HTAN KMT2A-r ALL single-cell multiome — 25 patients (18 infants), scRNA-seq + scATAC-seq + snmC-seq2, with 2 paired pre/post lineage-switch cases and analysis code on GitHub
also called Tan lab KMT2Ar dataset · Chen 2022 Blood single-cell multiomics resource · HTAN KMT2A-rearranged ALL atlas
This is the single most useful public dataset for the questions this project asks about which cells a KMT2A leukaemia is made of: twenty-five patients, eighteen of them babies, with three different measurements made on the same individual cells — which genes are on, which bits of DNA are open, and which bits are chemically silenced. Two of the patients were sampled before and after their leukaemia changed lineage. The analysis code is public too.
The same thing, for a scientist
The resource behind PMID 34864916. **Deposited at the Human Tumor Atlas Network (HTAN) data portal** (`https://data.humantumoratlas.org/`), with the linkage between HTAN parent IDs and sample IDs given in the paper's supplemental Table 1, and analysis scripts at `https://github.com/tanlabcode/KMT2Ar_Paper` [E1]. Contents: **25 KMT2A-rearranged ALL patients** — 18 diagnosed under 1 year, 7 diagnosed after 1 year — plus **5 healthy paediatric donors** as reference, profiled by **scRNA-seq, scATAC-seq and snmC-seq2** (single-nucleus methylcytosine sequencing), with **2 patients contributing paired longitudinal samples spanning a documented lineage switch** [E1][E2]. This is not a GEO series and will not be found by a GEO search, which is one reason it has stayed invisible to this project: **it is cited zero times in this corpus**, and a GEO sweep for `KMT2A AND single-cell` returns eighteen other series and not this one [E3]. **Why it is the right resource for this graph specifically:** the three modalities are matched, so the same cell has a transcriptome, an accessibility profile and a DNA-methylation profile — which means questions this project currently asks with bulk data and a caveat (is the fusion's target set bivalent in the cell that matters; is the HSPC-like fraction methylation-distinct; does myeloid enhancer accessibility precede myeloid transcription) become within-cell questions. The 5 healthy paediatric donors are the control arm that most reanalyses of leukaemia single-cell data lack.
What is in it, and the two caveats that decide whether a given reanalysis is possible
| patients | 25 KMT2A-rearranged ALL |
| of which infants (<1 y at dx) | 18 |
| of which older | 7 |
| healthy paediatric donors | 5 |
| paired longitudinal (lineage switch) | 2 patients |
| modalities | scRNA-seq, scATAC-seq, snmC-seq2 |
| portal | HTAN, https://data.humantumoratlas.org/ |
| code | https://github.com/tanlabcode/KMT2Ar_Paper |
Caveat one, and it is the important one for this project: the partner gene mix is not stated in the abstract. 'KMT2A-rearranged' spans KMT2A::AFF1, ::MLLT1, ::MLLT3 and others, and this graph's central question is index-fusion-specific. Whether the 18 infants are predominantly KMT2A::AFF1 — which the epidemiology would predict, since t(4;11) is the commonest rearrangement in infant B-ALL — has to be read off the sample metadata before any index-fusion claim is made from it. Do not assume it.
Caveat two: HTAN access is not necessarily open. Level-1/2 human sequencing data in HTAN is typically controlled-access via dbGaP, with processed levels open. A reanalysis that needs per-cell fusion-spanning reads may require an access application; a reanalysis that needs cluster-level expression or accessibility matrices probably does not. Check the access tier before scoping the work, because this is the difference between a two-week analysis and a three-month one.
The questions this graph already has, that this dataset can answer
Each of these is an open question written elsewhere in this corpus, and each becomes tractable here:
- Do the myeloid-primed diagnostic blasts carry the fusion? Scoring fusion-spanning reads per cell separates 'selection of a leukaemic clone' from 'outgrowth of normal myelopoiesis after B-cell ablation'. This decides between the two readings of
mechanism-myeloid-primed-blasts-preexist-at-kmt2a-r-b-all-diagnosis. - Is the HSPC-like fraction partner-specific? With partner annotation, the KMT2A::AFF1 patients can be separated from the myeloid-partner patients and the fraction compared — which is the single biggest limitation on
cell-state-hspc-like-blast-fraction. - Is methylation or accessibility the earlier signal of lineage priming? snmC-seq2 and scATAC on the same cells is the only design that can order them.
