Module significance audit · EA-20260806-50
Eight modules, one evidence-to-experiment loop.
Every entry point now declares the question it answers, its evidence unit, the object it produces, the claim it cannot support and the next legal handoff. The map also makes the DMD-only interactive surfaces explicit.
Research flow
A page is useful only when its output has a lawful next use.
01DiseasesTyped disease summary, dataset registry, role-typed candidate objects, strict evidence states and a disease-specific next experiment→02QuestionsAnswer state, available evidence layers, missing truth, claim boundary and the next discriminating experiment→03BrowseA context map or stable evidence record that shows what is observed, contextual and missing→04Gene compareA stable comparison object and rule-traced gap-closing study suggestion without a scalar winner→05DesignA downloadable draft design object with explicit registration gaps and links to its evidence inputs→06ModelsModel cards, a six-run execution ledger, benchmark contracts, negative results and explicit readiness gates→07RegistryLifecycle states for comparisons, Study Cards, protocols, ModelRuns, predictions and outcomes, including honest zero states
Data & APIVersioned schemas, provenance and checksums support every step.Machine readability does not expand the biological claim.
Eight semantic contracts
What each module means—and where interpretation must stop.
01four disease typed evidence available
What has been measured, induced, corrected, perturbed or remains missing in each disease?
- Evidence unit
- Dataset-specific biological unit—donor, biopsy, clone, well or donor-derived line; individual cells are not treated as independent biological replicates by default.
- Output
- Typed disease summary, dataset registry, role-typed candidate objects, strict evidence states and a disease-specific next experiment.
- Claim ceiling
- No pooled cross-disease score, disease winner, treatment ranking, calibrated patient simulation or clinical recommendation.
Next · Choose a disease route; use the DMD interactive workflow where supported or follow the disease-specific next experiment. →02dmd interactive gate other diseases evidence only
Can the current release answer this prespecified DMD cell-state perturbation question?
- Evidence unit
- One question object with disease, cell state, perturbation, time, endpoint and requested output fixed together.
- Output
- Answer state, available evidence layers, missing truth, claim boundary and the next discriminating experiment.
- Claim ceiling
- No calibrated DMD response, candidate pathway outcome, patient trajectory, efficacy estimate or clinical recommendation.
Next · Inspect the relevant context in Browse, then compare DMD candidate evidence or freeze a Study Card. →03dmd context map and source linked records available
Which gene, cell compartment, dataset and measured context are relevant to the question?
- Evidence unit
- A source-linked gene, dataset or compartment record with explicit provenance and biological/statistical unit.
- Output
- A context map or stable evidence record that shows what is observed, contextual and missing.
- Claim ceiling
- No cell-specific causal effect, cell–cell interaction estimate or perturbation-conditioned DMD fate is inferred from browsing.
Next · Open a gene record, compare DMD candidates, or return to a disease module for disease-level evidence. →04dmd evidence dimension comparison available
How do selected genes differ across evidence depth, source agreement, context and the experiment-blocking gap?
- Evidence unit
- One mapped gene record per identifier, preserving original input, mapping provenance and separate evidence dimensions.
- Output
- A stable comparison object and rule-traced gap-closing study suggestion without a scalar winner.
- Claim ceiling
- No target rank, therapeutic priority, pooled cross-disease score or silent conversion of missing evidence to zero.
Next · Send the selected evidence gap to Design and freeze the estimand, endpoint and stopping rule. →05twenty one dmd draft study cards available
What falsifiable experiment would close the named evidence gap and what result would change the decision?
- Evidence unit
- One versioned Study Card with estimand, endpoint, replication, QC, analysis and stopping rule.
- Output
- A downloadable draft design object with explicit registration gaps and links to its evidence inputs.
- Claim ceiling
- A draft is not a registered study, executed experiment, measured outcome, target validation or efficacy result.
Next · Inspect lifecycle state in Registry; registration and execution remain prospective. →06bounded modelrun ledger no calibrated disease model
What actually ran on a frozen task, against which baseline, with which split, metric, failure reason and permitted use?
- Evidence unit
- One ModelRun bound to a task, data revision, split, baseline, seed, metrics and gate decision.
- Output
- Model cards, a six-run execution ledger, benchmark contracts, negative results and explicit readiness gates.
- Claim ceiling
- Published external performance is not inherited; no calibrated DMD or multi-disease predictor is released.
Next · Carry only a frozen, gate-passing run into prospective registration and outcome comparison. →07read only lifecycle and evidence registry available
What is draft, frozen, registered, running, measured, released, superseded or intentionally absent?
- Evidence unit
- One immutable versioned research object; a missing record is never interpreted as a negative result.
- Output
- Lifecycle states for comparisons, Study Cards, protocols, ModelRuns, predictions and outcomes, including honest zero states.
- Claim ceiling
- The public registry is read-only; it does not accept uploads, create persistence or promote zero outcomes into validation.
Next · Contribute a traceable outcome or evidence object, then apply the formal transition rules. →08versioned machine readable objects available
Which exact bytes, schema, identifier, checksum and release boundary support a displayed result?
- Evidence unit
- One versioned file or API object with a declared schema, provenance path and checksum identity.
- Output
- JSON, TSV, schemas, manifests, checksums, release bundles and a portable RO-Crate contract.
- Claim ceiling
- Machine readability or file availability does not expand biological claims; hosting exclusions remain declared rather than silently omitted.
Next · Verify the object, then return to the module whose scientific question it supports. →Disease × module scope
Four diseases do not imply four identical product capabilities.
| Disease | Diseases | Questions | Browse | Gene compare | Design | Models | Registry | Data & API |
|---|
| DMD | full typed module | interactive six input evidence gate | dmd muscle context map and gene records | twenty one candidate evidence comparison | twenty one draft study cards | bounded assay and dmd run ledger no calibrated model | prospective lifecycle zero registered outcomes | v1 1 workbench and v2 disease objects |
|---|
| FSHD | full typed module with registered single cell extension | bounded evidence route no interactive predictor | disease datasets and single cell state localization | not supported no cross disease gene pooling | next experiment recommendations not frozen study cards | no calibrated disease model | released evidence objects no prospective outcomes | v2 typed disease objects |
|---|
| DM1 | full typed module with splicing specific correction | bounded evidence route no interactive predictor | disease datasets and bulk splicing evidence | not supported no cross disease gene pooling | next experiment recommendations not frozen study cards | no calibrated disease model | released evidence objects no prospective outcomes | v2 typed disease objects |
|---|
| SMA | full typed module with registered single cell extension | bounded evidence route no interactive predictor | disease datasets and line aware state localization | not supported no cross disease gene pooling | next experiment recommendations not frozen study cards | no calibrated disease model | released evidence objects no prospective outcomes | v2 typed disease objects |
|---|
Reading rule: “evidence route” means a typed, bounded disease record—not an interactive predictor, a formal Study Card or a calibrated model.
Audit conclusions
The scientific boundary now travels with the navigation.
- Disease comparison and gene comparison are separate operations.
- The DMD checker, cell map, candidate comparison and Study Cards are not silently generalized to FSHD, DM1 or SMA.
- Model architecture names never substitute for a frozen local ModelRun.
- Draft, missing and null remain distinct lifecycle states.