start / graph / ana-q113-every-bounded-absence-classified-by-the-instrument-that-bounded-it

AnalysisstrongQ113-bounded-absence-audit

All 39 bounded absences in this graph, classified by WHAT BOUNDS THEM: 23 rest on an abstract-level literature search, 7 on a threshold or a query filter, and only 2 on reading the primary source

also called bounded absence instrument census · Q113 census

This project writes down things nobody has measured, which is useful only if the search that found nothing was capable of finding something. Sorting all thirty-nine of those write-ups by the tool that produced them shows that nearly six in ten rest on a search of titles and summaries only, and just two rest on somebody reading the actual papers.

The same thing, for a scientist

Complete census (39 of 39, no sampling) of every entry in this corpus whose id begins `absence-`, classified not by topic but by the instrument whose output the headline claim rests on. Seven classes were used: reading the PRIMARY SOURCE in full (safe by construction), reading a FULL TEXT or an archive's own records/file headers, a REGISTRY query, a DATABASE-ENTITY or metadata-label enumeration, a FILTERED database query, a THRESHOLDED peak set, and an ABSTRACT-LEVEL literature search. The distribution is 2 / 5 / 2 / 3 / 3 / 1 / 23. The two failure modes this project has already proven - a threshold that manufactured an absence (cycle 63, pre-IDR peaks present at globalIDR 0.0588 and 0.227 where the IDR-called set was empty) and a query filter that manufactured one (ENCODE `assay_title=TF ChIP-seq` turning a live BARD1 target record into an HTTP 404) - together account for 4 entries, and a third mode (the absence is real for the spelling used) is a co-factor in at least 6 more. The 23 abstract-level entries are not thereby wrong; rule 80 says only that they are bounded by abstracts, and 14 of the 23 say so explicitly. This census is the deliverable whether or not anything overturns; the re-runs that followed it are recorded in the sibling entries.

Go deeper

Why classify by instrument and not by topic

A bounded absence is an inference from a measurement whose subject is the search, not the biology. Zero called peaks in 13 of 13 experiments is a true statement about a peak caller. No occupancy is a statement about a protein. Cycle 63 published the second from the first and the second was false 1. Sorting by topic hides that; sorting by instrument makes it the first thing you see. The classes below are ordered from safest to least safe, and the ordering is a property of what each instrument can and cannot see - not of how carefully any beat worked.

The seven classes, safest first

1. Reading the PRIMARY SOURCE in full - 2 of 39 (5.1%). The only class safe by construction. An absence established by opening every candidate paper and tabulating what it actually contains cannot be defeated by a query form, a threshold or a spelling. absence-no-affinity-measurement-of-a-kmt2a-fusion-or-full-length-kmt2a-for-dna read six papers in full (PMIDs 11842107, 16990798, 18483194, 20010842, 23990460, 34280370) and built a construct table; absence-the-cd33-splice-genotype-has-never-been-crossed-with-kmt2a-status-in-aaml0531 rests on reading two published analyses of one trial. Neither can be overturned by re-running a search.

2. Reading a FULL TEXT, an archive's own records, or a file header - 5 of 39 (12.8%). absence-gse210480-processed-tracks-are-1-kb-binned is the model: it opened the bigWig headers and measured the bin size, so the bound is a physical property of the deposited files 2. Also here: the four-route readability test in absence-af4-mll-affinity-purification-unreadable-by-all-four-routes, the 60 GSM records read in absence-no-h3k9-methylation-measurement-at-the-kmt2a-locus-in-a-blood-cell, absence-h3k9me3-at-the-kmt2a-locus-was-assayed-and-never-reported, and absence-no-human-ctcf-or-cohesin-occupancy-measurement-in-a-blood-cell-at-kmt2a, which closed itself in-run by reading ENCODE rather than asserting the gap.

3. A REGISTRY query - 2 of 39 (5.1%). ClinicalTrials.gov and drug labels. Bounded by a drug NAME LIST, which is rule 72's exact failure. Re-run this cycle and the name list was short by four drugs; the headline absence survived and a supporting count did not 3.

4. A DATABASE-ENTITY or metadata-label enumeration - 3 of 39 (7.7%). absence-intact-holds-seven-mll-fusion-proteins-and-not-the-index-one enumerates the IntAct interactor namespace; absence-no-h3k9me1-or-h3k9me2-in-a-cd34-progenitor-in-the-ihec-federation enumerates 178 assay labels over 49,663 IHEC datasets. These are bounded by a controlled vocabulary, and the second says so in its own title - the absence is a consortium artefact, and GEO has had the data since 2017.

