Dataset: cBioPortal - open, easy, and containing exactly two KMT2A::AFF1 fusion records
also called cBioPortal access plan · public cohort pull · TARGET in cBioPortal
cBioPortal is the friendliest cancer-genomics database there is: no login, no forms, a simple web interface. We checked what it actually holds for our disease, and the answer is sobering - only two records of the exact fusion this project is about.
The same thing, for a scientist
The cBioPortal public instance is fully open (546 studies, anonymous REST, no key) and is the correct first stop for co-mutation questions. For KMT2A::AFF1 specifically it is close to empty, and a phase-2 plan that assumes otherwise will fail. Direct queries this cycle found that the only study in the public instance carrying a structural-variant profile relevant here is aml_stjude_2024, which holds 237 KMT2A structural-variant records across 180 samples - of which **two** are KMT2A::AFF1, against 67 KMT2A::MLLT3 and 65 KMT2A::MLLT10. The TARGET studies in cBioPortal carry no fusion or structural-variant profile at all; KMT2A status there exists only as a clinical annotation (171 of 886 patients labelled 'MLL' in aml_target_2018_pub; 48 MLL_STATUS-positive samples and 29 labelled MLL-Rearranged in all_phase2_target_2018_pub). This is a hard ceiling on what the open tier can answer.
What question it answers
Co-mutation landscape, copy-number context, and - in one study - fusion partner distribution. The graph's existing co-mutation entries (KRAS, NRAS, FLT3, PTPN11, NF1) are the sort of claim this resource is for.
What it cannot answer for this project: anything at breakpoint resolution, anything about infants specifically, and anything requiring more than a couple of KMT2A::AFF1 cases.
How to reach it, and what the queries return
Anonymous REST, no key. GET https://www.cbioportal.org/api/studies returns 546 public studies 1.
Fusion data requires a structural-variant profile, and most studies do not have one. The productive query is a POST to /api/structural-variant/fetch with {"molecularProfileIds":["aml_stjude_2024_structural_variants"],"entrezGeneIds":[4297]} - Entrez 4297 is KMT2A. That returns 237 records across 180 distinct samples, with the partner distribution: MLLT3 67, MLLT10 65, AFDN 25, ELL 15, MLLT1 8, MLLT6 5, AFF1 2 1.
The relevant TARGET studies - all_phase2_target_2018_pub (1,978 samples) and aml_target_2018_pub (1,025 samples) - expose only cna, methylation, mirna, mrna and mutation profiles. KMT2A status is a clinical field, not a molecular one.
The access barrier
None for the public instance. No registration, no agreement, no key, and it is safe to automate. This makes it the right resource to exhaust before applying for anything controlled.
The barrier is content, not access, and it is severe for this project: two KMT2A::AFF1 fusion records is not a cohort. Any phase-2 question that needs KMT2A::AFF1 patient molecular data at scale must go to St. Jude Cloud or TARGET-controlled data, both of which require agreements.
The one thing this resource is genuinely good for here
Partner-distribution context. The 237-record breakdown from a single modern AML study - MLLT3 and MLLT10 dominant, AFF1 almost absent - is itself a useful fact, because it says that a KMT2A-rearranged AML cohort is a different disease population from the infant ALL population this project cares about. Any dataset that is majority KMT2A::MLLT3 AML should not be used to make claims about infant KMT2A::AFF1 ALL, and cBioPortal makes that mismatch measurable in one query.
