Four YEATS antagonists exist, four public transcriptome datasets of YEATS inhibition exist, and not one paper or dataset has ever asked what happens to the AFF2- or AFF3-scaffolded complexes — and the analysis needs no wet work

Rules 70-72: a counted absence about a drug class is bounded by the name list, not by the query. The 18 strings tried, each tagged individually as [tiab] (never (A OR B)[tiab], rule 94), run 2026-09-24:

Union: 18 distinct records. Singular returns 0 and plural returns 5 for the same concept, and hyphenated and unhyphenated forms differ (TDI-11055 3, TDI11055 2) — rule 72 firing twice, exactly as documented.

Where the list came from, because that is what the bound rests on. Thirteen strings were built from the four YEATS-antagonist papers this beat surfaced (PMIDs 30288907, 34079898, 36053276, 36645983). GNE-064 was not among them; it came from reading ana-q159-eleven-abstract-level-absences-retested-against-agent-and-model-names-and-none-was-overturned, which had already assembled a four-name ENL-YEATS agent list in an earlier cycle. Adding it changed no zero, but a name list assembled only from the papers you happened to find is not a bound — the corpus's own prior lists have to be read first.

The OR group audits itself and it must be read back. PubMed silently deletes non-matching quoted phrases from a group; the quotedphrasesnotfound list names the five zero-records strings above, so the group ran as 13 of 18 terms. Every deleted term returns 0 on its own, so no recall was lost — but that had to be checked, not assumed, and the querytranslation is the only place it is visible.

Query shape: (<13 surviving name strings, each [tiab]>) AND <target>[tiab].

And the bare-operand controls, so the zeros cannot be blamed on either side: AFF2[tiab] = 126, AFF3[tiab] = 123, YEATS[tiab] = 233, "SEC-L2"[tiab] = 2, "SEC-L3"[tiab] = 5.

Every zero came back with a non-empty querytranslation and outputmessages: ["No items found."], which is the genuine-zero signature; none was an instrument failure. Scope: titles and abstracts only. A supplementary figure showing AFF3 protein unchanged under SGC-iMLLT would be invisible to all of this (rule 80), and no full text was searched for any of the nine targets.

The literature absence is real and the data absence is not. Verified 2026-09-24 through the SOFT brief record and attributed by !Series_pubmed_id, never by topic:

The fourth row is the one that matters. MV4;11 carries KMT2A::AFF1, and somebody has already measured scaffold occupancy under a YEATS inhibitor — for AFF1 and AFF4. The missing arm is AFF2 and AFF3, which nobody has ever ChIPped in any cell.

The analysis that closes most of this, with no wet work:

1. Take the AFF2-responsive and AFF3-responsive gene sets from GSE34097 (ana-q187-sec-family-target-set-overlap-measured reconstructs their sizes; the identities need the deposit).
2. Ask whether they move in the GSE185091 / GSE162592 YEATS-inhibitor RNA-seq, against the AFF4-responsive set as the internal comparator and a size-matched, expression-matched random set as the null.
3. If the sibling sets move as much as the AFF4 set, scaffold-blindness has a measurable transcriptional footprint. If they do not, it does not, and mechanism-q187-target-specificity-is-a-scaffold-property-while-the-reader-is-shared should be downgraded to a composition curiosity.

Two things this cannot settle, which is why the absence is only mostly closable by reanalysis. The gene sets come from HEK293T and the drug data from AML lines, so a null result is confounded with cell type. And nothing in any of these deposits touches imprinting, allele-specific expression or neural tissue, which is where risk-q187-yeats-inhibition-touches-an-imprinting-complex-in-an-infant places the liability.

• PMID 22895430, Luo, Lin and Shilatifard, Nat Rev Mol Cell Biol 2012, The super elongation complex (SEC) family in transcriptional control. On topic, and one of only five records naming SEC-L2 or SEC-L3. Not cited as evidence anywhere in this beat because it is a review by the same authors of the same primary paper, and the brief's quality bar prefers the primary. Recorded so a later cycle knows it was seen and why it was passed over.
• PMID 39327672, ENL mutation and AML: a new model that reveals oncogenic condensate's function in leukemogenesis, Mol Oncol 2025. Surfaced by the "ENL degrader" string. Not used: it concerns ENL hotspot mutations and condensates, a different axis from complex membership, and this graph already holds the condensate-assay caveats.
• PMID 36645983, benzimidazole ENL inhibitors, Eur J Med Chem 2023. Surfaced by the name list. Not used: a medicinal-chemistry series with no complex-membership or paralogue content; it would only add a fifth name to the name list, which it now has.
• "SEC-like 2"[tiab], "SEC-like 3"[tiab] and "super elongation complex-like"[tiab] all return 0 with quotedphrasesnotfound. The spelled-out names are not in the index at all; only the hyphenated abbreviations are. A reader repeating this search must use SEC-L2 and SEC-L3.