5. A FILTERED database query - 3 of 39 (7.7%). PROVEN FAILURE MODE. All three used the ENCODE portal with assay_title=TF ChIP-seq pinned to the query. That filter turns a live target record into an HTTP 404 for any protein ENCODE has not ChIPped, and three entries recorded the 404 as no target record exists 4. absence-bard1-and-exo1-are-the-only-two-resection-proteins-with-no-occupancy-data-in-any-archive caught and corrected this; absence-no-resection-factor-localised-at-the-kmt2a-locus and absence-no-h3k9-writer-or-reader-localised-at-the-kmt2a-locus still carry the uncorrected form.

6. A THRESHOLDED peak set - 1 of 39 (2.6%). PROVEN FAILURE MODE. absence-no-damage-state-occupancy-measurement-for-any-resection-factor-at-the-kmt2a-locus states that the resting state is MEASURED and negative in 13 ENCODE datasets. That statement is an output of the IDR peak caller, and the cycle-63 pre-IDR check found the peak present in both biological replicates at globalIDR 0.0588 (BRCA1 HeLa-S3, ENCFF457RFT) and 0.227 (CBX3 K562, ENCFF271ZAZ) 1. The signal-level re-read of those tracks is a separate beat this cycle and is deliberately not duplicated here.

7. An ABSTRACT-LEVEL literature search - 23 of 39 (59.0%). Not bounded, per rule 80. These are the majority. They are not wrong - 18 of the 23 store at least one of the three hygiene artefacts (the query strings with their counts, an operand or positive control, or the alias/name list), and 5 store none: absence-enl-function-in-developing-brain, absence-kdm2b-and-the-kmt2a-fusion-are-both-nucleolar-and-nobody-has-put-them-together, absence-ledgf-agent-never-tested-in-menin-inhibitor-resistant-disease, absence-enl-yeats-and-condensate-phospho-gating and absence-no-proximity-degron-or-crispr-study-of-wrad-in-kmt2a-fusion-leukaemia (the last stores query strings but no counts and no control). 18 + 5 = 23. But no list of synonyms reaches a measurement that appears only in a supporting figure, and several make a claim whose natural instrument is not PubMed at all (see the next section).

The sharpest defect the census found: four entries ask PubMed a question PubMed cannot answer

Four absences are bounded by an abstract-level literature search but make a claim about an object that is not indexed by abstracts:

entry	claim is about	instrument used	the matching instrument
`absence-no-kmt2a-fusion-junction-peptide-reported-in-any-mass-spectrometry-dataset`	mass-spectrometry DATASETS	PubMed `[tiab]`	PRIDE / ProteomeXchange (the entry names this in `not_swept`; swept this cycle, 3 MLL-AF4 datasets exist 5)
`absence-no-junction-specific-antibody-exists-for-any-kmt2a-fusion`	a commercial REAGENT	PubMed `[tiab]`	Antibody Registry, vendor catalogues - not searched
`absence-no-phosphosite-mapped-on-any-kmt2a-fusion-protein`	PTM database records	PubMed `[tiab]`	PhosphoSitePlus, PRIDE - not searched
`absence-no-menin-dependence-measurement-in-kmt2a-r-sarcoma`	whether a MODEL exists	PubMed + Cellosaurus	Cellosaurus WAS queried - this one has the right instrument and is listed only for contrast

This is the same shape as the two proven modes: the answer returned is a property of the instrument, and the instrument was never the one that could see the object.

What this census does NOT say

It does not say 59% of these absences are wrong. Seven were re-run against their live instruments this cycle and five survived intact, one survived with a corrected supporting count, and one had half of it overturned 6. The class distribution is a statement about exposure, not about error rate. It also does not cover the 332 node files that carry absence_class, never-measured or measured-not-analysed in a data field without an absence- id - rule 84's sibling problem, and the open question absence-h3k79me2-decay-half-life-never-measured already asks it.

Facts and figures

script	no script; the classification is a hand-read of all 39 entries and is reproduced in full in data.census below
status	run
run date	2026-09-22
run by	Q113-bounded-absence-audit
result summary	39 of 39 classified, 0 skipped, 0 failed. 23 rest primarily on an abstract-level literature search, 7 on an instrument with a proven failure mode (1 threshold, 3 filtered query, 3 database-entity enumeration), 7 on reading a full text or an archive record, 2 on a registry, and only 2 on reading the primary source in full.

