GroupstrongQ9-landscape

NCI TARGET — a finished paediatric leukaemia genomics resource, not an active partner

also called TARGET · Therapeutically Applicable Research to Generate Effective Treatments · TARGET-ALL · TARGET-AML · NCI Office of Cancer Genomics TARGET

TARGET was a national programme that sequenced the genomes of thousands of children with leukaemia and other cancers, and its data still sits in a public archive anyone can apply to use. It is not an organisation you can collaborate with any more; it is a completed dataset with a front door.

The same thing, for a scientist

TARGET (Therapeutically Applicable Research to Generate Effective Treatments) was an NCI Office of Cancer Genomics initiative, run in partnership with the Children's Oncology Group, that molecularly characterised selected childhood cancers in a Discovery-then-Validation design using sequencing and array-based methods across genomes, transcriptomes and, for some diseases, epigenomes [E1]. Its leukaemia arms are the largest: **TARGET-AML holds 2,492 cases and 52,213 files; the three ALL phases hold 1,802 cases (P1 n=24, P2 n=1,587, P3 n=191) and 22,486 files** in the Genomic Data Commons [E2]. All TARGET data is hosted at the GDC [E1]. The AML arm characterised nearly 1,000 participants from COG AML trials by whole-genome, targeted DNA, mRNA and microRNA sequencing plus CpG methylation profiling [E3]. The practical status today is legacy: a completed, citable, applicable-for resource, not a partner with staff and a pipeline.

Go deeper

What is in it, precisely1 deeper

Counts taken directly from the GDC projects API on 2026-09-18 2:

Project	Disease	Cases	Files
TARGET-AML	Acute myeloid leukaemia	2,492	52,213
TARGET-ALL-P2	Acute lymphoblastic leukaemia, phase II	1,587	18,193
TARGET-ALL-P3	Acute lymphoblastic leukaemia, phase III	191	4,184
TARGET-ALL-P1	Acute lymphoblastic leukaemia, phase I	24	109
TARGET-NBL	Neuroblastoma	1,132	13,525
TARGET-WT	High-risk Wilms tumour	652	6,417
TARGET-OS	Osteosarcoma	383	4,314
TARGET-RT	Rhabdoid tumour	69	1,129
TARGET-CCSK	Clear cell sarcoma of the kidney	13	185

For this project the relevant total is 4,294 leukaemia cases across the AML and ALL arms.

Data types. NCI describes TARGET as employing sequencing and array-based methods examining genomes, transcriptomes and, for some diseases, epigenomes 1. The AML arm is the best-documented: whole-genome, targeted DNA, mRNA and microRNA sequencing, plus CpG methylation profiling, on nearly 1,000 COG AML trial participants 3.

Why the ALL phase structure matters. The three ALL projects are not equal. P2 (1,587 cases) is the bulk; P1 (24 cases) is a pilot. Anyone quoting 'TARGET ALL' as a single cohort is aggregating three programmes with different designs, eras and assay panels. The phase labels are a warning, not a detail.

Where it lives and how access works

Hosting. NCI states that all data collected and processed by TARGET is available at the Genomic Data Commons, with web-based analysis and visualisation tools 1. The GDC is the single front door; TARGET does not maintain a separate portal.

Access model — verified and unverified components. The GDC serves TARGET data in two tiers, an open tier (processed, non-identifiable summary-level data such as gene-level expression and mutation calls) and a controlled tier (raw sequence, BAMs, germline variants) requiring authorisation. The NCI pages read in this session did not state the access model, did not name dbGaP accession numbers and did not discuss authorisation requirements 14. The general two-tier GDC model with dbGaP-mediated authorisation for controlled data is standard practice across GDC-hosted NCI programmes, but for TARGET specifically it is inference here, not verified fact. An applicant should confirm the exact dbGaP accession and the authorisation route on the GDC portal before planning a timeline around it.

Practical consequence. The open tier is usable immediately and anonymously. The controlled tier requires institutional signing authority and a dbGaP application, which in the general NCI system is measured in weeks rather than days.

