GSE159608 — 419 automated CUT&Tag libraries across 7 KMT2A-rearranged primary patient samples, 5 lines and a normal CD34+ donor, including KMT2A fusion occupancy, a menin-inhibitor arm in SEM, and three single-cell CUT&Tag libraries

absence-no-menin-inhibitor-occupancy-dataset-in-an-infant-derived-kmt2a-aff1-line states, in its title, that every chromatin measurement under a menin inhibitor is in an adult-derived line, and none is in an infant KMT2A::AFF1 model, and in its body that 'SEM — the canonical infant-derived t(4;11) KMT2A::AFF1 line … appears in the menin-inhibitor literature only as a viability and synergy system, never as a chromatin-profiling system under drug.'

GSE159608 contains SEM treated with 30 µM VTP-50469 for 72 hours, profiled by H3K4me3 CUT&Tag, two replicates, with matched DMSO controls — GSM5139713 and GSM5139714 against GSM5139709 and GSM5139710 1. The same design exists for RS4;11 and for KOPN-8, which is also infant-derived.

The absence entry cites the very paper this data belongs to, at its evidence[2]. The enumeration behind the absence was done by reading methods and figure legends of full texts — a careful and correct method for what it can see — and a deposited sample that a paper does not discuss in its main text is invisible to it. That is the generalisable lesson: for a bounded absence about whether a measurement exists, the GEO sample metadata is a better index than the paper, because depositors annotate arms that the narrative never returns to.

What survives of the absence, and it is not nothing. Under the menin inhibitor GSE159608 profiles H3K4me3 only — no KMT2A occupancy, no menin, no H3K27ac, no RNAPII, no H3K79me2. So the narrower and still-true claim is: no fusion-occupancy measurement under a pharmacological menin inhibitor exists in an infant-derived KMT2A::AFF1 line; the only deposited chromatin readout is H3K4me3 CUT&Tag in SEM and RS4;11, n = 2 replicates each, GSE159608. That is a materially different and much more actionable statement than the one on record, and reanalysing those six libraries is a day's work.

The paper's own framing is that integrating automated and single-cell CUT&Tag 'uncovers lineage heterogeneity within patient samples' 2. That is the epigenetic counterpart of the transcriptomic plasticity result in mechanism-myeloid-primed-blasts-preexist-at-kmt2a-r-b-all-diagnosis: a patient sample that looks like one chromatin state in bulk resolves into more than one when read cell by cell.

The limits are real and should be stated before anyone plans on this. Three libraries, three marks, one platform run (iCell8, 2020-07-23), and the sample annotation is a disjunction — 'sorted CD45+ cells from patient sample or KMT2Ar cell lines' — so which cells are in which library has to be recovered from the per-cell FASTQ identifiers rather than read off the metadata. None of the three single-cell marks is KMT2A itself, so the fusion's own occupancy heterogeneity across cells is not in this dataset. Single-cell CUT&Tag also yields very sparse per-cell coverage, so any per-cell claim needs aggregation over cells and the effective n is much smaller than the cell count suggests.

Seven primary patient samples, each with its fusion partner named in the GEO metadata, spanning AF4, AF9, ENL, AF10 and SEPT6, plus a CD34+ normal donor on the same platform in the same lab 1. This corpus's recurring structural problem — measured in its own partner-provenance audit, where 0 of 24 sampled load-bearing experiments rested on a dedicated human KMT2A::AFF1 experiment — is that mechanism is imported across partners without the import being tested.

GSE159608 is a within-lab, within-platform, partner-resolved panel. It is one of the very few datasets in which the question does this hold for AF4 as well as for AF9 can be asked without a cross-study batch confound. Two of the primary samples are KMT2A-AF4 (ALL-1 and MPAL-2) and one of them is MPAL, which is directly on the lineage-plasticity axis this beat is about.