Bounded, with every archive named: H3K9me1 and H3K9me2 in a CD34+ progenitor are absent from BLUEPRINT, EGA, Roadmap/Human Epigenome Atlas, the IHEC Data Portal and ENCODE - 49,663 IHEC datasets carry 178 assay labels and H3K9me2 is not one of them - and the absence is a consortium artefact, because GEO has had the CD34+ H3K9me2 data since 2017
also called the archives do not carry H3K9me1/2 in blood progenitors · why six core marks is the whole story · the bounded absence, and the dataset that escapes it
The big public epigenome projects only ever measured six standard chemical tags on DNA-packaging proteins, and the two tags this project cares about are not among them. So none of the big archives has them for blood stem cells. A single 2017 paper in a different archive does have one of them, which shows the gap is about what the consortia chose to measure, not about what is knowable.
The same thing, for a scientist
The dispatch predicted a bounded absence and asked for every archive to be named. **The absence holds for the curated consortium archives and is bounded four different ways** [E1, E2, E3]. **It does not hold for the literature**, and the escape is recorded in the same cycle: GEO's GSE71809 carries H3K9me2 in normal human CD34+ bone-marrow progenitors and has been public since 2017-03-20 (`ana-q111-cd34-h3k9me2-over-the-kmt2a-locus-measured`) [E4]. **IHEC Data Portal, all 50 data releases, 49,663 datasets, 178 distinct assay labels.** H3K9me2 **does not appear as a label at all**, under that spelling or any K9-dimethyl spelling. H3K9me1 accounts for **9 datasets in four biosamples** - IMR-90/IMR90, HUVEC, K562 and NHEK keratinocyte. H3K9me3 has 4,574 and H3K9ac 466 [E1]. **BLUEPRINT's own file index** (`current_release/homo_sapiens/20160816.data.index`, 11,278 file rows, 15 distinct `EXPERIMENT_TYPE` values): **zero H3K9me1, zero H3K9me2 rows anywhere in the project**, and the only CD34-containing `CELL_TYPE` string in the entire ChIP index is `CD34-NEGATIVE, CD41-positive, CD42-positive megakaryocyte cell`. **BLUEPRINT's current release carries no CD34-POSITIVE histone ChIP-seq of any mark** - its CD34-positive samples are WGBS and RNA only [E2]. **EGA, enumerated rather than searched**, because its public metadata API silently ignores `query=`, `search=` and `title=`: **all 10,649 study records pulled by offset paging, of which 19 mention any H3K9 species and ZERO mention H3K9me1 or H3K9me2** [E3]. **ENCODE, queried directly and confirming the dispatch's own figures exactly**: H3K9me1 = **4** experiments (IMR-90, K562, endothelial cell of umbilical vein, keratinocyte), H3K9me2 = **19**, no CD34+ in either. H3K9me3 = 480, including **6 in CD34-positive common myeloid progenitor** [E4]. **Roadmap / Human Epigenome Atlas CD34+ panels** (`Mobilized_CD34_Primary_Cells`, `CD34_Primary_Cells`, `CD34_Cultured_Cells`, `common myeloid progenitor, CD34-positive`) carry H3K4me1/me3, H3K27ac/me3, H3K36me3, H3K9me3, chromatin accessibility, RRBS and WGBS - **never H3K9me1 or H3K9me2** [E1]. Same for CEEHRC, UCLA_ChromImpute, UCLA_ChromHMM and UCL_Hygeia CD34+ entries [E1].
The cause, which makes this a predictable absence rather than an accident
IHEC's reference-epigenome standard is six core marks: H3K4me1, H3K4me3, H3K9me3, H3K27me3, H3K36me3, H3K27ac, plus input, WGBS and RNA. Two EGA study records state it in those words - the CEEHRC/CEMT reference-epigenome study EGAS00001000552 lists exactly that panel, and the DEEP August 2016 release EGAS00001001937 lists the same six 3. H3K9me1 and H3K9me2 are excluded by design, in every consortium that adopted the standard.
