Oncofusion

start / graph / ana-phase2-dataset-stjude-cloud

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q14-phase2-readiness

Dataset: St. Jude Cloud - the right cohort for this disease, behind a Data Access Agreement we have not applied for

also called St Jude Cloud access plan · PanALL DAU · PCGP access

St. Jude Cloud holds the largest collection of childhood cancer genomes in the world, including the infant leukaemias this project is about, but the raw data needs a signed agreement reviewed by a committee. Nobody in this project has applied yet, and that free application is the single cheapest thing we could do.

The same thing, for a scientist

St. Jude Cloud is the best-matched cohort resource for infant KMT2A-rearranged ALL and the most important controlled-access application phase 2 should make. Access is tiered: the PeCan knowledgebase and Visualization Community are open anonymously; a free account permits running workflows and accessing RNA-seq expression; raw WGS, WES and RNA-seq require a signed Data Access Agreement reviewed by a Data Access Committee, granted per Data Access Unit. Seven DAUs exist, of which **PanALL** ('ALL subtypes across the age continuum') and **PCGP** are the relevant ones. The founding publication reports 12,104 whole genomes, 7,697 whole exomes and 2,202 transcriptomes across more than 10,000 paediatric cancer patients, with a median DAA approval time of seven days and an 18.8% rejection rate. Current totals and, critically, the number of infant-ALL or KMT2A-rearranged cases held are UNVERIFIED - no published count exists and the portal is a JavaScript application with no countable endpoint.

Go deeper
What question it answers

Several that nothing else in this beat can:

  • The breakpoint prognostic reanalysis (ana-interfant-breakpoint-prognostic-reanalysis) needs genomic breakpoints with outcome data on real infants. That is exactly a PanALL/PCGP question.
  • The methylome cell-of-origin hypotheses (hyp-methylome-of-origin-sets-fusion-target-set, hyp-pentimento-ground-predicts-post-drug-destination) need infant methylation arrays.
  • The lineage-switch signature work needs paired diagnosis-relapse transcriptomes in KMT2A-rearranged cases.

No open resource supplies any of these at the required n. cBioPortal's public instance holds two KMT2A::AFF1 fusion records.

How to reach it, and the exact barrier

Three tiers 1:

  1. Open, anonymous: PeCan knowledgebase and Visualization Community.
  2. Registered, free account: workflows and RNA-seq expression.
  3. Controlled: raw WGS/WES/RNA-seq, requiring a signed Data Access Agreement reviewed by a Data Access Committee, granted at the level of a Data Access Unit. The documentation states plainly: there is a request process to protect this data even though it is de-identified.

Seven DAUs exist: PCGP, SJLIFE, Clinical Genomics, Sickle Cell Genome Project, CCSS, PanALL, and CReATe. One approval covers future data within that DAU, which makes applying to PanALL and PCGP a durable investment rather than a per-study cost.

There is no dbGaP dependency here - the DAA is St. Jude's own.

Scale, and what is genuinely unknown

From the founding publication 2: 12,104 whole genomes, 7,697 whole exomes, 2,202 transcriptomes, 1.25 PB, covering more than 10,000 paediatric cancer patients and survivors. Median DAA approval turnaround seven days; 18.8% of requests rejected.

UNVERIFIED and material: the current (2026) totals, and the number of infant-ALL or KMT2A-rearranged cases held. No page or API publishes these. PCGP and PanALL certainly contain such cases; how many is unknown. A phase-2 plan should therefore treat cohort size as a question to ask the Data Access Committee during the application, not as a number to assume.

The recommendation

Apply now, before any analysis needs it. The application is free, the median turnaround is a week, one in five is rejected, and approval is durable at the DAU level. Every one of the three phase-2 questions above is blocked on it. An application submitted at the moment an analysis is ready is an application submitted a week too late - and if rejected, a month too late.

This is the highest-value zero-cost action identified in this readiness review after the desk analyses.

