The eleven-line panel is nine patients, seven of them myeloid, and none an infant

Nine individuals, not eleven. SEMK2 (CVCL_S906) is a Cellosaurus child record of SEM (CVCL_0095); MOLM-14 of MOLM-13. Every per-line statistic computed over the eleven assumes an independence those pairs do not have.

Seven of the nine are myeloid or monocytic. Only SEM/SEMK2, RS4;11 and HB11;19 are B-lineage. The disease this project exists for is infant B-ALL, and the panel that supplies its dependency conclusions is dominated by adult monocytic leukaemia.

The index fusion is three patients: SEM/SEMK2 (5Y), RS4;11 (32Y, whose Cellosaurus disease term is literally Adult B acute lymphoblastic leukemia), and MV4-11 (10Y, monocytic). THP-1 is KMT2A::MLLT3 and HB11;19 is KMT2A::MLLT1.

A sweep found 138 entries that quantify this panel without the patient-count caveat. Appending the same boilerplate to 138 entries would be a bulk mutation that adds noise to every one of them and makes the graph harder to read, not easier — and it would put the correction in 138 places that then drift apart.

So the correction lives here, once, and the entries that carry the most weight point at it. A reader who wants to know what 'the 11 KMT2A-rearranged lines' means has one place to look. The alternative — quietly leaving it uncorrected because correcting it properly is expensive — is the failure mode this project has been auditing itself for.

What the 138 entries say is not wrong: there are eleven lines in the DepMap panel. What is wrong is the inference a reader draws about sample size, lineage and age, and that inference is what this entry exists to block.

Assembling the table, the coordinator mapped NOMO-1 to CVCL_1274 from memory. That accession resolves to HEC-1, a 71-year-old's type II endometrial adenocarcinoma. NOMO-1 is CVCL_1609, a 31-year-old woman's adult acute monocytic leukaemia.

The error was caught because the API returns the primary name and the disease term, and an endometrial adenocarcinoma in a leukaemia panel is conspicuous. Had the recalled accession been another leukaemia line it would have passed. This is the fifth identifier-from-recall error the coordinator has made in four cycles, and it happened inside the check designed to enforce the rule against it. Resolve by accession and read back the primary name; the accession alone is not self-validating.

And again, in the same entry. The SEM founding paper's DOI was also typed from memory (...tb04912.x); the real one is 10.1111/j.1365-2141.1994.tb04726.x. Caught by verify-citations.py --strict on the run immediately after. That is two recall errors inside the single entry whose subject is recall errors, which settles the question of whether care is a sufficient control. It is not; the checker is.

This entry's own open question was whether the SEM patient was diagnosed in infancy — Cellosaurus records age at sampling and SEM is relapse-derived, so if she had been an infant at diagnosis the most-used index-fusion line in the world would be infant material and the verdict would move.

The founding paper was found and read, and it does not say. PMID 8199015 states the line was "established from the peripheral blood of a 5-year-old girl in relapse with acute lymphoblastic leukaemia". Age at sampling, at relapse; no age at diagnosis anywhere in the abstract. The full text is unreachable: no PMCID, not in Europe PMC, checked under the corrected rule against both fullTextXML and efetch db=pmc. Same for the companion breakpoint paper.

So the question is bounded-unresolvable by this project's routes, not open. One could reason that infant ALL relapses early and a relapse at five years is unlikely to follow an infant diagnosis — but that is inference, and this graph has a rule about presenting inference as evidence.

The reading returned something better than it was sent for. SEM is biphenotypic: B-lineage CD19, CD22 and CDw75 in the absence of CD20, CD24 and immunoglobulin, together with myeloid CD13, CD15, CD33 and CDw65. That refines this entry's own lineage arithmetic. Calling the panel 'seven myeloid, four B-lineage' overstates how clean the B-lineage arm is: its largest member co-expresses four myeloid antigens. And the companion paper places SEM's crossover in KMT2A intron 7 — the 5' edge of the major breakpoint cluster region, which this graph elsewhere gives as intron 7 to exon 13.

This entry's own open question was which of the 138 entries actually draw an inference the true composition would change. Audited on 2026-09-20.

A first pass for sample-size and generalisation language returned 20 candidates. Reading three of them showed eighteen were a false positive: they match on the phrase "computed by us from per-line API values", which is a statement about data provenance, not about sample size. A blanket edit would have annotated eighteen entries with an irrelevant caveat.

The answer is one. mechanism-kat8-mof-requirement-is-myeloid-not-fusion-specific E2 computes "n = 11 KMT2A-rearranged versus 42 other leukaemia gives a minimum detectable effect around d = 1" — a power calculation on eleven assumed-independent samples when there are nine patients. It is now corrected, and the correction runs against that entry's own conclusion: a wider detectable effect makes its null weaker, not stronger.

So the proportionate response to 138 matches was one citable entry and one edit. The instinct to bulk-annotate would have been wrong by a factor of 138, and the instinct to stop at 'too many to fix' would have missed the one that mattered.