Enumeration of candidate cooperating lesions in KMT2A-rearranged leukaemia - what was written, what was recorded as a gap, and what was read and rejected

The last row is the point of this entry. A beat that returns nine entries out of ten candidates and explains the tenth is more useful than one that returns ten.

Queries run on 2026-09-20, with querytranslation inspected and every title read:

The complex form was run deliberately, because the project has twice written a wrong counted absence on a complex subunit whose literature is framed at complex level. It returned more records and no co-mutation studies: the seventeen are dependency, interaction and inhibitor papers - BRG1 inhibition in Ph-like B-ALL (40672274), BAF/PBAF subunit roles in MLL-AF9 leukaemogenesis (35022500), Baf200 in normal haematopoiesis and leukaemogenesis (29482581), a SWI/SNF complex containing an MLL translocation partner (12665591), and ALL-1 SET-domain interaction with INI1 (9539705).

And the one squarely on-topic paper is on-topic for a different beat. PMID 26571505, 'SWI/SNF Subunits SMARCA4, SMARCD2 and DPF2 Collaborate in MLL-Rearranged Leukaemia Maintenance' 2, reports a selective requirement for these three subunits for leukaemic cell expansion and self-renewal, with overlapping gene-expression effects and a requirement for haematopoietic stem-cell-associated gene expression - and, contrary to earlier mouse work, no requirement for c-MYC-regulated genes. That is a vulnerability finding, not a cooperating lesion.

Correction owed outside this folder. A repository grep on 2026-09-20 found PMID 26571505 cited in no entry of this graph, while SMARCA4 is named in 22 node files. A dependency established in MLL-rearranged leukaemia maintenance, uncited in a graph with dedicated vulnerabilities and interactome folders, is a gap the coordinator should route rather than something this beat should absorb.

Recorded because a reader would reasonably expect them to be cited and they are not.

• PMID 8824725, Distribution of TP53 mutations among acute leukemias with MLL rearrangements (1996). Used, but not as evidence for a frequency. It reports TP53 mutation in 9/29 MLL-rearranged leukaemias including 3/4 infants, which no modern cohort reproduces. Recorded as an open dispute in gene-tp53 rather than as a finding or as a superseded result. What would change the decision: a modern whole-genome study of infant KMT2A-rearranged AML, which does not exist.
• PMID 32154134, Mutations predictive of hyperactive Ras signaling correlate with inferior survival across high-risk pediatric acute leukemia. Cited at weak only. n = 43, single institution, ALL and AML pooled, cohort selected for high-risk features, and NRAS, NF1 and PTPN11 pooled into one exposure - so it cannot support a gene-level claim. What would change it: the same analysis in an unselected cohort with per-gene effects.
• PMID 42440143, The clinical relevance of RAS pathway gene mutations in childhood B-cell acute lymphoblastic leukemia (488 patients). Cited at weak/moderate. Its headline FLT3/NF1 result rests on 11 NF1-mutated patients with an OS hazard ratio confidence interval of 1.40-109.46, and the authors state the effect was clearest compared with patients without KMT2A rearrangements - the population this graph is not about. What would change it: replication in a KMT2A-rearranged cohort.
• PMID 40196700, the TP53/revumenib resistance report. A bioRxiv preprint, cited at weak and labelled as a preprint in its citation type. What would change it: peer-reviewed publication, and a replication in KMT2A-rearranged rather than mixed AML material.
• PMID 39116419, the paediatric therapy-related neoplasm series. Its 18% TP53/17p figure is for the whole cohort, about half of which is KMT2A-rearranged, and the abstract states that the non-KMT2A-rearranged cases had heterogeneous drivers - so the 18% may be concentrated where the fusion is not. Cited with that caveat rather than dropped.
• Adult KMT2A-rearranged AML literature generally (HARMONY, PMID 38965370; and the broader adult AML landscape). Used only for the age-contrast argument, never for frequencies transferred into paediatric disease. What would change it: nothing - the transfer is the error this beat is guarding against.

Not pursued, and why. ChEMBL activity-row counts for MEK, SHP2 and CDK4/6 inhibitors were not fetched. A row count measures how much medicinal chemistry has been done and would have added nothing to a beat whose therapeutic conclusion turns on patient access and clonality rather than on chemical matter. DepMap dependency scores for KRAS and NRAS were not quoted for the same class of reason, stated in gene-kras: a dependency score measures lines carrying clonal RAS mutations, which is the opposite of the situation this beat describes.

All ten accessions were resolved this session from gene_exact:<SYMBOL> AND organism_id:9606 AND reviewed:true with gene_primary and length requested, and both accession and length are asserted in every entry so verify-accessions.py can catch a swap.

Two findings the brief's rules did not already cover:

1. The RAS family defeats the length check entirely. KRAS, NRAS and HRAS are all 189 aa, they share hotspot codons, and NRAS's UniProt record lists HRAS1 as a synonym. The brief notes that SMC1A and SMC3 are 16 residues apart and calls that a limit on the length defence; the RAS case is the limiting case, at zero residues apart. Only gene_primary separates them.
2. gene_exact:CDKN2A returns two reviewed records that BOTH carry the requested symbol as primary. Rule 62 says to stop if more than one row matches, and it fires here on a case that is not an error: p16INK4a and p14ARF are two proteins translated in different frames from one locus. The correct handling is to record both and state which one an entry means, which gene-cdkn2a does.