What if human trials were predictable?
VERDICT forecasts how a treatment will work in people by simulating human biology, so you can decide with evidence before the trial exists.
From a trial description to a forecast, through human biology.
Every forecast runs through the same chain: the data we assembled, the generative models trained on it, and a trial model that reads their output alongside the registered design.
- 01
Describe your trial.
Start with a public trial ID, a protocol document, or define an asset, therapeutic area, patient population, comparator and primary outcome.
- 02
Generate tissue responses.
GEM-2 simulates how the treatment changes human biology, tissue by tissue.
- 03
Generate gene expression.
A whole-transcriptome view of the response, from virtual human models.
- 04
Predict trial outcomes.
VERDICT brings the biology together with the registered design, the drug’s targets and prior results in the disease to forecast each planned analysis.
Four forecasts for every planned analysis.
Each ships with the uncertainty it earned on held-out trials: a calibration band showing how often trials predicted alike succeeded, and prediction intervals labelled with the coverage they achieved.
- Significance
- Will the primary between-arm analysis be statistically significant?
- Direction
- If significant, which arm does it favor?
- Effect size
- How large is the posted ratio?
- Adverse events
- Incidence of 976 serious-event concepts in each arm.
Frontier trial prediction
35%
Relative increase in observed success rate.
Among trials VERDICT called successful, compared with all evaluated trials.
In retrospective testing on 1,717 held-out later trials.
Success means at least one significant primary endpoint, not approval.
All evaluated trials
52%
Trials VERDICT called successful
70%
What you can do with it.
Pharma, biotech and investors forecast trials. Frontier labs and virtual cell teams build on the biology underneath.
In the browser, or in your pipeline.
Run a trial in the VERDICT application, or call the same models through the API with a workspace key.