class counts

denominator

{
  "node_files_in_corpus": 2192,
  "files_whose_basename_starts_with_absence_": 39,
  "files_containing_absence_class_or_never_measured_or_measured_not_analysed": 332,
  "classified_by_instrument": 39,
  "skipped": 0,
  "failed": 0,
  "closes": "39 classified + 0 skipped + 0 failed == 39 total",
  "re_run_against_the_live_instrument": 7,
  "not_re_run": 32,
  "re_run_closes": "7 re-run + 32 not re-run == 39 total",
  "counted_how": "python glob over data/graph/nodes/*/*.json, basename prefix test, counted by this beat rather than quoted from the inventory header"
}

class definitions

{
  "primary-source-read": "every candidate paper opened in full and tabulated. Safe by construction.",
  "full-text-or-archive-record-read": "a full text, an archive's own sample records, or a deposited file's header was read. Bounded by the files enumerated.",
  "registry-query": "ClinicalTrials.gov or a drug label. Bounded by the drug NAME LIST (rule 72).",
  "database-entity-enumeration": "an interactor namespace or a metadata assay-label vocabulary was enumerated. Bounded by the controlled vocabulary.",
  "filtered-database-query": "an archive query carrying a facet filter. PROVEN to manufacture absences (ENCODE assay_title).",
  "thresholded-peak-set": "a called-peak set, usually IDR. PROVEN to manufacture absences (cycle 63).",
  "abstract-level-literature-search": "esearch over titles and abstracts. NOT a bound on the literature (rule 80)."
}

class counts — the numbers behind the chart

{
  "primary-source-read": 2,
  "full-text-or-archive-record-read": 5,
  "registry-query": 2,
  "database-entity-enumeration": 3,
  "filtered-database-query": 3,
  "thresholded-peak-set": 1,
  "abstract-level-literature-search": 23,
  "sums_to": 39
}