Why this is a resource and not a partner

This distinction determines how a laboratory should spend its time, so it is worth arguing rather than asserting.

The design was finite by construction. TARGET ran in two phases — Discovery, in which cohorts were used to identify molecular alterations, and Validation, in which independent patient cohorts and separate sequencing methods confirmed findings and estimated mutation frequency 4. That is the architecture of a characterisation project with an end state, not of a standing platform.

The publication record brackets it. The landmark TARGET ALL output is from 2009 5; the landmark TARGET AML output is from 2017-2018 3; the joint TARGET/PCGP pan-cancer synthesis is from 2018 6. The NCI programme page describes TARGET in completed terms — that 'a rich genomic data resource has been generated' and that the programme 'operated' in two phases 14.

Contrast with St. Jude Cloud. St. Jude Cloud is explicitly described by its builders as expanding, with regular uploads from prospective clinical genomics programmes. Nothing equivalent is claimed for TARGET. One is a living resource; the other is an archive.

Confidence caveat. The NCI pages read here did not contain an explicit statement that TARGET has ceased generating data, so 'completed legacy resource' is a strong inference from the design description, the publication record and the framing, not a quoted fact. Treat it as strong rather than established.

What follows for a small lab. Do not write to TARGET; there is nobody to write to. Write an application to the GDC. And do not treat a TARGET citation as evidence of a relationship — using the data creates no connection to the Children's Oncology Group that generated the specimens, which is a separate approach requiring a separate node and a separate person.

What it is genuinely good for, and what it is not

Good for. Establishing frequencies and co-occurrence in paediatric leukaemia at a scale a laboratory cannot assemble: 2,492 AML and 1,802 ALL cases with molecular characterisation 2. The TARGET AML analysis established that paediatric AML is structurally driven — somatic structural variants, new gene fusions and focal deletions were disproportionately prevalent in the young relative to adults, while DNMT3A and TP53 mutations common in adults were 'conspicuously absent from virtually all pediatric cases', and deletions, mutations and promoter hypermethylation converged on a cluster of zinc-finger genes associated with KMT2A rearrangements 3. That last finding is directly on this project's target and is the single most relevant result TARGET produced.

Good for, second. A denominator when making an argument. A grant or a specimen request that says 'in 2,492 TARGET AML cases, X co-occurs with KMT2A rearrangement at rate Y' is materially stronger than one that cites cell lines.

Not good for. Anything requiring treatment exposure to a menin inhibitor — TARGET predates the entire drug class. Anything requiring live cells, prospective sampling or a new assay. Anything requiring current-generation sequencing quality: the ALL data in particular spans array-based and early sequencing eras and the TARGET-ALL-P1 pilot has 24 cases and 109 files, which is not analysable on its own. And the pan-cancer synthesis itself cautions that the cohort is shaped by what the contributing programmes chose to sequence — high-risk, research-consented disease — so frequency estimates are not population estimates 6.

Is collaboration possible? Not in any meaningful sense. The correct framing is that TARGET is infrastructure a laboratory consumes, sitting alongside St. Jude Cloud as the second of two open resources that can be used before any human relationship exists.

