Public data
Sparse
Too little interaction coverage
Most public interaction data is binary, inconsistent across sources, and covers only a fraction of the space models need.
Ground truth for agentic biology.
Agents design candidates. Models predict interactions. There is still no way to measure, at that same scale, which ones function in the real world. Lagomics runs one pooled, sequencing-based experiment that measures interactions across your entire candidate library at once: protein-protein, peptide-target, or binder-antigen. You get a quantitative interaction matrix back in weeks.
The bottleneck isn't design anymore. It's validation.
AI protein design and peptide discovery pipelines now produce candidates faster than any lab can test them with SPR, ITC, ELISA, or one-off pulldowns. Most teams validate the 5 to 20 candidates they have time for, not the ones most likely to work. Real hits get missed. Negative results, which are just as valuable for training your next model, never get generated at all.
Affinity
Rare
Quantitative data at scale
Dense binding-affinity matrices barely exist because traditional methods are too slow and expensive to generate them.
Negatives
Lost
The most informative misses
Teams test their best guesses one at a time. Non-binders and off-targets are filtered out before they can improve the next model.
What we're building
The measurement layer biology never had.
01
Candidate library in
Sequences and targets. The interaction question defined upfront.
02
One pooled experiment
Barcoded, cell-free, all pairs in one tube.
03
Interaction matrix out
Ranked binding, off-targets, structured data for teams and models.
Who this is for
You generate more candidates than one-at-a-time validation can keep up with.
AI-designed proteins and binders
Validate your model's output at the scale it is actually produced.
Peptide discovery
Screen full candidate libraries against your target in one experiment.
Interaction datasets for model training
Generate the dense, quantitative ground-truth data your model is starved for.
Activate Fellowship · MIT Engine · NVIDIA · Based at MBC Biolabs
Have a candidate library sitting in a spreadsheet?
Tell us what you're screening and against what. We'll tell you if a pilot makes sense within one call.