census

[
  {
    "id": "absence-no-affinity-measurement-of-a-kmt2a-fusion-or-full-length-kmt2a-for-dna",
    "primary_instrument": "primary-source-read",
    "also": [],
    "re_run": false,
    "note": "six papers read in full, construct table built. The safest entry in the set."
  },
  {
    "id": "absence-the-cd33-splice-genotype-has-never-been-crossed-with-kmt2a-status-in-aaml0531",
    "primary_instrument": "primary-source-read",
    "also": [],
    "re_run": false,
    "note": "two published analyses of one trial, read."
  },
  {
    "id": "absence-gse210480-processed-tracks-are-1-kb-binned",
    "primary_instrument": "full-text-or-archive-record-read",
    "also": [],
    "re_run": false,
    "note": "bigWig headers measured. A physical bound."
  },
  {
    "id": "absence-af4-mll-affinity-purification-unreadable-by-all-four-routes",
    "primary_instrument": "full-text-or-archive-record-read",
    "also": [],
    "re_run": false,
    "note": "four retrieval routes tested and each failure recorded separately."
  },
  {
    "id": "absence-no-h3k9-methylation-measurement-at-the-kmt2a-locus-in-a-blood-cell",
    "primary_instrument": "full-text-or-archive-record-read",
    "also": [],
    "re_run": false,
    "note": "all 60 GSM records of the founding deposit read; names its own weak point (ENCODE/BLUEPRINT CD34+ not queried)."
  },
  {
    "id": "absence-h3k9me3-at-the-kmt2a-locus-was-assayed-and-never-reported",
    "primary_instrument": "full-text-or-archive-record-read",
    "also": [
      "abstract-level-literature-search"
    ],
    "re_run": false,
    "note": "CLOSED in cycle 59."
  },
  {
    "id": "absence-no-human-ctcf-or-cohesin-occupancy-measurement-in-a-blood-cell-at-kmt2a",
    "primary_instrument": "full-text-or-archive-record-read",
    "also": [],
    "re_run": false,
    "note": "closed itself in-run by reading ENCODE instead of asserting the gap. The pattern other entries should copy."
  },
  {
    "id": "absence-nup98-response-rate-in-menin-inhibitor-trials",
    "primary_instrument": "registry-query",
    "also": [
      "full-text-or-archive-record-read"
    ],
    "re_run": false,
    "note": "counts the NUP98 string in four registry records and both drug labels - a document-level bound, not a keyword bound."
  },
  {
    "id": "absence-no-menin-inhibitor-trial-in-kmt2a-r-sarcoma",
    "primary_instrument": "registry-query",
    "also": [
      "abstract-level-literature-search"
    ],
    "re_run": true,
    "verdict": "SURVIVES with a corrected supporting count: three solid-tumour menin-inhibitor trials exist, not two."
  },
  {
    "id": "absence-intact-holds-seven-mll-fusion-proteins-and-not-the-index-one",
    "primary_instrument": "database-entity-enumeration",
    "also": [],
    "re_run": true,
    "verdict": "SURVIVES. Its own named next step - a federated PSICQUIC sweep - is not executable as written; see ana-q113-psicquic-free-text-cannot-retrieve-an-interactor-label."
  },
  {
    "id": "absence-no-h3k9me1-or-h3k9me2-in-a-cd34-progenitor-in-the-ihec-federation",
    "primary_instrument": "database-entity-enumeration",
    "also": [],
    "re_run": false,
    "note": "the model self-aware entry: it names the vocabulary that bounds it, states the absence is a consortium artefact, and points at the GEO deposit that contradicts the consortium."
  },
  {
    "id": "absence-no-menin-inhibitor-occupancy-dataset-in-an-infant-derived-kmt2a-aff1-line",
    "primary_instrument": "database-entity-enumeration",
    "also": [
      "abstract-level-literature-search"
    ],
    "re_run": false,
    "note": "ANSWERED in cycle 55; the inventory flags it as already resolved elsewhere."
  },
  {
    "id": "absence-no-resection-factor-localised-at-the-kmt2a-locus",
    "primary_instrument": "filtered-database-query",
    "also": [
      "abstract-level-literature-search"
    ],
    "re_run": true,
    "verdict": "The literature half SURVIVES. The archive half carries the uncorrected ENCODE 404 claim for BARD1, which is false and which q37 already corrected in a sibling entry."
  },
  {
    "id": "absence-no-h3k9-writer-or-reader-localised-at-the-kmt2a-locus",
    "primary_instrument": "filtered-database-query",
    "also": [
      "abstract-level-literature-search",
      "thresholded-peak-set"
    ],
    "re_run": true,
    "verdict": "PARTLY OVERTURNED. Literature half survives, including under the untried alias KMT1A (22 records, 0 with KMT2A or MLL). Archive half falls: GEO holds six human SUV39H1 occupancy samples in two cell types."
  },
  {
    "id": "absence-bard1-and-exo1-are-the-only-two-resection-proteins-with-no-occupancy-data-in-any-archive",
    "primary_instrument": "filtered-database-query",
    "also": [
      "full-text-or-archive-record-read"
    ],
    "re_run": true,
    "verdict": "SURVIVES. ChIP-Atlas returns 404 for BARD1 and EXO1 with the CTCF positive control passing at HTTP 200. This entry also contains the correction of the ENCODE 404 error, which is why it is the best of the three in this class."
  },
  {
    "id": "absence-no-damage-state-occupancy-measurement-for-any-resection-factor-at-the-kmt2a-locus",
    "primary_instrument": "thresholded-peak-set",
    "also": [
      "filtered-database-query"
    ],
    "re_run": false,
    "note": "NOT re-run here by design - the signal-level re-read of these tracks is a concurrent beat. Its 'resting state MEASURED and negative in 13 ENCODE datasets' is the corpus's clearest instance of the threshold class and the cycle-63 pre-IDR finding bears directly on it."
  },
  {
    "id": "absence-cxcr4-axis-never-measured-under-menin-inhibition",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "database-entity-enumeration"
    ],
    "re_run": true,
    "verdict": "SURVIVES. menin[tiab] AND CXCR4[tiab] = 0 on two independent runs, non-empty translation, warninglist 'No items found.', positive control menin[tiab] AND HOXA9[tiab] = 33. The untried symmetric form MEN1[tiab] AND CXCR4[tiab] returns 1 and it is a neuroendocrine cell-line paper, off topic."
  },
  {
    "id": "absence-h3k79me2-decay-half-life-never-measured",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": true,
    "verdict": "SURVIVES. H3K79[tiab] AND half-life[tiab] = 0 (genuine zero); H3K79me2[tiab] AND kinetics[tiab] = 3, all title-checked off topic (kinetoplastid DOT1, an elongation-rate preprint, a methylome-persistence paper); positive control H3K27me3[tiab] AND half-life[tiab] = 7."
  },
  {
    "id": "absence-no-kmt2a-fusion-junction-peptide-reported-in-any-mass-spectrometry-dataset",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": true,
    "verdict": "SURVIVES as written, but the claim is about MS datasets and the instrument was PubMed. PRIDE swept this cycle: 3 projects under MLL-AF4 and 2 under RS4;11. The class measured-not-searched is now backed by accessions."
  },
  {
    "id": "absence-no-cell-state-for-aff4",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "full-text-or-archive-record-read"
    ],
    "re_run": false,
    "note": "alias list recorded; one full text opened; GEO descriptive half checked."
  },
  {
    "id": "absence-no-cell-state-for-ash2l",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "full-text-or-archive-record-read"
    ],
    "re_run": false,
    "note": "as above."
  },
  {
    "id": "absence-no-cell-state-for-ell",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "full-text-or-archive-record-read"
    ],
    "re_run": false,
    "note": "alias list includes ELL2, which is a different gene; that widens the net so the zero stays conservative."
  },
  {
    "id": "absence-no-cell-state-for-kmt2b-and-the-mll4-alias-contaminates-the-count",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "no full text opened; the entry is mostly about a symbol collision it found."
  },
  {
    "id": "absence-enl-function-in-developing-brain",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "THE WEAKEST BOUND IN THE SET, by its own open question: a two-query floor, no MeSH terms, no full alias set, no full-text search. It also carries no confidence_overall and no status."
  },
  {
    "id": "absence-kdm2b-and-the-kmt2a-fusion-are-both-nucleolar-and-nobody-has-put-them-together",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "an intersection-of-publication-sets absence, the form most exposed to rule 80."
  },
  {
    "id": "absence-no-experiment-tests-cxxc-reader-competition-at-a-kmt2a-fusion-target",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "ALREADY SELF-CORRECTED 2026-09-21 and carries status disputed. A worked example of this class failing."
  },
  {
    "id": "absence-no-junction-specific-antibody-exists-for-any-kmt2a-fusion",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "INSTRUMENT MISMATCH: a reagent's existence is a catalogue fact. Antibody Registry and vendor catalogues were not searched."
  },
  {
    "id": "absence-no-study-has-tested-for-gained-fusion-occupancy-under-menin-inhibition",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "primary-source-read"
    ],
    "re_run": false,
    "note": "full texts read for the on-topic hits, which lifts it well above the class average."
  },
  {
    "id": "absence-no-menin-dependence-measurement-in-kmt2a-r-sarcoma",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "database-entity-enumeration"
    ],
    "re_run": false,
    "note": "Cellosaurus and GEO elink both queried; bounded by two fusion spellings (YAP1::KMT2A, VIM::KMT2A) with LDB1::KMT2A checked in the repository but not the literature."
  },
  {
    "id": "absence-ledgf-agent-never-tested-in-menin-inhibitor-resistant-disease",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "registry-query"
    ],
    "re_run": false,
    "note": "bounded by a drug name list; same exposure as the sarcoma registry entry."
  },
  {
    "id": "absence-ledgf-removal-from-established-kmt2a-leukaemia",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "the stored queries carry NO field tag at all, so they ran as All Fields - broader than labelled, which makes the zeros conservative (rule 94's safe direction)."
  },
  {
    "id": "absence-no-idh-inhibitor-has-been-reported-in-a-kmt2a-rearranged-cohort",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "registry-query"
    ],
    "re_run": false,
    "note": "records its own caught instrument failure - a first-pass false zero - which is the behaviour this census wants to see."
  },
  {
    "id": "absence-no-infant-kmt2a-cohort-has-reported-idh-status",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "every querytranslation stored beside its count. Best hygiene in the abstract-level class."
  },
  {
    "id": "absence-no-kmt2a-mllt1-analysis-stratified-by-reciprocal-or-spliced-status",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "translations inspected; alias list includes the deliberate exclusion of bare ENL as contaminated."
  },
  {
    "id": "absence-partner-stratified-response-in-menin-inhibitor-trials",
    "primary_instrument": "abstract-level-literature-search",
    "also": [
      "registry-query"
    ],
    "re_run": false,
    "note": "the blocking object is a non-retrievable supplement, which no query form can reach."
  },
  {
    "id": "absence-enl-yeats-and-condensate-phospho-gating",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "counts recorded per sub-question."
  },
  {
    "id": "absence-mllt10-kinase-site-relations",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "DEFECT: the fourth stored query reads 'partners AND (CK2 OR CSNK2 OR casein kinase 2)[tiab]' - the word 'partners' is a placeholder that leaked into the stored string, so that query cannot be re-run as written. The other three are intact."
  },
  {
    "id": "absence-no-phosphosite-mapped-on-any-kmt2a-fusion-protein",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "INSTRUMENT MISMATCH: claims a PTM-database-shaped absence; PhosphoSitePlus and PRIDE were not searched."
  },
  {
    "id": "absence-no-proximity-degron-or-crispr-study-of-wrad-in-kmt2a-fusion-leukaemia",
    "primary_instrument": "abstract-level-literature-search",
    "also": [],
    "re_run": false,
    "note": "three method-word queries; the method-word list is the bound and it is recorded."
  }
]

