78% of the noncoding beat rests on a claim that nobody has done something

Five hypotheses in the genome beat named ana-noncoding-gap-ledger-kmt2a-r as the analysis that tests them. That node did not exist. The reference was dangling, and the schema validator did not check ids referenced from inside data{} — only edges. So the rule every hypothesis needs at least one experiment or analysis was being satisfied by a pointer to nothing, in five places at once.

The validator now checks these references (13 dangling refs were found the first time it ran), and this is the analysis that was promised.

78% sounds alarming and mostly is not. A beat about what has not been measured will legitimately say so, and 15 of the 41 sit in open questions or caveats, which is what those fields are for.

The useful split is bounded versus unbounded. An absence claim that names a date, a database and a count — "as of 2026-09-18, PubMed returns 23 records for HOTAIRM1 with leukemia" — is reproducible: a reader can re-run it and see whether it still holds. One that says "nobody has measured X" is a bare assertion that decays silently as the literature moves, and cannot be checked without redoing the work.

All three absence claims falsified in the cycle-36 sweep were unbounded. The bounded ones in this beat have so far survived. That is one beat and a small sample, but it is the axis worth acting on: 8 unbounded HIGH claims against 4 bounded ones.

Each of these asserts an absence to the reader as fact, with nothing behind it that can be re-run. The fix is not to delete them — several are probably true — but to give each a date, a database and a count, or to soften it to what the evidence supports.

• comut-ptpn11-kmt2a-r · summary_expert — “An independent query of the public curation of the same study returns zero PTPN11 mutations across 93 sequenced samples, a discrepancy this node reports rather than resol…”
• genome-enhancer-promoter-contacts-kmt2a-r · summary_expert — “Combined with E4's finding that contacts persist when transcription is switched off, contact frequency is a poor read-out of regulatory state in both directions, and no e…”
• hyp-r-loop-landscape-kmt2a-r · summary_lay — “Fusion proteins force genes to be read very hard, so these knots should pile up - but they pile up in any hard-read gene in any cell, the main way of measuring them is di…”
• hyp-splicing-landscape-kmt2a-r · summary_lay — “The fusion keeps that half, so it should change RNA stitching - but the RNA-stitching changes that have actually been found in this family of leukemias are explained by o…”
• hyp-splicing-landscape-kmt2a-r · summary_expert — “So the surviving hypothesis is narrower and sharper than the one originally written: **splicing is rewired in KMT2A-rearranged leukaemia, and that is no longer in doubt;…”
• hyp-transposable-element-reactivation-kmt2a-r · summary_expert — “Nobody has measured TE or ERV expression, dsRNA sensing or interferon output in a KMT2A-rearranged leukemia, before or after hypomethylating-agent exposure.…”
• hyp-trna-ribosome-biogenesis-kmt2a-r · summary_lay — “This matters in other blood diseases and has never been examined in KMT2A-fusion leukemia - but every fast-growing cancer cell needs more of these machines, so finding th…”
• noncoding-mir-9 · summary_expert — “Every functional experiment is nonetheless myeloid AML; there is no KMT2A::AFF1 B-ALL experiment, and PubMed returns zero records for miR-9 with KMT2A rearrangement b…”

This is a text audit, not a fact check. It does not know whether any of these claims is true. It finds the sentences whose form makes them fragile; confirming or refuting each one is the literature work it is meant to prioritise, not replace.

It also cannot distinguish a load-bearing absence from a passing mention, and it will flag a counter-evidence note that correctly reports an absence having been refuted — which is the opposite of a defect. And it only matches phrasings in its pattern, so the count is a floor, not a total.