So an absence search over consortium archives for these two marks is bounded by a protocol decision made once, around 2010, and no amount of widening the query inside those archives can reach past it. That is a different kind of bound from a literature zero: the instrument was never pointed at the question, in any tissue, and the four H3K9me1 experiments that do exist are ENCODE's own from the Broad, not consortium reference epigenomes.
The practical consequence for this graph. Any future question of the form "is mark X present in cell Y" where X is outside the six core marks should go to GEO first and the consortium portals second. This project has had it the other way round, and exp-h3k9me3-and-ctcf-at-the-kmt2a-anchor-in-a-cd34-progenitor names ENCODE and BLUEPRINT as the cheap first step for exactly this question - correctly identifying the step, and pointing it at the two archives least likely to have the answer.
Four traps hit while bounding this, each of which returns a plausible wrong answer
1. CD34 matches CD34-negative. A grep for CD34 in BLUEPRINT's index returns 142 ChIP rows, all of them the CD34-NEGATIVE CD41+/CD42+ megakaryocyte. Counting those as CD34+ progenitor data would have reported that BLUEPRINT holds H3K9me3, H3K4me1, H3K4me3, H3K36me3, H3K27ac and H3K27me3 in CD34+ cells, which is the opposite of the truth 2. Match the sign, not the antigen.
2. EGA's metadata API ignores filter parameters without error. ?query=H3K9me2, ?search=H3K9me2 and ?title=H3K9me2 all return the same unfiltered first page beginning EGAD00000000001 WTCCC1 1958BC control dataset. A script that trusted the parameter would report whatever the first page contains and it would read as a result. The only safe route is to page the whole collection by offset and filter locally, which is what was done - 10,649 studies 3. (The collection also begins with a literal DUMMY_STUDY record whose title is q; any count taken from row 1 is off by one and semantically wrong.)
3. An aggregator portal bounds the aggregator, not its members. ENCODE has 19 H3K9me2 experiments; the IHEC Data Portal, which ingests ENCODE, shows zero under any K9-dimethyl label across all 50 releases. So the portal's completeness cannot be assumed for a non-core assay, and the member archives must be queried separately - which is why ENCODE's own API and BLUEPRINT's own file index are both cited here rather than the portal alone 14.
4. A consortium label is not a cell. The IHEC portal's CD34+ strings include common myeloid progenitor, CD34-positive, CD34-positive, CD38-negative hematopoietic stem cell, CD34-positive, CD38-positive common myeloid progenitor OR CD34-positive, CD38-positive common lymphoid progenitor (a disjunction, in a single label), Mobilized_CD34_Primary_Cells, CD34_Cultured_Cells and CD34Mobilized. Six strings, at least four different biological objects, and one of them is explicitly ambiguous between two lineages. Any count of "CD34+ datasets" depends on which strings are accepted 1.
And a correction to the premise this beat was given
The dispatch stated that ENCODE's H3K9me2 holdings are 19 experiments and "none haematopoietic". The count is exactly right; the characterisation is not. Nine of the 19 are haematopoietic-lineage cell lines: DOHH2, Karpas-422, OCI-LY1, OCI-LY3, OCI-LY7 and SU-DHL-6 (B-cell lymphoma), Loucy (T-ALL), and MM.1S and KMS-11 (multiple myeloma) 4. The rest are HCT116, MCF-7, PC-3, PC-9, SK-N-SH, GM23248, GM23338 and three in-vitro-differentiated populations.