Facts and figures
resourceSt. Jude Cloud Genomics Platform
question answeredBreakpoint-resolution genomics, methylation and transcriptomes on real paediatric and infant ALL patients with outcome data - the cohort no open resource supplies.
recommendationSubmit PanALL and PCGP applications immediately, ahead of need. Free, durable, median one week, one in five rejected. This is the highest-value zero-cost action in the readiness review after the desk analyses.
statusplanned
computeSt. Jude Cloud runs workflows in its own cloud environment; local compute is not the constraint
access tiers
{
  "open_anonymous": "PeCan knowledgebase, Visualization Community",
  "registered_free": "workflows, RNA-seq expression",
  "controlled": "raw WGS/WES/RNA-seq via signed Data Access Agreement reviewed by a Data Access Committee, granted per Data Access Unit"
}
relevant daus
[
  "PanALL (ALL subtypes across the age continuum)",
  "PCGP (Pediatric Cancer Genome Project)"
]
all daus
[
  "PCGP",
  "SJLIFE",
  "Clinical Genomics",
  "Sickle Cell Genome Project",
  "CCSS",
  "PanALL",
  "CReATe"
]
scale 2021 publication
{
  "whole_genomes": 12104,
  "whole_exomes": 7697,
  "transcriptomes": 2202,
  "patients": ">10000",
  "total_size": "1.25 PB"
}
daa process
{
  "median_approval_days": 7,
  "rejection_rate_pct": 18.8,
  "dbgap_required": false,
  "durability": "approval covers future data within the same DAU"
}
unverified
[
  "current 2026 totals",
  "number of infant-ALL cases held",
  "number of KMT2A-rearranged cases held"
]
blocked analyses in this graph
[
  "ana-interfant-breakpoint-prognostic-reanalysis",
  "ana-infant-methylome-hoxa-predictor",
  "ana-pentimento-methylation-ground-destination",
  "ana-lineage-switch-signature-scoring"
]
How this could bite us back
  • Building a phase-2 plan around a cohort whose size is unverified risks designing an analysis that the data cannot power. The application should ask for counts before the analysis is designed, not after.
  • A one-in-five rejection rate means the plan needs a fallback: if PanALL is refused, the breakpoint and methylome questions have no open substitute and those hypotheses become untestable by this lab.
Still open
  1. How many KMT2A::AFF1 infant cases does PanALL actually contain, and can that be asked of the Data Access Committee before applying?
  2. Does PanALL include methylation arrays, or only sequence - because two of the blocked analyses need methylation specifically?
  3. What are the commonest reasons for the 18.8% of applications that are rejected, and can the application be written to avoid them?
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    St. Jude Cloud operates three access tiers, with raw WGS, WES and RNA-seq requiring a signed Data Access Agreement reviewed by a Data Access Committee and granted at the level of one of seven Data Access Units, of which PanALL covers ALL subtypes across the age continuum.

    What this evidence cannot show

    Fetched this cycle by a verification worker from the documentation site. Access policy can change, and DAU composition has changed historically.

    Evidence that argues against this (1)
    • Access tiering is not the only barrier; the data access unit boundary is also an analytical one. PanALL spans ALL subtypes across the age continuum, which means any infant-versus-older comparison drawn from it crosses assay generations, sequencing platforms and annotation vintages within a single approved unit. Leek et al. document that such technical structure is widespread, frequently correlated with the biological variable of interest, and has produced incorrect conclusions - so harmonising across the unit is not sound by default. Leek JT et al. 2010
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The St. Jude Cloud founding publication reports 12,104 whole genomes, 7,697 whole exomes and 2,202 transcriptomes across more than 10,000 paediatric cancer patients and survivors, with a median data-access approval turnaround of seven days and an 18.8% rejection rate.

    St. Jude Cloud: A Pediatric Cancer Genomic Data-Sharing Ecosystem · McLeod C, Gout AM, Zhou X, et al. · Cancer Discovery 2021
    PMID 33408242DOI
    What this evidence cannot show

    2021 figures. Current totals are UNVERIFIED: the portal is a JavaScript application and publishes no countable endpoint, and no updated count could be fetched this cycle.

    Evidence that argues against this (2)
    • The headline totals are a poor guide to what is usable for this project. The 2,202 transcriptomes are spread across the whole of paediatric oncology and survivorship, not concentrated in KMT2A-rearranged infant ALL, and the subtype classifications demonstrated in the founding paper are expression-based calls whose fusion evidence depends on the caller used - a choice Haas et al. show changes recall substantially. A count of transcriptomes is not a count of confirmed KMT2A::AFF1 cases. Haas BJ et al. 2019
    • The resource is also single-institution in origin, with regular uploads from St. Jude's own prospective clinical genomics programmes, so its cohort composition carries that referral pattern. The paper's own reported 18.8% data-access rejection rate and seven-day median turnaround are the access facts; the composition fact is that this is one hospital's patient population, which is not interchangeable with a cooperative-group cohort such as TARGET or Interfant. McLeod C et al. 2021
Connected to