Dataset: TARGET via the GDC - open expression and mutations, but the fusion calls themselves are controlled
also called TARGET access plan · GDC pull · dbGaP phs000464 phs000465
TARGET is a large US programme that sequenced childhood cancers, and its data now lives in the NIH's Genomic Data Commons. Much of it is free to download. The one thing a fusion project most wants - the file listing which fusion each patient has - is locked behind an application.
The same thing, for a scientist
TARGET is the largest open-tier paediatric leukaemia genomics resource and the natural second stop after cBioPortal. Its data lives in the GDC, whose API is open and key-free. Two projects matter: TARGET-ALL-P2 (dbGaP phs000464.v23.p8, 1,587 cases, 18,193 files) and TARGET-AML (dbGaP phs000465.v23.p8, 2,492 cases, 52,213 files). The open/controlled split was computed directly this cycle: TARGET-ALL-P2 has 2,408 open and 15,785 controlled files; TARGET-AML has 10,479 open and 41,734 controlled. Open covers gene/isoform/miRNA expression quantification, masked somatic mutation, gene-level and allele-specific copy number, and methylation beta values. **Transcript Fusion files are controlled** - 1,064 for TARGET-ALL-P2 and all 12,908 for TARGET-AML - which means breakpoint-level fusion calls require a dbGaP Data Access Request.
What question it answers
On the open tier: expression and methylation phenotypes of paediatric ALL and AML at cohort scale. This is enough to test an expression-based signature, to look at methylation beta values, and to ask co-mutation questions from masked somatic mutation calls. It is enough for the lineage-switch signature scoring analysis if that analysis is reframed to use expression rather than fusion status from the same files.
Only on the controlled tier: anything that needs the fusion call itself, the breakpoint, the aligned reads, or unmasked somatic variants.
How to reach it, and the exact split
GDC API, open and anonymous: https://api.gdc.cancer.gov/projects?filters=...program.name=TARGET and https://api.gdc.cancer.gov/files with project and access filters 1.
| Project | dbGaP | Cases | Open files | Controlled files |
|---|---|---|---|---|
| TARGET-ALL-P2 | phs000464.v23.p8 | 1,587 | 2,408 | 15,785 |
| TARGET-AML | phs000465.v23.p8 | 2,492 | 10,479 | 41,734 |
Also present: TARGET-ALL-P1 (phs000463, 24 cases) and TARGET-ALL-P3 (191 cases, no dbGaP accession recorded in the GDC, which is itself worth knowing before planning around it).
The access barrier
A dbGaP Data Access Request against phs000464 and phs000465, requiring institutional signing-official approval and a Data Use Certification. This is heavier than the St. Jude Cloud DAA: it goes through NIH, needs an eRA Commons account and an institutional signing official, and renewals are annual.
The honest trade-off: St. Jude Cloud is better matched to this disease (infant ALL, ALL across the age continuum) and has a lighter process. TARGET is larger on the open tier. If only one application is made, make it St. Jude Cloud. Apply to dbGaP only when a specific analysis needs TARGET fusion calls or aligned reads, and note that the paediatric ALL fusion population in TARGET is dominated by subtypes other than KMT2A::AFF1.
| resource | TARGET (Therapeutically Applicable Research to Generate Effective Treatments), hosted in the NCI Genomic Data Commons |
|---|---|
| question answered | Cohort-scale paediatric ALL and AML expression, methylation and masked somatic mutations on the open tier; breakpoint-level fusion calls and aligned reads only on the controlled tier. |
| recommendation | Exhaust the open expression and methylation tier first. Apply to dbGaP only when an analysis specifically needs fusion calls or aligned reads. If only one controlled-access application is made this phase, make it St. Jude Cloud PanALL, which is better matched to infant KMT2A-rearranged disease and has a lighter process. |
| status | planned |
| compute | laptop for the open expression tier; controlled-tier reanalysis would need Explorer Slurm |
access
{
"api": "https://api.gdc.cancer.gov",
"auth_open_tier": "none",
"auth_controlled_tier": "dbGaP Data Access Request with institutional signing official and Data Use Certification"
}projects — the numbers behind the chart
[
{
"project_id": "TARGET-ALL-P2",
"dbgap": "phs000464.v23.p8",
"cases": 1587,
"files": 18193,
"open_files": 2408,
"controlled_files": 15785,
"transcript_fusion_files_controlled": 1064
},
{
"project_id": "TARGET-AML",
"dbgap": "phs000465.v23.p8",
"cases": 2492,
"files": 52213,
"open_files": 10479,
"controlled_files": 41734,
"transcript_fusion_files_controlled": 12908
},
{
"project_id": "TARGET-ALL-P1",
"dbgap": "phs000463",
"cases": 24
},
{
"project_id": "TARGET-ALL-P3",
"dbgap": null,
"cases": 191,
"note": "no dbGaP accession recorded in the GDC"
}
]open data types
[ "Gene Expression Quantification", "Isoform Expression Quantification", "miRNA Expression Quantification", "Masked Somatic Mutation", "Gene Level Copy Number", "Allele-specific Copy Number", "Methylation Beta Value" ]
controlled data types
[ "Aligned Reads", "Raw and Annotated Simple Somatic Mutation", "Splice Junction Quantification", "Transcript Fusion" ]
- A dbGaP application is an institutional commitment with annual renewal; starting one for a cohort that turns out to hold only a few dozen relevant infants would consume administrative capacity that the St. Jude application needs.
