NMD-VCell Research Workbench Module: Research Workbench · evidence-to-experiment workflow NMD = neuromuscular disorders
Search Release v1.0 candidate
v1.0 candidate Frozen 25 Jul 2026 DOI pending
v1.0.0-database-resource Schema 1.1 Model ridge-safe-v2.3 Benchmark repeated-fold-v2.2 Build EA-20260726-13 v1.0 is a database and evidence-governance release; it is not a validated disease-prediction or clinical decision-support release.
Current evidence ceiling L2 observed HepG2 limited L3a context No independent DMD perturbation validation Open boundary

NMD-VCell Research Workbench

From evidence to a registered next experiment.

Explore one gene, compare a candidate set, create an action-specific Study Card, and preserve the result as a versioned research object.

Explore a gene Open a worked comparison
Current evidence ceiling L2 observed HepG2 limited L3a context P1/P2 locked Claim boundary

Results command panel

Four-click verification console

A compact reviewer surface for the release identity, evidence ceiling, benchmark gate and machine-readable proof routes. It summarizes what is available now and what remains locked.

Open guided tour
Releasev1.0.0-database-resource Evidence freeze2026-07-25 Schema1.1 Benchmark strict gate0/16 pass

Claim ceiling

Where the evidence stops

L2 HepG2 observations21/21
L3a muscle-context screens9/21
L3b independent DMD replications0/21
L4/L5 therapeutic/clinical validation0/21

The panel is descriptive: no composite score, winner, clinical action or P1/P2 disease-prediction unlock is generated.

Candidate evidence matrix preview

A heatmap-style audit of all 21 governed hypotheses

Each cell is a rendered frozen field, not a new score: observed HepG2 support, muscle-expression context, source agreement, rescue sign span, strict benchmark gate and missing independent DMD/muscle perturbation replication.

Open full candidate dashboard
7Computational replication7Risk/context review4Verify expression2Muscle-context assay1Provisional hold · external report not yet audited
available/supporting partial/context conflicted below expression threshold missing/locked
Current display order is action then gene; no numeric current rank or winner is generated.
Gene Current action L2 obs. Muscle expr. Source agree Rescue span Strict gate L3b rep.
CPEB1 Verify expression yeslow0.670.01…0.120/160/21
DNAAF3 Verify expression yeslow1-0.04…0.060/160/21
EPS8L1 Verify expression yeslow1-0…0.110/160/21
ZFP69B Verify expression yeslow0.67-0.05…0.060/160/21
INTS13 Risk/context review yesTPM>11-0.03…0.060/160/21
NAGLU Risk/context review yesTPM>10.75-0.04…0.050/160/21
RAC3 Risk/context review yeslow0.750.01…0.090/160/21
RNASEH2C Risk/context review yesTPM>11-0.05…0.040/160/21
RNF8 Risk/context review yesTPM>10.5-0.03…0.040/160/21
WDR4 Risk/context review yesTPM>11-0.02…0.060/160/21
ZNF236 Risk/context review yesTPM>10.5-0.07…0.090/160/21
MON1A Muscle-context assay yesTPM>10.5-0.07…0.050/160/21
ZNF133 Muscle-context assay yesTPM>10.75-0.04…0.070/160/21
ADAM10 Computational replication yesTPM>10.5-0.01…0.080/160/21
CALR Computational replication yesTPM>10.75-0.01…0.070/160/21
DDX19B Computational replication yesTPM>10.5-0.05…0.020/160/21
DNM1 Computational replication yesTPM>10.5-0.07…0.020/160/21
EHMT2 Computational replication yesTPM>10.75-0.03…0.030/160/21
LMO2 Computational replication yesTPM>10.75-0.08…0.020/160/21
MPHOSPH6 Computational replication yesTPM>11-0.04…0.080/160/21
GFOD2 Provisional hold · external report not yet audited yesTPM>10.5-0.05…0.080/160/21
Compare any listUse the same mapped evidence fields without ranking. Worked comparisonZNF133 · MON1A · GFOD2, frozen and checksummed. Prediction gateConfirm why P1/P2 claims remain locked.

Bioinformatics result snapshots

Results you can read before opening the ledgers

One data-driven SVG turns the frozen release objects into a visual summary: object scale, candidate action queue, evidence-depth funnel, DMD source-agreement strip and benchmark gates. It is descriptive, not a ranking or prediction claim.