- Do the 5 healthy paediatric donors contain the normal counterpart of the arrested state? This corpus's cell-type entries (
cell-type-pro-b,cell-type-lmpp,cell-type-mpp) are mostly sourced from adult or fetal atlases; a paediatric normal reference measured on the same platform removes a batch confound that every cross-atlas comparison currently carries.
A fifth that the dataset cannot answer and should not be asked to: nothing in it is under menin inhibition or under DOT1L inhibition. Every drug question needs a different resource.
Where it sits among the alternatives
| resource | what it is | accession | in this corpus? |
|---|---|---|---|
| this one | 25 KMT2A-r ALL, 3 matched single-cell modalities | HTAN portal | no |
| GSE159608 | automated single-cell CUT&Tag, MLL leukaemia, 419 samples | GEO | paper cited 12x, accession never named |
| GSE251658 | KMT2A::AFF1 ontogeny-specific induction, CD24+ preleukaemic state | GEO | paper cited 3x, accession never named |
| GSE297908 / GSE283157 etc. | enhancer heterogeneity in ALL, incl. snATAC-seq and CUT&Tag | GEO | paper cited 2x, accessions never named |
| GSE132396 / GSE130142 | epigenetic regulators direct lineage switching in MLL/AF4 | GEO | paper cited 11x, accessions never named |
The pattern in the right-hand column is itself a finding. This project cites the papers and does not record the accessions, so every one of these reanalyses has to be rediscovered from scratch by whoever wants it. A dataset entry per resource, with the accession in a structured field, is the cheap fix and this beat writes two of them.
| repository | Human Tumor Atlas Network (HTAN) |
|---|---|
| portal url | https://data.humantumoratlas.org/ |
| code url | https://github.com/tanlabcode/KMT2Ar_Paper |
| n patients | 25 |
| n infants under 1y | 18 |
| n older than 1y | 7 |
| n healthy paediatric donors | 5 |
| n paired lineage switch patients | 2 |
| disease | KMT2A-rearranged acute lymphoblastic leukaemia |
| fusion partner breakdown | NOT STATED in the sources read; must be taken from supplemental Table 1 or portal metadata before any KMT2A::AFF1-specific claim |
| access tier | unverified by this beat; HTAN raw human sequence is typically controlled-access, processed levels typically open |
| source pmid | 34864916 |
| cited in this corpus | 0 |
| status | planned |
| result summary | Resource inventory, not a computation. Locates and characterises the 25-patient KMT2A-r ALL single-cell multiome (scRNA + scATAC + snmC-seq2, 18 infants, 5 healthy paediatric donors, 2 paired lineage-switch patients) as the deposited resource this project could reanalyse. It is on the HTAN portal, not GEO, and is cited zero times in this corpus. |
assays
[ "scRNA-seq", "scATAC-seq", "snmC-seq2" ]
inputs
[ "https://data.humantumoratlas.org/", "https://github.com/tanlabcode/KMT2Ar_Paper", "PMID 34864916 supplemental Table 1" ]
- A reanalysis of somebody else's single-cell atlas inherits their cell-type labels, their doublet filter and their ambient-RNA correction, and each of those is a modelling choice that can create or destroy a minority population. The HSPC-like fraction is a minority population. Any reanalysis that reports a different fraction size than the paper should be assumed to be measuring the pipeline until proven otherwise.
- Building a project claim on a controlled-access resource creates a reproducibility asymmetry: the lab can check it and readers cannot. Where an open GEO series can answer the same question it should be preferred, and where it cannot, the entry should say that the check is not independently repeatable.
- What is the fusion-partner breakdown across the 25 patients, and how many are KMT2A::AFF1? Every index-fusion claim from this dataset depends on it and it is one table lookup.
- Which HTAN levels are open-access and which need a dbGaP application? Scoping any reanalysis needs this answer first.
- Do the two lineage-switch patients have all three modalities at both timepoints, or only the transcriptome at the later one? The pre/post design is the most valuable part of the resource and also the most likely to be incomplete.
- Is there an equivalent resource under a menin inhibitor? Nothing in this dataset is drug-treated, and this corpus's absence entry on menin-inhibitor occupancy datasets in an infant system remains open.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
The data are deposited at the Human Tumor Atlas Network data portal with HTAN parent-ID to sample-ID linkage in supplemental Table 1, and analysis scripts at github.com/tanlabcode/KMT2Ar_Paper.