The claim that survives is narrower and is the one that matters: no ENCODE H3K9me2 experiment is in a CD34+ progenitor or in any primary haematopoietic sample - all nine haematopoietic ones are transformed lines, and none is KMT2A-rearranged. But Loucy is a leukaemia line and its H3K9me2 map is public, which makes "H3K9me2 has never been measured in a leukaemia cell" false, and this graph should not write that sentence. The nine were not read here and are the obvious next set.
| absence class | measured-not-searched |
|---|---|
| absence class note | measured-not-searched for H3K9me2: the CD34+ data existed in GEO for nine years and the search space this project used (ENCODE plus BLUEPRINT) excluded it by protocol. For H3K9me1 in a CD34+ progenitor the class is never-measured in every archive queried - no such dataset exists anywhere checked, so the two marks sit in different classes inside one entry and the field carries the weaker (more closable) of the two. |
| what is absent | any H3K9me1 or H3K9me2 ChIP-seq dataset in a CD34-positive haematopoietic progenitor in BLUEPRINT, EGA study metadata, Roadmap/Human Epigenome Atlas, the IHEC Data Portal federation, or ENCODE |
| access verdict | The question of controlled versus open access does not arise for these two marks, because the datasets do not exist in these archives. Where the marks that DO exist are concerned: BLUEPRINT's processed files are OPENLY downloadable from the EBI FTP and its raw sequence is CONTROLLED in EGA; ENCODE and Roadmap are open; the IHEC portal serves open metadata and links. And the dataset that answers the question - GEO GSE71809 - is fully OPEN, no DAC, no data-access agreement. |
| why the absence exists | IHEC's reference-epigenome standard is six core marks (H3K4me1, H3K4me3, H3K9me3, H3K27me3, H3K36me3, H3K27ac) plus input, DNA methylation and RNA. H3K9me1 and H3K9me2 are outside it, so every consortium that adopted the standard excludes them by protocol - stated verbatim in EGA study records EGAS00001000552 and EGAS00001001937. |
| what closes it | For H3K9me2 in a CD34+ progenitor: nothing - it is CLOSED, by GEO GSE71809, at domain resolution, and at coverage resolution by aligning that series' 31 SRA libraries. For H3K9me1 in a CD34+ progenitor: a new ChIP, because no archive queried holds one in any haematopoietic cell other than K562. |
| cost tier | low for the H3K9me2 half (already done); med for a coverage-resolution re-run from SRA; high for new H3K9me1 wet work |
| archive check run | true |
| archive queried | IHEC Data Portal metadata API (all 50 releases); BLUEPRINT DCC FTP file index; EGA public metadata API (all studies, offset-paged); ENCODE portal search API; NCBI GEO via elink pubmed->gds and acc.cgi SOFT records |
| localization | not applicable - this entry concerns archive holdings and asserts nothing about a cellular location |
| subnuclear localization | not applicable - an archive inventory has no subnuclear coordinate and none is claimed |
| cell type | the cell type SOUGHT is the CD34-positive human haematopoietic progenitor; the cell types FOUND carrying H3K9me1 anywhere in these archives are IMR-90 lung fibroblast, HUVEC, NHEK keratinocyte and K562, and for H3K9me2 in ENCODE a set of nine transformed haematopoietic-lineage lines plus eight non-haematopoietic lines and three in-vitro-differentiated populations |
| cell state | not applicable to an archive inventory; the datasets found are unperturbed steady-state cultures or primary cells except where a consortium recorded a treatment |
| epigenetic readout | the inventory IS the readout: which H3K9 methylation states have been assayed by ChIP-seq in which human cells across five archives, with H3K9me3 and H3K9ac as positive controls demonstrating the query can return large numbers |
