Interaction matrix
Every candidate by every target, ranked by relative binding strength.
Platform
One experiment. Every candidate. A real number for each pair.
SPR, ITC, ELISA, and Y2H were built for careful measurement of one interaction at a time. Modern design pipelines were not. Lagomics pools every candidate against every target in one reaction and uses sequencing to read out which pairs actually bound.
Mechanism
1
Barcode every candidate
Cell-free expression. Unique DNA tag per sequence.
2
Pool in one reaction
All candidates and targets together. No per-pair setup.
3
Enrich binders
Interacting pairs linked and kept. Non-binders washed away.
4
Sequence the pool
One Illumina run reads which barcodes survived. Pooled, barcoded, sequencing-based.
What you get back
Off-target profile
Which candidates bind targets outside the intended set.
Epitope binning
Which candidates compete for the same binding site.
Structured data
Raw sequencing and matrices for teams, models, or downstream tools.
Comparison
| Traditional (SPR / ITC / ELISA / Y2H) | Lagomics (pooled, sequencing-based) | |
|---|---|---|
| Candidates per run | 1 | Thousands |
| Cost driver | Scales linearly with candidate count | Scales with sequencing depth, not candidate count |
| Off-target data | Rarely collected | Generated automatically |
| Turnaround for a full library | Months | Weeks |
| Input required | Purified protein per pair | A list of sequences |
Building toward AI-native validation
Lagomics is building validation infrastructure for a world where candidates come from models and agents, not just human intuition. Pooled, barcoded, sequencing-based readouts return structured interaction matrices designed to feed the next design cycle, the next model training run, and eventually autonomous workflows.
We would love to hear from teams working on this problem. A conversation is the best way to explore whether we can help.
FAQ
Questions teams ask us
Straight answers about the approach, the deliverable, and how it compares to what exists today.
What would working together look like?
Usually a short conversation to understand the candidate library, targets, and what decision the interaction data needs to support. From there we can explore whether a collaboration or letter of intent makes sense. No pressure on a first call.
How many candidates could you screen in one run?
The design targets pools from dozens to several thousand candidates, depending on the target and how much resolution is needed. Scope gets sized once we understand the library.
Do candidates need to be synthesized first?
No. The workflow is built for cell-free expression directly from sequences. No physical peptide or protein synthesis required upfront.
What would Lagomics need to get started?
Candidate sequences, target(s), library size, and context on what decision the interaction matrix would inform. That helps us understand whether the approach fits the problem.
What comes back from an experiment?
A quantitative interaction matrix: every candidate by every target, ranked by relative binding strength, plus off-target profiling and raw sequencing data that can be reprocessed or fed into models.
How does this compare to SPR, ITC, or ELISA?
Those methods give a precise number for one pair at a time. Lagomics is built for ranked, relative binding signal across an entire library in a single run. Breadth and speed first; top hits can still be confirmed with SPR or ITC afterward.
How does this compare to yeast two-hybrid or other library screens?
Y2H can surface binders at scale, but in a yeast context, often as binary hit/no-hit, without quantitative ranking for a specific designed library against specific targets. Our readout is built for a structured interaction matrix on the candidates and targets that matter.
What if none of the candidates bind?
That is still useful data. Negative results across a library are hard to generate any other way, and many teams need them as much as hits to filter a design pipeline or retrain a model.
Questions about the approach?
Lagomics would love to hear about the validation problems worth solving and explore whether we can help.
Get in touch