The KMT2-family beat's own ledger: every query run, every gene dropped, every paper read and not used

All run 2026-09-20 through NCBI E-utilities with querytranslation inspected on every count.

Rule 64 discharged and it mattered. The decisive paper of the whole beat, van Nuland 2013, is unreachable from any KMT2B symbol query - it says MLL2 and frames everything as 'SET1/MLL complexes'. The 32-record KMT2B leukaemia count would have supported a confident, bounded and wrong absence. The complex-form query that found it was (MLL2 OR MLL4) AND menin, not a complex name.

Rule 64's second failure mode also appeared. "COMPASS-like complex" is the ordinary-word contamination case: 37 records, the majority about flowering time in plants.

Subtract-before-report discharged. The repository was grepped before every absence recorded here. Two results: (a) PMID 30315824 - the paper that directly tests whether menin inhibitors work through MLL1/MLL2 - is cited in zero entries in this graph; (b) 85 entries name a KMT2 family member other than KMT2A and none of those genes had an entry.

The dispatch asked for 14-18 entries and warned that a padded roster is worse than a recorded gap. Three candidates were enumerated and not written:

• PAGR1 / PA1. The other MLL3/4-specific subunit, co-equal with PTIP in the quantitative assignment 1 and the FLAG-tagged bait used to purify the complex 4. Dropped because everything this beat could say about it is already said by gene-paxip1, with which it forms a subcomplex, and it has no independent leukaemia literature. What would change this: a PA1 result separable from PTIP's.
• KMT2F / KMT2G (SETD1A / SETD1B). Already written as gene-setd1a and gene-setd1b. Not duplicated. Noted here because the KMT2 numbering includes them and a reader enumerating KMT2A-G will otherwise think two are missing.
• HCFC1 / HCFC2. Named as components of the KMT2B complex by the KMT2E and KMT2B reviews. Dropped because HCFC1 is already recorded in this graph as floor-compressed in DepMap (depleted in 100% of lines, index-versus-comparator difference -0.073), which makes a selectivity claim uninterpretable, and because the KMT2B-specific evidence is review-level only. What would change this: an HCF1 requirement demonstrated for KMT2B and not for KMT2A.

Recorded because a reader would reasonably expect them to be cited.

• PMID 39806204, Loss of Kmt2c or Kmt2d primes urothelium for tumorigenesis and redistributes KMT2A-menin to bivalent promoters. Used in gene-kmt2c but deliberately not promoted: urothelium, no leukaemia, no fusion, no menin inhibitor. It is already the mechanistic backbone of resistance-mll34-utx-tumour-suppressor-failure, and this beat's view is that a tissue-crossing inference is carrying more weight there than it can bear.
• PMID 41970232, The Histone-Lysine N-Methyltransferase (KMT2) Family in Health and Disease (2026). A current family review that would have been a convenient map. Not cited anywhere in this beat: every structural and functional claim here was taken from a primary source, and a 2026 review adds citation surface without adding evidence. What would change this: needing a claim no primary paper states.
• PMID 36598580, Clustered PHD domains in KMT2/MLL proteins are attracted by H3K4me3 and H3 acetylation-rich active promoters and enhancers. Squarely on-beat by title and a genuine family-level chromatin result. Not used because this graph already carries the KMT2A PHD biology in depth (interface-kmt2a-phd3-h3k4me3, subtraction-kmt2a-phd3-bromo-cassette, localization-phd-cluster-1416-1640-is-lost-ten-residues-past-the-junction) and adding a family-level PHD entry without reading the paper in full would be the exact family-attribution error this beat exists to warn about. What would change this: a beat dedicated to reading it.
• PMIDs 32365172 and 31485071, H2B ubiquitylation enhances H3K4 methylation activities of human KMT2 family complexes and Structural basis of nucleosome recognition and modification by MLL methyltransferases. Both are genuine family-wide structural biochemistry and both are on-beat. Not used because mechanism-h2bub-compass-h3k4-arm already occupies that ground and this beat did not read either in full. What would change this: a structural beat.
• PMID 34304711, Loss of KMT2C reprograms the epigenomic landscape in hPSCs resulting in NODAL overexpression and a failure of hemogenic endothelium specification. Haematopoietic development, KMT2C, and directly relevant to this graph's fetal-origin arguments. Not used for space; the fetal-barrier paper carries the same argument with a leukaemia endpoint. What would change this: it is the best candidate for the next beat's KMT2C work and should be read.
• PMID 30936419's non-leukaemia half, and the whole ODLURO syndrome literature (KMT2E germline). Enumerated, judged developmental-neurology rather than leukaemia, and dropped.
• KMT2E-ASNS letter, PMID 28069604. Cited at weak confidence with the caveat that the PubMed record has an empty abstract and the title is the entire claim. It is recorded rather than dropped because a KMT2-family fusion gene in acute leukaemia is on this project's central beat and a future cycle should read the letter.

The standing brief says the literature wins and the agent says so. Two cases.

1. 'Every menin inhibitor in this graph hits two proteins and the graph has written one of them - part of what a menin inhibitor does in a patient may be KMT2B biology rather than fusion biology.' The binding half is right. The therapeutic half has been tested directly and largely fails: Men1 deletion and Mll1;Mll2 co-deletion produce distinct consequences in MLL-AF9 cells, the compounds phenocopy menin loss rather than Mll1;Mll2 loss, and Mll1;Mll2-deficient cells are more sensitive to menin inhibitors, not less 2. What survives is the normal-tissue version, which nobody has tested and which is now hyp-menin-inhibitor-normal-tissue-effect-is-kmt2b.

2. (KMT2B OR MLL2 OR MLL4) AND (leukemia OR leukaemia) returns 258.' True as a count and not attributable to KMT2B. PubMed maps MLL2 to the KMT2D supplementary concept and MLL4 to the KMT2B one, so that query is a KMT2B/KMT2D mixture 3. KMT2B AND (leukemia OR leukaemia) alone returns 32. The KMT2B leukaemia literature is about a quarter the size the brief implied - which makes the gap more striking, not less, because the few papers that exist are decisive.