Facts and figures

group	TARGET (Therapeutically Applicable Research to Generate Effective Treatments) — NCI Office of Cancer Genomics / Center for Cancer Genomics initiative, executed in partnership with the Children's Oncology Group
institution	National Cancer Institute, Bethesda, Maryland, USA; specimens and clinical annotation contributed largely by Children's Oncology Group trials; data hosted at the NCI Genomic Data Commons
focus	Molecular characterisation of selected childhood cancers — ALL (three phases), AML, neuroblastoma, Wilms tumour, osteosarcoma, rhabdoid tumour, clear cell sarcoma of the kidney — across genomes, transcriptomes and, for some diseases, epigenomes, in a Discovery-then-Validation design
status today	Completed legacy resource rather than an active partner. Strong inference, not an explicit NCI statement: the two-phase Discovery/Validation architecture is finite by design, the NCI pages use past-tense framing, and the landmark publications run 2009-2018 with nothing equivalent since. There is no standing organisation, staff or pipeline to collaborate with; the resource is consumed through the GDC.
data types	Sequencing and array-based methods across genomes, transcriptomes and, for some diseases, epigenomes (NCI). Documented in detail for the AML arm: whole-genome, targeted DNA, mRNA and microRNA sequencing plus CpG methylation profiling on nearly 1,000 COG AML trial participants.
where it lives	NCI Genomic Data Commons (GDC) — the single front door; TARGET maintains no separate portal.
access model	VERIFIED: all TARGET data is hosted at the GDC. NOT VERIFIED in this session: the specific dbGaP accession numbers, the open/controlled tier split for TARGET specifically, and the authorisation route. The standard GDC pattern — an open tier of processed, non-identifiable data usable immediately and a controlled tier of raw sequence and germline data requiring dbGaP authorisation with institutional signing authority — is inference from general NCI practice and must be confirmed on the GDC portal before being relied on.
partnering angle	There is no partner. TARGET is infrastructure to consume, not an organisation to approach, and a TARGET citation creates no relationship with the Children's Oncology Group that supplied the specimens. Its practical value is as a denominator: 2,492 AML and 1,802 ALL cases with molecular characterisation, including the finding that paediatric AML is structurally driven and that convergent deletion, mutation and promoter hypermethylation affect a zinc-finger gene cluster associated with KMT2A rearrangements. Use it to build a quantitative argument before asking COG or a company for anything.

key papers

cohort sizes

{
  "TARGET-AML": {
    "cases": 2492,
    "files": 52213
  },
  "TARGET-ALL-P1": {
    "cases": 24,
    "files": 109
  },
  "TARGET-ALL-P2": {
    "cases": 1587,
    "files": 18193
  },
  "TARGET-ALL-P3": {
    "cases": 191,
    "files": 4184
  },
  "TARGET-NBL": {
    "cases": 1132,
    "files": 13525
  },
  "TARGET-WT": {
    "cases": 652,
    "files": 6417
  },
  "TARGET-OS": {
    "cases": 383,
    "files": 4314
  },
  "TARGET-RT": {
    "cases": 69,
    "files": 1129
  },
  "TARGET-CCSK": {
    "cases": 13,
    "files": 185
  },
  "leukaemia_total_cases": 4294,
  "source": "GDC Projects API, retrieved 2026-09-18"
}

key papers — the numbers behind the chart

[
  {
    "title": "The molecular landscape of pediatric acute myeloid leukemia reveals recurrent structural alterations and age-specific mutational interactions",
    "year": 2018,
    "venue": "Nature Medicine",
    "pmid": "29227476",
    "doi": "10.1038/nm.4439"
  },
  {
    "title": "Pan-cancer genome and transcriptome analyses of 1,699 paediatric leukaemias and solid tumours",
    "year": 2018,
    "venue": "Nature",
    "pmid": "29489755",
    "doi": "10.1038/nature25795"
  },
  {
    "title": "Deletion of IKZF1 and prognosis in acute lymphoblastic leukemia",
    "year": 2009,
    "venue": "New England Journal of Medicine",
    "pmid": "19129520",
    "doi": "10.1056/NEJMoa0808253"
  }
]

How this could bite us back

A legacy resource cannot answer a question about a drug class that did not exist when it was built; no TARGET analysis will say anything about menin-inhibitor response or resistance.
Cohorts assembled from trial-enrolled, research-consented, high-risk disease produce frequency estimates that look authoritative and are not population frequencies — a mistake that is easy to make and hard to notice at n=2,492.
Assay heterogeneity across three ALL phases and a decade of technology means cross-cohort comparisons can generate batch artefacts that survive as apparent biology.
Because the data are open, any finding derivable from TARGET alone has almost certainly been looked for by better-resourced groups; its value to a small lab is in framing a question, not in answering one.