| resource | cBioPortal for Cancer Genomics, public instance |
|---|---|
| question answered | Co-mutation landscape and copy-number context; partner distribution in one modern AML cohort. |
| access barrier | None. The barrier is content: two KMT2A::AFF1 structural-variant records in the entire public instance. |
| status | run |
| result summary | The open tier cannot support any KMT2A::AFF1 cohort analysis: two fusion records exist. Its real value is the partner-distribution query, which quantifies how different a KMT2A-rearranged AML cohort is from infant KMT2A::AFF1 ALL and therefore which published cohorts must not be used as proxies. |
| compute | laptop; no GPU |
access
{
"auth": "none",
"endpoint": "https://www.cbioportal.org/api",
"n_public_studies": 546,
"safe_to_automate": true
}relevant studies — the numbers behind the chart
[
{
"id": "aml_stjude_2024",
"name": "Pediatric Acute Myeloid Leukemia (St. Jude, Nat Genet 2024)",
"n": 887,
"has_sv_profile": true
},
{
"id": "all_phase2_target_2018_pub",
"name": "Pediatric Acute Lymphoid Leukemia - Phase II (TARGET, 2018)",
"n": 1978,
"has_sv_profile": false,
"mll_status_positive": 48,
"molecular_subtype_mll_rearranged": 29
},
{
"id": "aml_target_2018_pub",
"name": "Pediatric Acute Myeloid Leukemia (TARGET, 2018)",
"n": 1025,
"has_sv_profile": false,
"cytogenetic_code_mll": 171,
"of_patients": 886
},
{
"id": "bll_target_gdc",
"n": 989
},
{
"id": "aml_target_gdc",
"n": 2766
},
{
"id": "alal_target_gdc",
"n": 251
}
]kmt2a sv partner distribution aml stjude 2024 — the numbers behind the chart
{
"total_records": 237,
"distinct_samples": 180,
"MLLT3": 67,
"MLLT10": 65,
"AFDN": 25,
"ELL": 15,
"MLLT1": 8,
"MLLT6": 5,
"AFF1": 2
}- The ease of this resource is a trap: it is tempting to build a phase-2 analysis on the cohort that is easy to get rather than the cohort that matches the disease, and a KMT2A::MLLT3-dominant AML cohort is a different disease from infant KMT2A::AFF1 ALL in cell of origin, age and outcome.
- Is there any public resource with more than a handful of KMT2A::AFF1 cases at molecular resolution, short of a data-access agreement?
- Do institutional cBioPortal instances hold infant ALL cohorts that the public instance does not, and is the Iqbal Lab able to reach one through a collaborator?
- How much of the field's KMT2A-rearranged evidence base is actually KMT2A::MLLT3 AML being generalised to infant KMT2A::AFF1 ALL?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
The cBioPortal public REST API is open with no key and returns 546 studies; the only relevant study with a structural-variant profile, aml_stjude_2024, holds 237 KMT2A structural-variant records across 180 samples of which only two are KMT2A::AFF1, while the TARGET studies carry no fusion profile and record KMT2A status only as a clinical annotation (171 of 886 'MLL' in aml_target_2018_pub; 48 MLL_STATUS-positive in all_phase2_target_2018_pub).
cBioPortal public REST API: /api/studies, /api/structural-variant/fetch (molecularProfileIds aml_stjude_2024_structural_variants, entrezGeneIds 4297) · cBioPortal for Cancer Genomics 2026What this evidence cannot show
Counts were computed by a verification worker from live API responses this cycle and reflect the public instance only; institutional cBioPortal instances may hold more. The clinical MLL annotations in TARGET are of unknown assay provenance and may under-call.
Evidence that argues against this (2)
- The two KMT2A::AFF1 records this query returned are not a measurement of rarity, they are a measurement of which pipeline was run. Haas et al. benchmarked 23 fusion-detection methods on simulated and real RNA-seq and found accuracy and recall differ substantially between callers, with de novo assembly-based methods notably less accurate. A portal study's structural-variant profile inherits whichever caller and filter set its submitters used, so absence of a fusion in cBioPortal is weak evidence of absence in the cohort. Haas BJ et al. 2019
- Pooling the aml_stjude_2024 structural-variant profile with the TARGET studies' clinical 'MLL' annotation is a harmonisation this project should not perform by default: one is a sequencing call and the other is a curated clinical flag with different ascertainment. Leek et al. document that technical and processing differences of exactly this kind are widespread, frequently correlate with the variable of interest and have produced incorrect published conclusions. Leek JT et al. 2010
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
Counter-evidence: only two KMT2A::AFF1 structural-variant records exist across cBioPortal's 546 studies, but fusion-caller benchmarking (PMID 31639029) shows recall varies widely between methods, so this is a pipeline artefact as much as a cohort fact.
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the public instance holds two KMT2A::AFF1 structural-variant records in total
the open tier must be exhausted first; it holds two relevant fusion records