| shown for fusion | Not shown for any fusion. No H3K9me1 or H3K9me2 dataset in any archive queried is in a KMT2A-rearranged cell, and none of ENCODE's nine haematopoietic-lineage H3K9me2 lines is annotated as carrying a KMT2A fusion. |
archives queried with counts
{
"IHEC Data Portal (all 50 releases; aggregates BLUEPRINT, ENCODE, Roadmap, DEEP, CEEHRC McGill/CEMT, CREST, KNIH, GIS, EpiHK, UCLA_ChromImpute, UCLA_ChromHMM, UCL_Hygeia)": {
"datasets": 49663,
"distinct_assay_labels": 178,
"H3K9me2": 0,
"H3K9me1": 9,
"H3K9me1_biosamples": [
"IMR-90/IMR90",
"HUVEC",
"K562",
"NHEK keratinocyte"
],
"H3K9me3": 4574,
"H3K9ac": 466,
"CD34_positive_H3K9me1_or_me2": 0
},
"BLUEPRINT DCC current_release file index": {
"file_rows": 11278,
"distinct_experiment_types": 15,
"H3K9me1": 0,
"H3K9me2": 0,
"H3K9me3": 944,
"only_CD34_celltype_string": "CD34-NEGATIVE, CD41-positive, CD42-positive megakaryocyte cell",
"CD34_positive_histone_chip_of_any_mark": 0
},
"EGA (study metadata, fully enumerated)": {
"studies_pulled": 10649,
"real_studies": 10648,
"mentioning_any_H3K9_species": 19,
"mentioning_H3K9me1": 0,
"mentioning_H3K9me2": 0
},
"ENCODE (portal search API, queried directly)": {
"H3K9me1_experiments": 4,
"H3K9me2_experiments": 19,
"H3K9me2_released": 17,
"H3K9me3_experiments": 480,
"H3K9me3_in_CD34_positive_common_myeloid_progenitor": 6,
"H3K9me1_or_me2_in_CD34_positive": 0,
"H3K9me2_in_haematopoietic_lineage_cell_lines": 9
},
"Roadmap / Human Epigenome Atlas CD34+ panels (via the IHEC portal)": {
"labels": [
"Mobilized_CD34_Primary_Cells",
"CD34_Primary_Cells",
"CD34_Cultured_Cells",
"common myeloid progenitor, CD34-positive"
],
"marks_present": [
"H3K4me1",
"H3K4me3",
"H3K27ac",
"H3K27me3",
"H3K36me3",
"H3K9me3",
"Chromatin_Acc",
"Input",
"RRBS",
"WGB-Seq",
"Footprinting",
"mRNA-Seq"
],
"H3K9me1_or_me2": 0
}
}king verification of the encode counts
{
"verified": "king re-queried the ENCODE API independently on 2026-09-22",
"H3K9me1": "4 experiments - MATCHES (K562, endothelial cell of umbilical vein, keratinocyte, IMR-90); 1 of 4 haematopoietic",
"H3K9me2": "19 experiments and 9 haematopoietic-lineage lines - BOTH MATCH EXACTLY, and the nine named lines are the nine returned",
"BUT THE COUNT DEPENDS ON A STATUS FILTER THAT WAS NOT STATED": "The 19 includes one REVOKED and one ARCHIVED experiment, and both are haematopoietic: KMS-11 (ENCSR900DFO) is revoked and OCI-LY3 (ENCSR856ACW) is archived. Restricting to status=released gives 17 experiments and 7 haematopoietic lines. So 9-of-19 and 7-of-17 are both correct statements about different denominators, and the haematopoietic fraction is precisely the part the filter moves - two of the nine are data you cannot simply download.",
"what_this_does_not_change": "Loucy (ENCSR186EAD, T-ALL) is RELEASED, so the load-bearing point survives intact: a leukaemia line H3K9me2 map is public, and this graph must not write that H3K9me2 has never been measured in a leukaemia cell. Checked: that sentence appears nowhere in the corpus.",
"the_lesson": "A count quoted without its status filter is a count nobody can reproduce. The king supplied this beat with the figure 19 and the characterisation none haematopoietic; the count was right and the characterisation was invented from never opening the list - which is this project standing lesson that a control count is a list nobody opens, recurring on the numbers I hand to agents rather than on the ones they hand back."
}- A bounded absence that names five archives reads as authoritative, and this one was wrong about the literature within the same afternoon it was written. The specific hazard for an intervention: if this graph concludes 'the resting mark state in a progenitor is unmeasurable', it will stop asking, and a G9a-inhibitor prophylaxis argument will keep resting on retinal pigment epithelium. The absence is real for four archives and false for the question, and the entry has to be read that way round.