Still open

What are the actual dbGaP accession numbers for TARGET-ALL and TARGET-AML, and what proportion of the files are controlled-tier? Neither NCI page read here stated this, and it determines the real cost of access.
Has TARGET generated any new data since roughly 2018? The programme framing and publication record say no, but no NCI page reviewed here says so explicitly.
How many TARGET-ALL cases are infants under 12 months with KMT2A rearrangement, and do any have RNA-seq? The GDC case counts are disease-level and this subgroup could be too small to use.
Are the TARGET-ALL P1, P2 and P3 cohorts harmonised to a common pipeline in the GDC, or do the phase differences persist as batch effects that would confound any cross-phase analysis?

Every source behind this entry (6)

E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
TARGET is an NCI initiative that molecularly characterised selected childhood cancers using sequencing and array-based methods across genomes, transcriptomes and, for some diseases, epigenomes, with the stated goal of guiding development of more effective, less toxic therapies; all data collected and processed by the programme is available at the Genomic Data Commons.
TARGET — Therapeutically Applicable Research to Generate Effective Treatments (NCI Center for Cancer Genomics programme page) · National Cancer Institute, Center for Cancer Genomics / Office of Cancer Genomics · cancer.gov 2026
What this evidence cannot show
A programme description page by the funder. It does not state cohort sizes, does not enumerate data types per disease, does not name dbGaP accessions, does not describe the access model, and does not explicitly say whether the programme is still generating data. Retrieved 2026-09-18.
Evidence that argues against this (1)
- TARGET's stated goal is programme description. Which childhood cancers were "selected" is the operative limitation for this graph: infant KMT2A-rearranged ALL is not one of the deeply characterised diseases, and the AML arm that does exist returned no KMT2A fusion records in its structural-variant profile. Cancer Discovery 2021
E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
In the Genomic Data Commons, TARGET comprises nine projects: TARGET-AML (2,492 cases, 52,213 files), TARGET-ALL-P2 (1,587 cases, 18,193 files), TARGET-ALL-P3 (191 cases, 4,184 files), TARGET-ALL-P1 (24 cases, 109 files), TARGET-NBL (1,132), TARGET-WT (652), TARGET-OS (383), TARGET-RT (69) and TARGET-CCSK (13) — a total of 4,294 leukaemia cases across the AML and ALL arms.
GDC Projects API query for program.name = TARGET, returning project_id, name, summary.case_count and summary.file_count for all TARGET projects · NCI Genomic Data Commons · api.gdc.cancer.gov 2026
What this evidence cannot show
Case counts are cases registered in the GDC, not cases with any particular assay; a case may have only one file type. File counts include derived and processed files, so they overstate the amount of primary data. The count does not distinguish open-tier from controlled-tier files. Counts can change as the GDC reprocesses or harmonises data. Retrieved 2026-09-18.
Evidence that argues against this (1)
- Qualifies the headline case counts in a way that changes their usefulness here. The GDC project-level counts are real, but for this project the relevant denominator is not 4,294 leukaemia cases; it is infant KMT2A-rearranged cases with usable molecular data, and TARGET's leukaemia arms were built from COG frontline protocols with a minimum enrolment age of 1 year (see the counter-evidence on E5). The only sizeable infant KMT2A-rearranged sequencing cohort identified anywhere in this beat is the St. Jude PCGP series: whole-genome, exome, RNA and targeted DNA sequencing of 65 infants, of whom 47 were MLL-rearranged, plus 20 older children with MLL-rearranged leukaemia, reported by Andersson et al. That cohort also supplies the number that governs what any such analysis can achieve: a mean of 1.3 non-silent mutations in the predominant leukaemic clone, among the lowest somatic mutation burdens of any sequenced cancer. Recording this here so that a reader planning a dbGaP or GDC application against TARGET does not discover the age structure after the application. Caveat: the PCGP cohort is a decade old with pipelines of its time, and the exact whole-genome subset is not stated in the abstract. Read: abstract only. Andersson AK et al. 2015
