Evidra · Validate
Every prediction, matched to what really happened
Evidra keeps an immutable record that links each simulated prediction to the clinical or real-world outcome it was meant to forecast, so accuracy is something you can check.
ev_20a1 · ORR at 12 weeks, 40 mg
- Predicted
- 31% (27–35%)
- Observed
- 29% (phase 2a, n = 64)
- Within interval
- Yes
- Calibration, this area
- 18 of 21 within 90% interval
- Recorded
- 2026-10-14 · signed
Capabilities
What Evidra does
Ledger
Prediction and outcome, linked
Each prediction is stored with its inputs and later joined to the matching outcome.
Calibration
Accuracy you can inspect
Calibration and error are tracked by therapeutic area, endpoint and model version.
Outcome graph
Drug, mechanism, subgroup, result
A knowledge graph on cuGraph links outcomes to the runs that predicted them.
Packages
Evidence packages
Export a structured package with lineage for MIDD documentation.
API
Governed data access
Approved partners query evidence and virtual-patient data through scoped APIs.
Feedback
Recalibration signals
Mismatches feed back to Cohorta, Insilon and Vigila.
How it works
Inside Evidra
Record
Store the prediction and its inputs.
Match
Join the realised outcome when it arrives.
Score
Update calibration by area and endpoint.
Feed back
Send recalibration to the models.
Specification
Specification
Items marked planned are in design with pilot partners.
| Item | Detail |
|---|---|
| Records | Prediction, inputs, versions, approvals, outcome, comparison |
| Integrity | Append-only store with signed entries |
| Compliance posture | Designed to support 21 CFR Part 11 and EU Annex 11 controls; not yet assessed |
| Exports | PDF and JSON evidence packages |
| API | REST with scoped tokens; partner access planned |
| Status | In development |
Place in the loop
Evidra in the Sanenza loop
- 01PopulateCohorta
Generates calibrated virtual-patient cohorts for an indication, region or rare subgroup.
- 02SimulateInsilon
Runs the compound across the whole cohort with PK-PD surrogates and biomolecular response.
- 03DesignProtova
Searches dose, cohort, endpoint and sample size, and scores every candidate protocol.
- 04Stress-testVigila
Oversamples rare subgroups to find adverse-event signals before a patient is exposed.
- 05LocalisePhysova
Renders organ-scale twins for tissue exposure and drug-device questions.
- 06ValidateEvidra
Links each prediction to the real outcome and sends recalibration back to every product.
Questions
About Evidra
Why does a ledger matter?
Simulation is only useful if you know how often it was right. The ledger makes that track record visible instead of claimed.
Can a record be edited?
No. Corrections are new entries that reference the original, so the history stays intact.
Who can see our records?
Only your organisation, unless you approve sharing with a partner or regulator.
See Evidra in a briefing
We walk your team through Evidra on a public example, then discuss where a pilot could test it against outcomes you already know.