candidate panel
Select and justify the intervention panel; model-watch results cannot choose it.
Ready for selection and freezeDMD Prospective Truth Pilot
Generate the DMD perturbation truth needed for future model calibration.
Can a prespecified intervention produce a reproducible molecular and functional response in disease-relevant DMD myogenic cells from independent biological donors?
Track A · Truth generation. A frozen prediction is not required for this study.
Registration bridge · v08.8
The design contract now has an accountable execution path.
The model pair anchors generalization, Stage A differentiations provide nested support, and the evidence-linked review collects the variance, effect and named approvals needed to determine Stage B.
Scientific jobGenerate new truth
Study designReady for review
RegistrationAwaiting decisions
Target outputMatched DMD truth dataset
Standardized operating characteristics · v08.7
Two independent pairs remain a calibration pilot.
The reference 1-SD concordant-call probability is 9.3% with two pairs and 43.3% with six. Within-pair guides and differentiations do not increase the number of disease backgrounds; 39 formal-lock inputs remain pending.
Design recommendation · not frozen
A minimum reviewable design—not fabricated assignments.
The numbers below are starting requirements for experimental review. Candidate, modality, reagent, dose, time and endpoint choices remain null.
- independent dmd donors minimum
- 3
- independent reference donors minimum
- 2
- intervention panel size range
- 6–12
- negative interventions minimum
- 2
- positive or assay controls minimum
- 1
- required response layers
- transcriptomic response · cell-state distribution · disease-relevant functional endpoint · viability/toxicity
- inference unit
- independent biological donor or independently randomized culture; never individual cells
- holdout
- complete donor and future experimental batch
- note
- These are design recommendations for review, not assigned samples or an approved protocol.
Registration blockers
Ten decisions require author or experimental authority.
Every field remains explicit so a computational system cannot silently choose the biological experiment.
intervention modality
Freeze CRISPRi, ASO, small molecule or another modality before registration.
Ready for selection and freezereagent identity
Record reagent sequence/identity, target, lot and off-target controls.
Ready for selection and freezedose
Freeze dose levels and the rule for excluding toxic exposure.
Ready for selection and freezepost intervention time
Freeze molecular and functional readout times.
Ready for selection and freezedonor inclusion
Assign independent DMD and reference biological donors with inclusion/exclusion rules.
Ready for selection and freezemolecular primary endpoint
Choose one prespecified molecular endpoint and its statistical unit.
Ready for selection and freezefunctional primary endpoint
Choose one disease-relevant functional endpoint and assay.
Ready for selection and freezetoxicity endpoint
Define viability/toxicity failure and abstention thresholds.
Ready for selection and freezedecision rule
Freeze success, null, discordant and stop rules before outcome access.
Ready for selection and freezeAfter truth returns
Calibrate first; validate predictions in a separate study.
The qualifying outcomes become a versioned truth dataset, update the evidence record and support model training or calibration. Only an eligible later ModelRun may enter prediction validation.