Single-cell multiomics reveals increased plasticity, resistant populations, and stem-cell-like blasts in KMT2A-rearranged leukemia — Data Availability Statement · Chen C, Yu W, Alikarami F, et al. (senior author Tan K) · Blood 2022What this evidence cannot show
Read from the PMC full text's Data Availability Statement, quoted verbatim in the entry. I did NOT log in to the HTAN portal or attempt a download, so the access tier, the exact file inventory and whether every modality is present at every level are UNVERIFIED by me. The statement names no dbGaP phs accession; HTAN's general practice of controlled access for raw human sequence is my inference from the consortium's policy and is flagged as such in the body, not asserted as a property of this deposit.
Evidence that argues against this (1)
- A resource that needs a data-access committee is not equivalent to a GEO series for a small lab's purposes. GSE159608 (single-cell CUT&Tag in MLL leukaemia, 419 samples) is openly downloadable today and covers the epigenetic axis, so it is the lower-friction starting point even though it has fewer patients and no transcriptome. Janssens DH et al. 2021
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The cohort is 25 KMT2A-rearranged ALL patients (18 diagnosed <1 year, 7 older) plus 5 healthy paediatric donors, profiled by scRNA-seq, scATAC-seq and snmC-seq2, with 2 patients contributing paired longitudinal samples across a lineage switch.
Single-cell multiomics reveals increased plasticity, resistant populations, and stem-cell-like blasts in KMT2A-rearranged leukemia · Chen C, Yu W, Alikarami F, et al. (senior author Tan K) · Blood 2022What this evidence cannot show
Cohort composition read from the PMC full text, not from the sample manifest itself. The per-patient FUSION PARTNER breakdown is not stated in the abstract or the sections I read, and it is the single field this project most needs; treat it as unknown until read from supplemental Table 1 or the portal metadata. Not every patient necessarily has all three modalities.
Evidence that argues against this (1)
- Cohort composition read from a paper's prose is not the same as cohort composition read from the deposited sample manifest, and the two disagree often enough that this project has a rule about it. Nothing in this entry's cohort description has been checked against the HTAN metadata itself, so the per-modality and per-patient completeness could be materially worse than the headline numbers suggest. Human Tumor Atlas Network 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
This resource is absent from GEO and from this corpus: a GEO gds search for KMT2A single-cell series returns 19 records, none of them this one, and a structural scan of all 2,603 node files finds zero citations of PMID 34864916.
NCBI GEO DataSets esearch and analyses/what-the-citation-index-cannot-see/run.py, both run 2026-09-23 · Q139 · NCBI GEO / Iqbal Lab oncofusion knowledge graph repository 2026What this evidence cannot show
GEO queries as executed, with translations: 'KMT2A[All Fields] AND "expression profiling by high throughput sequencing"[DataSet Type] AND single-cell[All Fields]' -> 19; '(MLL-AF4[All Fields] OR "KMT2A-AFF1"[All Fields] OR "t(4;11)"[All Fields]) AND single-cell[All Fields]' -> 19; 'infant[All Fields] AND leukemia[All Fields] AND "single cell"[All Fields]' -> 30. GEO's [All Fields] is not [tiab] and the searches are of series metadata, not papers. A dataset not in GEO is not thereby unavailable — which is the whole point of this entry.
Evidence that argues against this (1)
- 'Absent from GEO' is a statement about where the data lives, not about whether it is findable or usable, and a GEO-shaped search strategy is the reason it looked absent. The generalisable risk runs the other way too: this project's dataset discovery is GEO-centric, so every resource deposited to HTAN, EGA, dbGaP, Synapse, cellxgene or a lab portal is systematically invisible to it, and the size of that blind spot has never been measured. Q139 2026
the deposited resource behind the HSPC-like observation
the deposited resource behind the pre-existing myeloid-primed blast observation, including both lineage-switch patients
- related to →KMT2A-rearranged acute lymphoblastic leukaemia
25 KMT2A-rearranged ALL patients, 18 of them infants
the resource that supplies the baseline fraction and the signature, and whose lack of a drug arm is why this must be an experiment rather than a reanalysis