Dataset Registry · governed subset
Which registered datasets can support which computational task? This page contains 5 task-audited datasets selected from the broader 126-record Data Universe. It is a computational-role registry—not a claim that the platform contains only five datasets.
Task-audited datasets 5
Same-context training substrates 1
DMD candidate-training status Not available yet
Product contract · DATASET-CARD-1.0
Dataset Card: What was measured, in which biological units, and what can it validly support? Identity + version Disease + cell context Donor + statistical unit Matrix + metadata + hashes Access + licence Readiness + linked runs
Data estate 12 workspace + 80 /share families Storage presence
→ Public inventory 126 mapped records Searchable summaries
→ Task registry 5 audited datasets Computational roles assigned
→ DMD prediction training Not qualified Matched outcomes required
No scalar readiness score. A dataset can qualify context, trajectory, disease correction or assay design without becoming direct evidence that a displayed candidate perturbation works in DMD.
Registry contract
Each selected dataset is audited against required metadata axes. The Data Universe answers “what exists?”; this registry answers “what exact task may this object support?” Inclusion here is deliberately narrower than storage or analysis presence.
Open machine-readable registry → accession and permanent URL disease, species and tissue/model donor, age, sex and disease stage modality, observation unit and cell/sample count perturbation, dose, time and endpoint replicates, batch, platform and hash chain leakage grouping and benchmark-use eligibility Training outcome boundary: disease-context evidence, perturbation evidence and calibrated prediction truth remain separate fields.
Dataset Disease context Perturbation outcome Cell-level outcome Time Donor / batch Current AI role internal-release-substrate ABSENT NOT DMD AVAILABLE SINGLE CRISPRI PARTIAL COMPONENT RESOLVED SINGLE FROZEN ENDPOINT LIMITED RELEASED DIAGNOSTICS BASELINE READY SAME CONTEXT ONLY GSE293514 MUSCLE CONTEXT NOT DMD RESCUE AVAILABLE KNOCKOUT SCREEN ENDPOINT NOT RELEASED AS CELL POPULATION TRUTH PUBLICATION ENDPOINT NOT QUALIFIED FOR CURRENT MODEL CONTEXT SCREEN ONLY GSE272233 AVAILABLE DMD CORRECTION REFERENCE DMD LOCUS CORRECTION NOT CANDIDATE PERTURBATION ABSENT BULK RNA SINGLE STUDY ENDPOINT GROUPED SAMPLE IDENTITIES AVAILABLE ORTHOGONAL REFERENCE ONLY GSE52529 ABSENT NON DISEASE REFERENCE ABSENT UNPERTURBED AVAILABLE SINGLE CELL REFERENCE AVAILABLE 0 24 48 72 HOURS NOT SUITABLE FOR UNSEEN DONOR PERTURBATION TRAJECTORY REFERENCE ONLY GSE133344 ABSENT K562 METHOD REFERENCE AVAILABLE CRISPRA SINGLES AND PAIRS AVAILABLE PUBLIC METHOD REFERENCE SINGLE STUDY ENDPOINT NOT A DONOR GENERALIZATION SET METHOD CALIBRATION ONLY
internal-release-substrate Observed HepG2 CRISPRi candidate responses BASELINE READY SAME CONTEXT ONLY
Disease / role Context-limited; not DMD
Cell type HepG2
Perturbation CRISPRi single-gene
Time Frozen endpoint aggregate
Size 21 governed candidate aggregates
Candidate coverage 21/21 Current AI role Train and audit the same-context baseline No matched muscle or DMD perturbation outcome
Permitted use: Direct same-context perturbation observation
Internal release licence pending institutional approval
GSE293514 Human myoblast fusion CRISPR screen CONTEXT SCREEN ONLY
Disease / role Muscle context; not DMD rescue
Cell type Human myoblast
Perturbation Knockout screen
Time Publication endpoint
Size 7,197 screen genes; 250 hits; 125 individually validated genes
Candidate coverage 9/21; 0 current fusion hits Current AI role Qualify muscle-context coverage and endpoint compatibility Not an independent candidate perturbation in a DMD model
Permitted use: Observed healthy-myoblast functional perturbation reference (Stage B0), not DMD replication
Source repository and publication terms apply; verify at accession
GSE272233 DMD-locus CRISPR-correction transcript reference ORTHOGONAL REFERENCE ONLY
Disease / role Duchenne muscular dystrophy
Cell type Patient-derived and corrected cellular models
Perturbation DMD-locus correction
Time Study endpoint
Size 21 samples; 3 per reported group
Candidate coverage Candidate transcripts observed; no candidate perturbation Current AI role Check disease-correction direction and context Candidate transcripts are observed but candidates were not perturbed
Permitted use: Orthogonal disease-correction reference
Source repository and publication terms apply; verify at accession
GSE52529 Human myogenic differentiation time course TRAJECTORY REFERENCE ONLY
Disease / role Non-disease trajectory reference
Cell type Primary human skeletal myoblast
Perturbation None
Time 0, 24, 48 and 72 hours
Size 271 cells
Candidate coverage 20/21 measured Current AI role Build an unperturbed human-myogenic state reference No perturbation-conditioned trajectory and no longitudinal DMD state
Permitted use: Unperturbed trajectory reference only
Source repository and package terms apply; verify at accession
GSE133344 CRISPRa single and combinatorial calibration reference METHOD CALIBRATION ONLY
Disease / role Non-muscle method calibration
Cell type K562
Perturbation CRISPRa singles and pairs
Time Study endpoint
Size 287 perturbations; 112 singles; 132 pairs
Candidate coverage 20/21 present as expression features Current AI role Calibrate activation and combination task design No matched muscle or DMD activation response for current candidates
Permitted use: Method calibration only; not muscle/DMD overexpression evidence
Source repository and publication terms apply; verify at accession