Download result SVG
Bioinformatics result snapshot figure summarising NMD-VCell object scale, candidate action queue, evidence depth, DMD source agreement and benchmark gates.
Static SVG generated from release manifests, candidate records and strict outcome objects. Gene symbols, API fields and numeric gates remain unchanged; no composite score or target ranking is generated.
Open WorkbenchInspect readiness and action queue Open 21 hypothesesFilter current actions and gaps Open BenchmarkAudit strict 0/16 gates

Reviewer guided tour

Search gene → Compare candidates → Build Study Card → Inspect visual card → Validate/register

This pipeline-style tour tells a first-time reviewer what to open, what visual result to expect and which boundary must remain intact.

Do not overclaimNo ranking, no therapeutic efficacy, no clinical decision support and no P1/P2 disease-prediction unlock.
1 · Search gene Start with one typed gene record Gene visual profile A single gene view reorganizes frozen fields only; it does not create a new rank. 2 · Compare candidates Contrast a small candidate set Comparison visual dashboard Input order is preserved; no winner, score, cluster or therapeutic priority is computed. 3 · Build Study Card Generate the next-experiment route Planner visual builder The Planner is a pipeline-style route, not an intervention recommender. 4 · Inspect visual card Read the stable Study Card object Study Card visual summary Cards remain DRAFT until required fields are registered and checksummed. 5 · Validate/register Confirm gates before any claim expansion Validation and registry gates P1/P2 prediction claims remain locked; formal v1.0 still requires external gates.

Three core research tasks

Start with the decision a researcher is trying to make

Explore evidence

Audit one gene

Search stable gene records, inspect source state, evidence level, DMD context and decision-blocking gaps.

 Compare candidates

Open a real comparison

Contrast ZNF133, MON1A and GFOD2 without generating a winner, score or therapeutic ranking.

 Design the next experiment

Review a Study Card

Move bounded evidence into estimand, controls, QC, replication and stop-rule fields.

Secondary actions

Validate and registerInspect readiness gates, lifecycle states and prospective outcome scaffolds. Reproduce dataOpen checksums, manifests, SQLite identity and rebuild objects without browsing the full ledger.

Governance

Publish packageCurrent release identity, checksums and DOI-pending boundary. AuthorizeExternal DOI, licence and maintainer gates remain explicit. Release gateInternal gates pass; formal v1.0 remains locked pending external inputs.

Evidence-to-experiment funnel

Evidence-to-experiment ladder

17,921gene recordsSearch and audit bounded evidence 2,160benchmark perturbationsSame-HepG2 model assessment 21governed hypothesesStudy Card / action queue 9external-context screensL3a context, not replication 0independent DMD replicationsRequired before P1 claims

External myoblast context is L3a screening, not same-perturbation replication. The 21 Study Cards remain drafts until their registered fields are complete.

Worked examples

Three real objects to try first

Gene object

ZNF133 evidence record

Inspect identifiers, source agreement, DMD context and bounded next action.

 Comparison object

ZNF133 · MON1A · GFOD2

Review an immutable comparison with mapping provenance and no generated ranking.

 Study Card

ZNF133 muscle-assay card

Use a governed design scaffold for preregistration-oriented discussion.

Immutable research objects

Evidence that can move without losing its boundary

Gene, Comparison, Study Card, Prediction, Outcome and Release objects carry stable IDs, schema and release identity. Registered revisions are never overwritten.

Open the lifecycle registry · Read the API 1.3 manifest

Evidence Atlas

A governed module inside the Workbench

The Atlas remains the evidence-audit layer. Compare, Design and Validate turn that bounded evidence into traceable downstream objects without upgrading the scientific claim.

Explore DMD evidence · Read the claim boundary

Current model result

Small same-HepG2 error reductions; no robust raw-direction claim

The safe ridge model modestly reduces prediction error inside the processed HepG2 dataset. Disease-context validity remains unestablished.

Inspect the benchmark

Why this remains beta

Independent DMD/muscle perturbation replication is absent. DOI deposition, approved licences and maintainer metadata also remain external gates.

Inspect every release gate · Inspect prediction readiness