Sensai gives labs execution truth at the run level. Odyssei extends the same measurement philosophy to molecules - mapping the interactions that actually decide outcomes, before programs commit to them.
Why this matters
01
AI can generate candidates faster than labs can verify them.
Candidate generation is no longer the bottleneck. Measurement is. Teams now have more designs than trustworthy evidence to rank them - and the gap is widening.
02
One-assay-at-a-time workflows delay real answers by weeks or months.
By the time one hypothesis resolves, the program has already moved. Decisions are made on partial signal, then corrected later at far higher cost.
03
Programs do not fail because the obvious variable was missed.
They fail because competitive, context-dependent interactions were never measured early enough. When they surface late, teams retrace months of work to understand what the system was doing all along.
The thesis
Most high-value biology is interaction-driven, not endpoint-driven. If those interactions stay unmeasured, teams optimize to proxies and discover real behavior too late. One multiplexed run returning a ranked interaction map changes the loop entirely.
Current loop vs. target loop
What happens today
What Odyssei aims to deliver
What gets measured
| Interaction signal | Why it matters | Current coverage |
|---|---|---|
| Binding strength | Determines whether a candidate engages the target at all | Partial |
| Competitive binding | Reveals off-target displacement that single-assay workflows miss | Missed |
| Selectivity profile | Separates specificity signal from promiscuous binding | Missed |
| Context response | How activity shifts under physiological conditions | Missed |
| Stability under assay | Distinguishes real signal from degradation artifact | Partial |
| Confidence score | Ranked evidence for program decisions | Missed |
How Odyssei works
01
Multiplex design
One experiment, thousands of conditions in parallel
Instead of running assays sequentially, Odyssei encodes each interaction pair with a barcode for sequencing. A single run interrogates an entire interaction matrix - candidates against targets, under multiple conditions simultaneously.
No expensive robotics or massive infrastructure required.
02
Data integrity
Quality control is built into the measurement workflow
Sequencing output is processed with automated QC, alignment, and reproducibility checks before interaction maps are ranked. The goal is to surface unreliable signal early, so teams do not make decisions on drift or assay artifacts.
03
Ranked output
A complete interaction map, not a single endpoint
The result is a versioned, reproducible interaction map ranked by evidence strength. Teams get the full picture of what the system is doing, early enough to act on it. Not a figure to interpret, but a structured output that feeds directly into program decisions.
Modular and extensible to immune mapping, neuroreceptors, and pathogen interactions.
We are working with teams where missing interaction signal is already a decision bottleneck. If that describes your workflow, we should talk.
Contact Us