- The correction to the dispatch's 'none haematopoietic' could be over-applied in the other direction. Nine public H3K9me2 maps in lymphoma, T-ALL and myeloma lines do not make the progenitor question answerable, and treating them as though they did would repeat the substitution this project already made when it used K562 for a CD34+ cell and said so.
- What do the nine ENCODE H3K9me2 experiments in haematopoietic-lineage lines say at the KMT2A locus? Loucy (T-ALL), DOHH2, Karpas-422, OCI-LY1/3/7, SU-DHL-6, MM.1S and KMS-11 are all readable public tracks and none has been opened by this project. They are transformed lines and not progenitors, but they are blood cells, and the K562 element-scale disagreement in sign (0.877 against 1.155) is exactly the kind of ambiguity more cell types would resolve.
- How many other non-core marks has this project bounded against ENCODE and BLUEPRINT and therefore under-searched? The six-mark core panel excludes H3K9me1, H3K9me2, H3K9me1/2, H3K79me2, H4K20me1 and every arginine mark, several of which this graph holds entries on. The check is one assay-label tabulation per mark against the 178-label list, and it is now cached.
- Is there a CD34+ H3K9me1 dataset anywhere outside these five archives - in GEO, ArrayExpress, DDBJ or a supplementary deposit? GEO was queried only via the one paper that answered the H3K9me2 half; a systematic GEO DataSets query on H3K9me1 crossed with CD34 was not run and is the obvious next step, because GEO is precisely where the escape from this absence was found.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Across all 50 data releases of the IHEC Data Portal - 49,663 datasets from BLUEPRINT, ENCODE, Roadmap, DEEP, CEEHRC (McGill/CEMT/other), CREST, KNIH, GIS, EpiHK, UCLA_ChromImpute, UCLA_ChromHMM and UCL_Hygeia - there are 178 distinct assay labels; H3K9me2 is not among them under any spelling, H3K9me1 accounts for 9 datasets in four biosamples (IMR-90, HUVEC, K562, NHEK keratinocyte), H3K9me3 for 4,574 and H3K9ac for 466; and every CD34-positive progenitor label in the portal - across BLUEPRINT, Roadmap, CEEHRC, UCLA_ChromImpute, UCLA_ChromHMM and UCL_Hygeia - carries only H3K4me1, H3K4me3, H3K27ac, H3K27me3, H3K36me3, H3K9me3, input, chromatin accessibility, WGBS/RRBS and RNA.
IHEC Data Portal metadata API: getReleases.py enumerated 50 releases and getDataHub.py?data_release_id=N pulled every dataset in each, 2026-09-22; assay-label and cell-type tabulation reproducible from the cached hubs · Bujold D, Morais DAL, Gauthier C, Côté C, Caron M, Kwan T, Chen KC, Laperle J, Markovits AN, Pastinen T, Caron B, Veilleux A, Jacques PÉ, Bourque G · Cell Systems 3:496-499.e2 (the IHEC Data Portal) 2016What this evidence cannot show
The portal is an AGGREGATOR and its coverage of any member consortium is incomplete for non-core assays - demonstrated here, since ENCODE's 19 H3K9me2 experiments appear in ENCODE's own API and in none of the portal's 50 releases. So this bounds the portal, not the consortia, and it is cited alongside BLUEPRINT's own index and ENCODE's own API for that reason. Assay labels were read from the `ihec_data_portal.assay` field; a dataset with a missing or non-standard label would be invisible to the tabulation, and 'NA' is itself one of the 178 labels. The portal mixes human and mouse releases; the H3K9me1 and CD34 tabulations above are human only.