the three proven failure modes and their corpus exposure

{
  "mode_1_a_threshold_manufactured_it": {
    "entries_primarily_exposed": 1,
    "entry": "absence-no-damage-state-occupancy-measurement-for-any-resection-factor-at-the-kmt2a-locus",
    "status": "handled by a concurrent beat, not by this one"
  },
  "mode_2_a_query_filter_manufactured_it": {
    "entries_primarily_exposed": 3,
    "of_which_already_corrected": 1,
    "still_carrying_the_error": [
      "absence-no-resection-factor-localised-at-the-kmt2a-locus",
      "absence-no-h3k9-writer-or-reader-localised-at-the-kmt2a-locus"
    ]
  },
  "mode_3_real_for_the_spelling_used": {
    "co_factor_in": 6,
    "entries": [
      "absence-no-menin-inhibitor-trial-in-kmt2a-r-sarcoma",
      "absence-ledgf-agent-never-tested-in-menin-inhibitor-resistant-disease",
      "absence-no-cell-state-for-kmt2b-and-the-mll4-alias-contaminates-the-count",
      "absence-no-cell-state-for-ell",
      "absence-mllt10-kinase-site-relations",
      "absence-no-menin-dependence-measurement-in-kmt2a-r-sarcoma"
    ],
    "tested_this_cycle": "yes, on three of the six; one name-list gap found (menin-inhibitor registry), two clean (KMT1A for SUV39H1, DNA2L for DNA2, HP1-gamma/HP1 gamma for CBX3 all return zero on-topic)"
  }
}

