NMD-VCell Research Workbench Module: Validate / registry · evidence-to-experiment workflow NMD = neuromuscular disorders
Research evidence only 21 observed HepG2 perturbations 0 independent DMD replications Boundary & release
Observed HepG2 perturbations with limited external myogenic context No validated DMD perturbation prediction v1.0.0-database-resource Frozen 25 Jul 2026 Schema 1.1 Model ridge-safe-v2.3 Benchmark repeated-fold-v2.2 Build EA-20260729-20 DOI pending Open evidence boundary →

NWLE-10 · partial feasibility audit

UniPert cross-domain representation feasibility

NMD-VCell can evaluate UniPert now as a frozen gene-and-compound identity representation. The first release separates verified source and identifier readiness from every encoder, embedding, transfer and DMD-response claim.

No embedding or predictive result. The 21 / 21 identifier preflight only confirms that the governed candidate list can be mapped to reviewed human UniProt primary accessions. It does not show that UniPert encodes them correctly, improves prediction or transfers to DMD.
Borrowed component, bounded roleWhere UniPert can enter NMD-VCell—and where it stopsThe centre block is a frozen perturbagen-identity representation, not a condition-response simulator.
Genetic cause spaceGene · UniProt · FASTA21/21 identifiers ready
Chemical cause spaceName · SMILESPublic compound inputs supported
Identity bridgeUniPert frozen representationEncoder runs in NMD-VCell: 0
U2Relation retrievalregistered
U3Held-out family G2registered
U4Gene ↔ compound transferregistered
Not encoded directlydirectiondosetimemodalitycell stateDMD contextfunction

Source and implementation audit

Useful immediately as a bounded representation baseline

Open the complete source, implementation and rights audit
Audit itemVerified stateOperational consequence
Formal paperCell 2026 · DOI 10.1016/j.cell.2026.06.005Citable method and reported benchmarks; not DMD validation.
Public repositoryAudited commit 2f5d46930dcbdeb92073a13e898abe6e363e679aFrozen model and encoding demos are present.
Reproduction surfaceFull G2CP training and paper-split reproduction pipeline not found in the audited public treeDo not label a local encoder run as paper reproduction.
Condition semanticsIdentity inputs are available; direction, dose, time, modality and DMD state are not direct tokensUse as an identity feature, not a condition-response simulator.
RightsArticle CC BY · code GPL-3.0 · article reports related patent filingRetain licence provenance and perform separate deployment/FTO review.

U0–U4 execution ladder

Only U0 and the identifier part of U1 are complete

StageTaskStateCurrent result
U0Paper, code, licence, patent-notice and implementation-surface auditEXECUTEDSource and implementation boundaries frozen in this contract.
U1Candidate identifier and sequence-input preflightPARTIAL_IDENTIFIER_PREFLIGHT_EXECUTED_ENCODER_NOT_RUN21/21 candidates map to reviewed human UniProt primary accessions.
U2Cause-space sanity checks and known-relation retrievalREGISTERED_NOT_EXECUTEDNo result
U3Leakage-safe G2 incremental benchmarkREGISTERED_NOT_EXECUTEDNo result
U4Public gene-to-compound cross-domain learning curveREGISTERED_NOT_EXECUTEDNo result

Candidate input preflight

21 governed candidates have reviewed human primary accessions

Open all 21 candidate mappings
GeneUniProtSequence lengthState
ADAM10O14672748IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
CALRP27797417IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
CPEB1Q9BZB8566IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
DDX19BQ9UMR2479IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
DNAAF3Q8N9W5541IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
DNM1Q05193864IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
EHMT2Q96KQ71210IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
EPS8L1Q8TE68723IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
GFOD2Q3B7J2385IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
INTS13Q9NVM9706IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
LMO2P25791158IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
MON1AQ86VX9652IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
MPHOSPH6Q99547160IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
NAGLUP54802743IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
RAC3P60763192IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
RNASEH2CQ8TDP1164IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
RNF8O76064485IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
WDR4P57081412IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
ZFP69BQ9UJL9534IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
ZNF133P52736654IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN
ZNF236Q9UL361845IDENTIFIER_PREFLIGHT_PASS_ENCODER_NOT_RUN

Leakage-safe primary test

Held-out family or pathway

Compare frozen UniPert features with native ridge, ESM2-only, nearest-neighbour and random-embedding controls. Report verified-unseen and unresolved-exposure results separately.

Adoption gate

Incremental value, not visual plausibility

Adopt UniPert as a feature only if it improves a predeclared held-out task over simple baselines with uncertainty, survives exposure and leakage audit, and does not degrade worst-group performance.

Hard boundary

An identity embedding is not a simulated perturbation

It does not specify CRISPRi, CRISPRa, KO, overexpression, dose, time, cell state, disease background, mechanism, function or safety. Those variables need explicit conditioning and independently evaluated outcomes.

Boundary. This release verifies sources, implementation limits and 21-candidate identifier readiness. It publishes zero UniPert embeddings, zero UniPert predictive benchmarks and zero DMD functional outcomes.