E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The COG-NCI TARGET AML initiative characterised nearly 1,000 participants in Children's Oncology Group AML trials by whole-genome, targeted DNA, mRNA and microRNA sequencing plus CpG methylation profiling, finding that somatic structural variants including new gene fusions and focal deletions were disproportionately prevalent in young patients relative to adults, that DNMT3A and TP53 mutations common in adults were conspicuously absent from virtually all paediatric cases, and that deletions, mutations and promoter hypermethylation convergently affected a cluster of zinc finger-encoding genes associated with KMT2A rearrangements.
The molecular landscape of pediatric acute myeloid leukemia reveals recurrent structural alterations and age-specific mutational interactions · Bolouri H, Farrar JE, Triche T, Ries RE, Lim EL, Alonzo TA, Ma Y, Moore R, Mungall AJ, Marra MA, et al. · Nature Medicine 2018
PMID 29227476 DOI
What this evidence cannot show
Cases are COG AML trial participants, so the cohort is trial-enrolled and research-consented rather than a population sample, and frequency estimates inherit that selection. Assay panels differ across cases. Published 2017-2018 with contemporaneous methods; fusion and structural-variant detection has improved materially since. The zinc-finger/KMT2A convergence is an association from integrative analysis, not a demonstrated mechanism. According to PubMed; DOI https://doi.org/10.1038/nm.4439
Evidence that argues against this (1)
- Nearly 1,000 participants across a battery of assays is impressive and the KMT2A-rearranged subset within it is small and heterogeneous by partner. The initiative's own structural-variant calling is also the limiting factor for a fusion question, and this graph has recorded that it returned no KMT2A fusion records where they were expected. Cancer Discovery 2021
E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
TARGET operated in a two-phase design in which Discovery cohorts identified molecular alterations and Validation phases used independent patient cohorts and separate sequencing methods to confirm findings and estimate mutation frequency, and NCI describes the programme in terms of a resource that has been generated.
Using TARGET Data (NCI Center for Cancer Genomics) · National Cancer Institute, Center for Cancer Genomics · cancer.gov 2026
What this evidence cannot show
The page does not contain an explicit statement that TARGET has ceased data generation, nor does it give dbGaP accessions, per-disease case counts or the access model. The characterisation of TARGET as a completed legacy resource is therefore a strong inference from the finite two-phase design, the past-tense framing and a publication record that peaks in 2017-2018 — it is not a quoted fact and should be treated as such. Retrieved 2026-09-18.
Evidence that argues against this (1)
- A Discovery-then-Validation design controls for discovery-cohort overfitting and not for the assays' shared blind spots: if both phases use sequencing approaches that miss a lesion class, independent cohorts confirm the same absence. Mutation-frequency estimates from such a design are lower bounds. Blood Cells, Molecules and Diseases 2010
E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
TARGET's ALL arm produced foundational paediatric leukaemia genomics: analysis of 221 children with high-risk B-progenitor ALL with validation in an independent cohort of 258 identified IKZF1 deletion as a predictor of very poor outcome, with NCI TARGET (Gerhard, Smith) and COG (Hunger, Carroll, Reaman, Willman) investigators as co-authors.
Deletion of IKZF1 and prognosis in acute lymphoblastic leukemia · Mullighan CG, Su X, Zhang J, Radtke I, Phillips LAA, Miller CB, Ma J, Liu W, Cheng C, Schulman BA, et al. · New England Journal of Medicine 2009
PMID 19129520 DOI
What this evidence cannot show
Explicitly excluded infant ALL, BCR-ABL1-positive ALL and hypodiploid ALL from the discovery cohort, so it contributes nothing directly to infant KMT2A-rearranged biology. SNP-array copy-number technology of 2009, superseded by sequencing. Cited here as evidence of what the TARGET ALL resource enabled, not for its clinical conclusion. According to PubMed; DOI https://doi.org/10.1056/NEJMoa0808253
Evidence that argues against this (1)