absence class vocabulary compliance 2026 09 22

{
  "checker": "python scripts/check-absence-class.py",
  "entries": 39,
  "classified_in_vocabulary": 36,
  "missing": 0,
  "out_of_vocabulary_free_text": 3,
  "the_three": [
    "absence-h3k9me3-at-the-kmt2a-locus-was-assayed-and-never-reported: 'measured-and-now-analysed-for-RPE; never-measured in any haematopoietic cell'",
    "absence-no-human-ctcf-or-cohesin-occupancy-measurement-in-a-blood-cell-at-kmt2a: 'reading gap, not a literature gap - CLOSED IN-RUN by reading ENCODE GM12878'",
    "absence-gse210480-processed-tracks-are-1-kb-binned: 'measured-not-analysed, with a newly measured RESOLUTION bound on what the processed deposit can answer'"
  ],
  "reading": "all three free-text strings say something TRUER than any vocabulary term could, and all three belong to the full-text-or-archive-record class - the best-bounded entries in the corpus. The vocabulary is too coarse for a per-cell-type or per-resolution verdict, and the entries that most deserve a precise class are the ones that broke it. Recommend a structured pair - class plus scope - rather than forcing these three into a single token.",
  "also": "1 entry is 'unclassified-not-checked' (absence-af4-mll-affinity-purification-unreadable-by-all-four-routes), which is the honest default, and 2 claim data exists without naming an accession"
}

How this could bite us back

The safest class is also the most expensive. Telling every future beat to read the primary source in full would reduce the number of bounded absences this project can write by roughly an order of magnitude, and a bounded absence that is never written is worse than an abstract-level one that is honestly labelled.
A census that ranks instruments invites beats to pick the class rather than the question. The right instrument is the one that can see the object, and for a question about what nobody has REPORTED, an abstract-level search is the correct instrument, not a weak one.

Still open

The 332 node files that carry `absence_class`, `never-measured` or `measured-not-analysed` in a data field WITHOUT an `absence-` id were not classified. Those are the absences held in data fields rather than in evidence claims, where no confidence tier and no caveat attaches - exactly the population `absence-h3k79me2-decay-half-life-never-measured` already asked about. How many of the 293 unclassified ones are load-bearing?
Should `primary_instrument` become a required field on every new `absence-` entry, so this census regenerates from a script rather than from a hand-read? Rule 90 says a figure with no reporter goes stale invisibly, and this one has no reporter.
Two entries claim a reagent or a PTM-database absence using PubMed. Is there any cheap catalogue-shaped instrument - Antibody Registry, PhosphoSitePlus - that a beat could be told to reach for, the way it is now told to reach for GEO?

Every source behind this entry (6)