- Masked somatic mutation calls on the open tier are masked precisely to prevent re-identification, and the masking removes exactly the low-allele-fraction variants that matter for resistance-mutation questions.
- How many TARGET-ALL-P2 cases are KMT2A-rearranged, and how many of those are infants - the cBioPortal clinical annotation suggests only 48 MLL-positive samples, which may under-call badly?
- Why does TARGET-ALL-P3 have no dbGaP accession in the GDC, and does that make its 191 cases easier or harder to reach?
- Can the open expression tier alone power the lineage-switch signature analysis if fusion status is taken from clinical annotation rather than from the controlled fusion calls?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
TARGET data lives in the GDC under dbGaP accessions phs000464.v23.p8 (TARGET-ALL-P2, 1,587 cases, 18,193 files) and phs000465.v23.p8 (TARGET-AML, 2,492 cases, 52,213 files); the GDC API is open, and the open/controlled split is 2,408/15,785 and 10,479/41,734 respectively, with Transcript Fusion files falling on the controlled side.
NCI Genomic Data Commons API, projects and files endpoints filtered to program TARGET; dbGaP study pages phs000464 and phs000465 · NCI GDC / NCBI dbGaP 2026What this evidence cannot show
File counts were computed by a verification worker from live API responses this cycle and change with each GDC data release. The dbGaP version suffixes (.v23.p8) will advance. TIGHTENED 2026-09-19: TARGET-ALL Phase II accrued through COG AALL03B1 (NCT00482352), registered minimum age 1 Year, so the 1,587 TARGET-ALL-P2 cases contain essentially no infant ALL; infants were enrolled separately on COG AALL0631 (NCT00557193, maximum age 1 Year).
Evidence that argues against this (3)
- TARGET-ALL Phase II is, for this project's purpose, the wrong cohort - and the reason is on the trial registry, not in the literature. Its cases were accrued through COG AALL03B1 (NCT00482352), whose registered eligibility is minimum age 1 Year, maximum 30 Years, 11,196 enrolled. Infants under one year were excluded by design. TARGET-ALL-P2's 1,587 cases therefore contain essentially no infant ALL, which is the population in which KMT2A::AFF1 is most common. ClinicalTrials.gov 2026
- Infants were enrolled on a separate protocol: COG AALL0631 (NCT00557193) carries a registered maximum age of 1 Year. So the infant KMT2A-r cohort exists as a COG dataset but sits outside the dbGaP accession this dossier describes. Anyone planning to draw infant KMT2A-r cases from phs000464 should expect to find none and should plan for a separate data request. ClinicalTrials.gov 2026
- The controlled/open split compounds this. Transcript Fusion files fall on the controlled side of both accessions, so fusion calls cannot be regenerated or audited without dbGaP approval - and fusion recall is caller-dependent to a degree that changes conclusions (Haas et al. benchmarked 23 methods and found substantial accuracy differences). The project would be inheriting somebody else's caller without being able to check it. Haas BJ et al. 2019
uses the TARGET paediatric ALL expression cohort
The same age gap runs through both resources: DepMap's KMT2A-r panel has no infant line, and TARGET-ALL Phase II draws from COG AALL03B1 whose registered minimum age is 1 year.
Harmonisation counter-evidence: St. Jude Cloud and TARGET differ in institution, assay generation and annotation, and batch structure of that kind is a documented source of incorrect conclusions (PMID 20838408). Pooling them is a decision that must be argued, not a default.
Hard counter-evidence from the trial registry: TARGET-ALL Phase II accrued through COG AALL03B1 (NCT00482352, minimum age 1 Year), so infants are excluded by design; infant ALL went to COG AALL0631 (NCT00557193, maximum age 1 Year).
if only one controlled-access application is made, make it St. Jude Cloud