional comparison
. Our unique design enables three benefits. You can use the binders you want. You keep a real antibody with its half-life, manufacturability, and familiar CMC. We took what we learned from our bispecific engineering approach and enhanced it, allowing direct comparison between B-Body bispecifics and T-Body trispecifics.
The B-Body platform, with the four engineering changes numbered
T-Body expands on the proven architectural principles used in our B-Body bispecific platform by enabling a third specific Fab
You bring the targets
We run the full workflow, starting from library screening. This is the path if you don't have binders yet, or if you want the design space explored properly rather than working from the few antibodies already available.
You bring the sequences
If you already have antibodies, your sequences go straight into the matrix. Because we don’t require a common light chain, the antibody you bring is the antibody we build with.
The green track is monoclonal discovery and characterization. The blue track is multispecific assembly and screening. Partner-supplied antibody panels enter at the marked point.
If you are starting with targets rather than sequences you have already identified. We start our antibody discovery engine. This is the front half of the workflow, and it’s the same whether you end up with a monoclonal, a bispecific, or a trispecific.
Before anything gets screened, the target must exist in a form worth screening against: cell lines or purified protein presenting the antigen as it appears in the body. Where you have already done this work while evaluating the target, we build on it.
We screen synthetic phage libraries rather than immunizing an animal. We use more than 30 proprietary libraries, each holding over a billion primary clones. The part that matters is the parallelism. We keep libraries separated by physical characteristics rather than pooling them and then querying each against several screening conditions at once, because it’s not known in advance which condition best represents what happens in a patient. We run more than 90 unique screens at once, delivering multiple shots on goal in a single pass rather than one hypothesis at a time.
Material produced by phage often moves into a yeast display module, which quantifies individual cells and provides additional dimensions for separating antibodies. This is how we steer selections toward a particular epitope using a blocking strategy, set gates that pass only the tightest binders, or pan against actual cells, so what survives is enriched for functional cell binding.
Every output from every selection goes into next-gen sequencing, telling us which sequences came from which condition in which well. That turns a dataset of material into a map. In the map, we select several hundred sequences and express them as antibodies from mammalian culture.
We don’t just evaluate whether they bind. We look at whether they bind to anything they should not, whether their surface properties suit manufacturing, whether they assemble correctly, whether they bind the analogous target in your safety toxicology species, and whether the panel covers different epitopes, established by binning rather than just assumed. At this point, a monoclonal program delivers antibodies, while a multispecific program moves to the next stage.
You now have characterized antibodies against each target. The conventional move is to pick a promising pair and hope it behaves. The trouble is that pairing is not additive: two antibodies that each look excellent alone can produce a molecule that expresses poorly, aggregates, or loses the activity you wanted.
So instead of choosing, we build all combinations. A typical bispecific campaign generates 288+ options, using a 12x12 matrix in two orientations, hundreds of individual molecules, each assembled in its final format. A single anti-CH1 capture purifies each of them, which makes running one method across hundreds of different molecules practical.
The same scaffold can build several shapes. Whether each arm binds once or twice changes how the molecule behaves, and the right choice depends on your biology. Target density, the therapeutic window, and manufacturing all drive format selection. Because the platform tests formats in parallel in their final configuration, you compare them on real data instead of guessing.
Key feature Monovalent binding
Key feature Bivalent
Key feature Three targets, one IgG
This is the front half of the workflow, and it’s the same whether you end up with a monoclonal, a bispecific, or a trispecific.
Stage 4
The matrix output undergoes the same characterization as the monoclonals did, now as multispecifics. Developability, binding, and function across hundreds to thousands of candidates, in plate-based and individual assays.
Stage 5
Function comes first. We start with the functional data, select the top performers, and then cross-reference them against developability. This is a working session, not a blind handoff. You see the full dataset, and we review it together because the trade-offs at this stage are program decisions rather than technical ones.
Stage 4
Nobody goes from a one-milliliter expression to a lead. Selected candidates are scaled up to meet your downstream testing requirements.
At the end, you don’t get a sequence and a shrug. You get a novel multispecific molecule in its final format with a complete data package, ready for cell line development — so you skip format engineering and move toward manufacturing.
Human germlines, natural CDR diversity, minimal liabilities, characterized in the format you will actually develop.
You get all the developability data, not a curated summary
Your leads and your data are yours, provided under a research use license. Clinical / Commercial access is available.