E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
An IDR peak set can be empty at an interval where the pre-IDR replicated peak is present, and at the KMT2A intervals the BRCA1 call missed the 0.05 cut by 0.009.
ENCODE portal file record ENCFF457RFT (BRCA1 ChIP-seq, HeLa-S3) and ENCFF271ZAZ (CBX3 ChIP-seq, K562) · ENCODE Project 2026
What this evidence cannot show
This determination was made by the coordinator in cycle 63 and is quoted here, not re-derived: the signal-level re-read of these tracks is a different beat running concurrently and this beat was instructed not to touch that locus at signal level. What this entry contributes is the classification, not a replication of that measurement.
Evidence that argues against this (1)
- A globalIDR of 0.0588 is a failed reproducibility test, not a suppressed peak. Whether the correct conclusion is 'there is occupancy' or 'the replicates do not agree well enough to say' is exactly what the concurrent signal-level beat is for, and this entry does not prejudge it. Li Q et al. 2011
E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
An absence bounded by a physical property of a deposited file - the bin size of a bigWig - cannot be defeated by a query form, a threshold or a spelling.
GSE210480, per-sample supplementary file inventory (NCBI GEO) · NCBI Gene Expression Omnibus 2026
What this evidence cannot show
The bound is on the PROCESSED deposit only. The raw reads remain and would answer a finer question, which that entry states.
Evidence that argues against this (1)
- A file-header bound is safe only for the files enumerated. If a series later gains a sample with a different pipeline, the bound does not extend to it - and the registry already records that one deposit can mix padded and unpadded chrom tables across its own files. NCBI Gene Expression Omnibus 2026
E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.trial_registry
The menin-inhibitor drug name list used to bound the sarcoma registry absence omits icovamenib/BMF-219, DS-1594b and BN104, and one of the omitted names retrieves a solid-tumour trial the entry does not know about.
ClinicalTrials.gov study record NCT05631574 - Study of Covalent Menin Inhibitor BMF-219 in Adult Patients With KRAS Driven Non-Small Cell Lung Cancer, Pancreatic Cancer and Colorectal Cancer (COVALENT-102) · ClinicalTrials.gov 2026
NCT05631574
What this evidence cannot show
Retrieved through the v2 API `query.intr` facet on 2026-09-22. The trial is TERMINATED and is not in sarcoma; it changes a supporting count in that entry, not its headline.
Evidence that argues against this (1)
- The headline absence it was tested against - no menin inhibitor has ever enrolled a KMT2A-rearranged solid tumour - survives the expanded name list: the union of 15 drug names returns 69 trials and not one enrolls a KMT2A-rearranged solid tumour. ClinicalTrials.gov 2026
E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The ENCODE portal returns HTTP 404 for a zero-result search, so pinning `assay_title=TF ChIP-seq` to a target query converts 'ENCODE has not ChIPped this protein' into 'no target record exists'; BARD1 and RAD50 both return HTTP 200 on the bare target record while EXO1, DNA2, SUV39H1 and SUV39H2 return 404.
ENCODE portal target record /targets/BARD1-human/ (HTTP 200) versus /targets/SUV39H1-human/ (HTTP 404) · ENCODE Project 2026
What this evidence cannot show
Fetched 2026-09-22. The 404 is a property of the whole query, not of the protein; the distinction only matters when the 404 is written down as a fact about the protein, which three entries do.
Evidence that argues against this (1)
- For SUV39H1 and EXO1 the 404 turns out to be honest - neither has an ENCODE target record under any label, tagged or untagged - so the filter did not manufacture those two. The mode is real and its hit rate here is 1 of 4 proteins, not 4 of 4. ENCODE Project 2026
E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
PRIDE holds three proteomics datasets indexed under MLL-AF4, so the junction-peptide absence is measured-not-searched with named accessions rather than an unbounded to-do.
PRIDE Archive project search, keyword MLL-AF4: PXD043920, PXD041875, PXD009338 · EMBL-EBI PRIDE Archive 2026
What this evidence cannot show
A keyword hit on a project record means the string appears in the project metadata, not that a junction peptide was searched for or reported. Establishing the absence still requires re-searching the raw files against a fusion-augmented database, which is what that entry's `not_swept` field asks for.
Evidence that argues against this (1)
- PXD009338 is already cited by this graph, so at least one of the three has been opened before and no junction peptide was reported from it - which is weak evidence that the sweep, when run, will confirm the absence rather than overturn it. EMBL-EBI PRIDE Archive 2026
E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Seven of the thirty-nine were re-run against their live instruments and one had half of it overturned: GEO holds human SUV39H1 chromatin-occupancy data in two cell types, which the ENCODE-only census could not see.
Chaetocin-mediated SUV39H1 inhibition targets stemness and oncogenic networks of diffuse midline gliomas and synergizes with ONC201 · Xin D, Liao Y, et al. · Neuro-Oncology 2024
PMID 38011799
What this evidence cannot show
The overturn is of the ARCHIVE half only. No SUV39H1 occupancy has been measured in a haematopoietic cell, and none has been read at the KMT2A locus, so the literature half of that absence stands.
Evidence that argues against this (1)
- Neither deposit is in a blood cell - one is patient-derived diffuse midline glioma, the other MCF-7 breast carcinoma - so a reader who wants the KMT2A-locus answer in the disease cell type still has nothing, and the corrected class buys compute, not the experiment. NCBI Gene Expression Omnibus 2024