Evidence that argues against this (1)
- A zero in an assay-label tabulation is a zero in a controlled vocabulary, and a mark measured under a non-standard label would be missed. The portal's own label list contains 'H3K9_14ac', 'H2A_Zac' and 'Chipmentation H3K9me3', so the vocabulary is not tidy, and a hypothetical 'H3K9diMe' or 'G9a mark' label would not have matched the regex used. The check that makes this safe is the positive control in the same tabulation: H3K9me3 returns 4,574 and H3K9ac 466 through the identical code path, so the instrument can return a large number for a K9 species. International Human Epigenome Consortium Data Portal 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
BLUEPRINT's own current_release file index (20160816.data.index, 11,278 file rows, 52 columns) contains 15 distinct EXPERIMENT_TYPE values - total-RNA-Seq, H3K27ac, DNA Methylation, H3K4me3, H3K4me1, H3K36me3, H3K9me3, H3K27me3, mRNA-Seq, Chromatin Accessibility, Input, mRNA-seq, H3K9/14ac, flRNA-seq and H2A.Zac - with zero H3K9me1 and zero H3K9me2 rows anywhere in the project; and of BLUEPRINT's 60 distinct CELL_TYPE strings the only one containing 'CD34' is 'CD34-negative, CD41-positive, CD42-positive megakaryocyte cell', so the release carries no CD34-POSITIVE histone ChIP-seq of any mark, its CD34-positive and stem/progenitor samples being WGBS and RNA only.
BLUEPRINT Data Coordination Centre, current_release/homo_sapiens/20160816.data.index (9.3 MB, 11,278 file rows), retrieved 2026-09-22 from the EMBL-EBI public FTP over HTTPS · BLUEPRINT Epigenome Project / EMBL-EBI 2016What this evidence cannot show
This is the 2016-08-16 release, which is BLUEPRINT's LAST - the directory named current_release points at it and the project has not released since (the releases directory lists 20130301 through 20160816 and nothing later). So it bounds BLUEPRINT as it finally stands, which is the right bound, but it cannot be refreshed. The index lists FILES rather than experiments, so row counts are not experiment counts; the EXPERIMENT_TYPE vocabulary is what is being counted and that is a per-file field. BLUEPRINT's raw sequence is in EGA under controlled access (studies ERP001663 and others); the processed files this index lists are openly downloadable from the EBI FTP, so for the marks that DO exist the access question is settled and open - it is simply moot for H3K9me1/2, which do not exist here.
Evidence that argues against this (1)
- The IHEC portal lists BLUEPRINT CD34-POSITIVE samples that this index does not contain - 'CD34-positive, CD38-negative multipotent progenitor cell', 'CD34-positive, CD38-positive common myeloid progenitor' and several 'from peripheral blood' progenitor strings - carrying WGB-Seq and RNA. So the two sources disagree about BLUEPRINT's sample inventory, presumably because the portal ingested a later or differently-curated cut. Neither source assigns H3K9me1 or H3K9me2 to any of them, so the conclusion is unaffected, but the disagreement means neither source should be quoted as BLUEPRINT's complete inventory. International Human Epigenome Consortium Data Portal 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
All 10,649 study records in the European Genome-phenome Archive were pulled by offset paging of the public metadata API (its query, search and title parameters are silently ignored and return the unfiltered first page); 19 study records mention any H3K9 species anywhere in their metadata and NONE mentions H3K9me1 or H3K9me2, the concrete ones naming H3K9me3 or H3K4me3; and the two IHEC reference-epigenome study records in EGA state the core panel explicitly as H3K4me1, H3K4me3, H3K9me3, H3K27me3, H3K36me3 and H3K27ac, which excludes H3K9me1 and H3K9me2 by protocol.