- Answers this entry's third open question and the answer is close to zero, for a structural reason that is checkable rather than inferred. The entry asks 'How many TARGET-ALL cases are infants under 12 months with KMT2A rearrangement?' TARGET-ALL's B-ALL discovery material came from Children's Oncology Group frontline high-risk B-ALL protocols, and those protocols exclude infants by eligibility. The cohort behind the IKZF1 finding cited here as E5 was COG P9906, whose registry record (NCT00005603, 'ALinC 17: Protocol for Patients With Newly Diagnosed High Risk Acute Lymphoblastic Leukemia - Evaluation of the Augmented BFM Regimen', COG-P9906, completed, n=276) states a minimum age of 1 year. The successor high-risk protocol AALL0232 (NCT00075725, n=3,154) likewise registers a minimum age of 1 year. Infant ALL in COG was run on separate protocols - P9407 and AALL0631 (NCT00557193, maximum age 1 year) - which were not TARGET arms. So TARGET-ALL is structurally an over-1-year resource, and a project that needs infant KMT2A-rearranged genomes should go to the St. Jude PCGP infant cohort (Andersson et al. 2015, PMID 25730765, DOI 10.1038/ng.3230; 47 MLL-rearranged infants) rather than to TARGET, and should not spend a dbGaP application finding this out. A second consequence follows for study design: because TARGET was run in partnership with COG and drew its cases from COG trials, TARGET and COG are not independent sources, and treating them as two cohorts for discovery-and-replication would be double-counting the same patients. Caveats: TARGET-ALL-P2 holds 1,587 cases from more than one COG study and the per-study composition was not enumerated here, so 'close to zero infants' is an inference from the eligibility of the contributing protocols rather than a count from the GDC; a small number of infants could enter through the P3 or P1 arms. Read: registry records for NCT00005603, NCT00075725 and NCT00557193, retrieved 2026-09-18. Children's Oncology Group; Pediatric Oncology Group 2026
E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
TARGET data was combined with the Pediatric Cancer Genome Project in a pan-cancer analysis of 1,699 paediatric leukaemias and solid tumours under a uniform analytical framework, with NCI Office of Cancer Genomics staff as co-authors, establishing that only 45% of the 142 paediatric driver genes match adult pan-cancer studies and that copy-number alterations and structural variants constitute 62% of events.
Pan-cancer genome and transcriptome analyses of 1,699 paediatric leukaemias and solid tumours · Ma X, Liu Y, Liu Y, Alexandrov LB, Edmonson MN, Gawad C, Zhou X, Li Y, Rusch MC, Easton J, Hermida LC, Davis S, Smith MA, Guidry Auvil JM, Gerhard DS, Zhang J, et al. · Nature 2018
PMID 29489755 DOI
What this evidence cannot show
Six histotypes only. The cohort is the union of what TARGET and PCGP chose to sequence — high-risk, research-consented disease — so frequencies are not population frequencies. Driver-gene lists depend on the statistical framework applied. 2018 methods. According to PubMed; DOI https://doi.org/10.1038/nature25795
Evidence that argues against this (1)
- A uniform analytical framework across 1,699 paediatric tumours buys comparability at the cost of sensitivity for any one disease: parameters that suit a solid tumour are not tuned for a balanced translocation in a low-mutation-burden leukaemia. Blood Cells, Molecules and Diseases 2010

Connected to

supports →Kinase/PI3K/RAS pathway activation — the only common co-lesion class in KMT2A-r leukemia, and it is mostly subclonal
refutes →Infant acute lymphoblastic leukaemia (ALL diagnosed under 1 year of age)
Counter-evidence (NCT00005603 COG-P9906 MinimumAge = 1 Year; NCT00075725 AALL0232 likewise): TARGET-ALL's B-ALL cohorts came from COG frontline high-risk protocols that exclude infants, so TARGET holds essentially no infant KMT2A-r cases
related to →St. Jude Cloud and the Pediatric Cancer Genome Project — raw paediatric leukaemia genomes a small lab can get without a collaboration
Counter-evidence: TARGET was run in partnership with COG and drew its cases from COG trials, so PCGP+TARGET is not discovery-plus-independent-replication; and the infant cohort lives in PCGP (47 MLL-r genomes), not TARGET
supports →Who holds the data and samples we cannot generate ourselves — an access map
open legacy paediatric cohort in GDC/dbGaP; useful now, not an active partner