Connected to

← supportsThe absence-class checker globbed a filename instead of the field, so it walked 40 of 243 entries and reported them as the whole population - and the 40 it called invalid turn out to be two independently re-invented vocabularies
that census classified 39 absence- entries; this shows the checker behind the field was walking 40 of 243, so the population both rest on was six times larger
← related toOnly 20% of this corpus's absences carry a machine-readable query, which I nearly published as an auditability crisis - and then read five of the other 80% and found every one bounded by a better instrument
the instrument-class census covered the 39 absence--named entries; this shows the population needing it is 260, and that reading the primary source in full is commoner than that census could see
tests →What is left of the resection absence after the compute was done: the resting state is now MEASURED and negative in 13 ENCODE datasets, so the whole question collapses onto the DAMAGE state, which is never-measured for all six proteins in every cell type on Earth
classified as the corpus's only entry whose headline rests primarily on a thresholded peak set - the proven mode-1 exposure
supports →The resolution floor: every deposited H3K9 track in GSE210480 is 1,000 bp binned, so the 147 bp exon-11 question cannot be answered from the processed files - and the planned analysis that asked for 500 bp windows was asking for something that is not there
the model entry for the file-header class: the bound is a physical property of the deposited bigWigs and no query form can defeat it
supports →Ten affinity measurements for this contact and not one uses a fusion protein, full-length KMT2A or a nucleosome
one of only two entries in 39 bounded by reading the primary source in full - the only class safe by construction
supports →BARD1 and EXO1 are the only two proteins in the resection layer with ZERO chromatin-occupancy datasets anywhere - ENCODE, ChIP-Atlas, ReMap and a hand-checked GEO title list all return nothing - and BARD1's silence is the loud one, because its obligate partner BRCA1 has thirty-one
re-run against ChIP-Atlas with a passing CTCF control: BARD1 and EXO1 both return zero, so this entry survives and is the best of the three filtered-query entries because it corrects the ENCODE 404 error rather than carrying it
refutes →No BRCA1, BARD1, EXO1, DNA2, HP1-gamma or KDM5A has ever been localised at the KMT2A locus, in any cell type: 21 queries, every querytranslation read back, 0 on-topic records - and the absence splits THREE ways by class, which is what decides what it costs to close
its archive census still records 'BARD1: no ENCODE target record exists (HTTP 404)'; /targets/BARD1-human/ returns HTTP 200 and the 404 was a property of the assay_title filter
supports →Nobody has reported what a menin inhibitor does to CXCR4: zero papers name menin and CXCR4 together, and the RNA-seq that would answer it is already public
re-run twice through pubmed_safe: genuine zero on both runs, non-empty translation, positive control 33 - and the untried symmetric form MEN1[tiab] AND CXCR4[tiab] returns one off-topic record
supports →Bounding an absence the graph already asserts: the H3K79me2 decay half-life, counted, dated and controlled - and why three of the six hits are about the drug's half-life
re-run and survives: the three H3K79me2-plus-kinetics records the original three-clause shape could not interpret are all off topic on title check
supports →No KMT2A fusion junction peptide has ever been reported - the one measurement that would prove the protein exists has not been published
instrument mismatch named: a claim about mass-spectrometry datasets bounded by PubMed abstracts. PRIDE swept this cycle and the archive half now has accessions
← supportsThe GEO sweep that q110 named as free and did not run: SUV39H1 chromatin occupancy HAS been measured in two human cell types, six samples, hg38 bigWigs already deposited - so the writer half of the H3K9me3 absence is measured-not-analysed, not never-measured
the one overturn out of seven re-runs, and the row that fixes the census's mode-2 exposure count
← supportsHow the three occupancy archives spell zero: ENCODE 404, ReMap 401, ChIP-Atlas 404 - and ChIP-Atlas spells a one-character URL typo exactly the same way, so a whole sweep can come back cleanly and be entirely false
the mechanism behind the filtered-database-query class: all three archives carry a data-layer answer on the transport layer
← supportsThe federated sweep this graph prescribes for the missing index-fusion interactor cannot be run: PSICQUIC free text returns ZERO for `mll_af9`, an entity that provably exists and returns 32 records under its accession - so every provider's zero in that sweep is uninterpretable
the database-entity-enumeration class audited: a recommendation stored in an entry is a claim, and this is the first one tested for executability
← supportsNCT05631574 (COVALENT-102): the third solid-tumour menin-inhibitor trial, missing from this graph because its drug is called BMF-219 - terminated after 13 patients for a strategic reason, explicitly NOT for efficacy or safety
the registry-query class re-run: a drug name list short by four names, and an INN that retrieves fewer trials than its own development code
← supportsThe 297 absence-bearing entries that are not named absence-* split cleanly by entry type: the 109 gene and protein entries are bounded by title/abstract search (7 of 7 hand-read), the analyses are bounded by deposited data (0 of 6), and the weighted abstract-level share is 60% - indistinguishable from the 59% measured on the 39 named ones
closes the largest gap that census named in itself: the 297 entries outside the absence-* naming convention have the same instrument profile, so the 59% generalises
← related toA congress abstract book is one indexed document containing thousands of unrelated abstracts, so it matches almost any conjunction: across 18 Europe PMC full-text intersections read to completion, 83 of 321 records (25.9%) are proceedings, ranging 0% to 50% per query
characterises the instrument that census recommends for re-bounding the 23 abstract-level absences, and names the breadth error it introduces
← supportsThe bounded-absence population went 39 to 73 in two days, and the abstract-level defect class went from 26 of 39 to 2 of 34 (Fisher p = 1.3e-5)
re-derives the census denominator from git and reproduces its 39 exactly, then extends it to 73