EGA public metadata API https://metadata.ega-archive.org/studies paged by limit=1000 and offset, 10,649 records, retrieved 2026-09-22; CEEHRC/CEMT reference-epigenome study EGAS00001000552 and IHEC DEEP release EGAS00001001937 read for the core-mark panel · European Genome-phenome Archive, EMBL-EBI / CRG 2026What this evidence cannot show
This searches STUDY-level metadata - accession, title, description, abstract, study type - not sample-level or experiment-level records, and not the files themselves. A study whose title and description do not name its marks would be invisible, and that is common: many of the 10,649 have a null description. So this is a bound on what EGA's study metadata SAYS, which is weaker than a bound on what EGA CONTAINS, and it is stated that way. It is cited alongside BLUEPRINT's own file index precisely because the file index reaches the assay level for the one EGA-hosted consortium that matters here. The collection's first record is a literal placeholder, 'DUMMY_STUDY' with the title 'q', so the true study count is 10,648.
Evidence that argues against this (1)
- A study-metadata search cannot see a mark that is only named in a sample record, and EGA's sample records were not enumerated - so this is the weakest of the four bounds and could in principle be wrong. The concrete reason to think it is not: the marks are absent for a PROTOCOL reason that two EGA study records state in their own words, and that reason predicts the zero rather than merely being consistent with it. European Genome-phenome Archive 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Queried directly against the ENCODE portal search API on 2026-09-22, ENCODE holds 4 H3K9me1 histone ChIP-seq experiments (ENCSR580LZO IMR-90, ENCSR000AKW K562, ENCSR000ALE endothelial cell of umbilical vein, ENCSR000ALQ keratinocyte), 19 H3K9me2 experiments and 480 H3K9me3 experiments of which 6 are in CD34-positive common myeloid progenitor (ENCSR128KIH, ENCSR979RHZ, ENCSR179MIQ, ENCSR245ZCU, ENCSR957WQX, ENCSR734ESI); no H3K9me1 or H3K9me2 experiment is in a CD34-positive or any other primary haematopoietic sample, but NINE of the 19 H3K9me2 experiments are in haematopoietic-lineage cell lines - DOHH2, Karpas-422, OCI-LY1, OCI-LY3, OCI-LY7, SU-DHL-6, Loucy, MM.1S and KMS-11 - which corrects the dispatch brief's statement that none is haematopoietic.
ENCODE portal search API, type=Experiment&assay_title=Histone+ChIP-seq&target.label=H3K9me1|H3K9me2|H3K9me3 with biosample_ontology and donor organism fields, retrieved 2026-09-22 · ENCODE Project, encodeproject.org 2026What this evidence cannot show
Counts are of EXPERIMENT records and include non-released statuses: KMS-11's H3K9me2 experiment is 'revoked' and OCI-LY3's is 'archived', so the usable H3K9me2 count is 17 released and the haematopoietic-lineage count is 7 released of 9. 'Haematopoietic-lineage cell line' is a lineage statement and not a claim that any of these is a normal blood cell or a progenitor; all nine are transformed, none is annotated KMT2A-rearranged, and Loucy is T-lineage while the index disease here is B-lineage. The search was scoped to assay_title=Histone ChIP-seq, so a Mint-ChIP or ChIPmentation experiment for the same target under a different assay_title would be excluded.
Evidence that argues against this (1)
- Calling nine of these haematopoietic weakens the absence less than it appears to. None is a progenitor, none is primary, none is KMT2A-rearranged, and the question this absence was built for - the RESTING-STATE mark level in a cell that could become an infant KMT2A-rearranged leukaemia - cannot be answered in a germinal-centre lymphoma line or a myeloma line any more than it can in retinal pigment epithelium. The correction is real and the practical gap is unchanged. ENCODE Project 2026
the consortium absence is real and does not bound the literature - GEO has had the CD34+ H3K9me2 data since 2017
9 datasets in the entire IHEC federation, four biosamples, none haematopoietic except K562; 4 experiments in ENCODE
not one of the IHEC portal's 178 assay labels; zero rows in BLUEPRINT's file index; zero EGA study records name it
that experiment named ENCODE and BLUEPRINT as the cheap first step; both exclude H3K9me1 and H3K9me2 by protocol
the cell type sought; CD34-positive panels in every consortium carry only the six IHEC core marks
the H3K9me2 arm of this design is the wet-work answer to that bounded absence