Bounded absence: no published CRISPR or dependency screen in any infant-derived KMT2A::AFF1 model
also called infant dependency screen gap
We searched for any published experiment that systematically knocked out genes in a leukaemia cell line taken from a baby with this fusion. There are none, and we record the exact searches so someone can check.
The same thing, for a scientist
A bounded-absence record with the queries written out, because this beat's whole point is that a confident wrong count was previously drawn in this area. **Two queries return zero records**: a CRISPR/dependency-screen term crossed with infant-leukaemia terms, and a CRISPR term crossed with the names of every infant KMT2A::AFF1 line identified in this beat. This is an absence of **testing**, not a negative result: nothing published says infant KMT2A::AFF1 cells lack a menin dependency, because nobody has looked at genome scale in those cells.
The queries, verbatim, with counts
Run against PubMed via NCBI eutils esearch on 2026-09-19.
| # | query | count |
|---|---|---|
| A | (CRISPR screen[tiab] OR CRISPR-Cas9 screen[tiab] OR genome-wide screen[tiab] OR dependency screen[tiab]) AND (infant leukemia[tiab] OR infant leukaemia[tiab] OR infant ALL[tiab] OR infant acute lymphoblastic leukemia[tiab]) | 0 |
| B | (PER-490[tiab] OR PER-485[tiab] OR PER-494[tiab] OR PER-785[tiab] OR PER-703[tiab] OR PER-910[tiab] OR PER-784[tiab] OR PER-826[tiab]) | 1 |
| C | (KOCL-45[tiab] OR KOCL-48[tiab] OR KOCL-58[tiab] OR KOCL-69[tiab] OR SCMC-L1[tiab] OR BLIN-3[tiab] OR ALL-PO[tiab]) | 41, noise-dominated |
| D | (CRISPR[tiab]) AND (PER-490[tiab] OR PER-485[tiab] OR PER-785[tiab] OR ALL-PO[tiab] OR SCMC-L1[tiab] OR KOCL[tiab]) | 0 |
| H | (patient-derived xenograft[tiab] OR PDX[tiab]) AND infant[tiab] AND (KMT2A[tiab] OR MLL[tiab]) | 10 |
Queries A and D are the bounded absence. Queries B, C and H are the recall and precision checks that say how much to trust it. Both plural and singular forms of the multi-word terms were covered by including the four infant-disease spellings in A.
Why query C is a worked example of the alias-precision rule, not a result
Query C returns 41 records, and the first page includes papers on SARS-CoV-2 spike protein membrane anchoring, genomic kinship in a selected animal population, Vietnamese coriander powder in fish feed, pulse oximeter bench tests, and oxygen imaging of beta-cell replacement devices. PubMed's parser splits ALL-PO, BLIN-3 and similar hyphenated short tokens and matches their parts.
So 41 is not a count of anything. Reported without title-checking it would look like a healthy literature on infant KMT2A::AFF1 lines, which is precisely the failure mode this project logged as standing rule 5: short, word-like or hyphenated aliases inflate counts, and the inflated number reads as reassurance. The usable observation from C is the opposite of its number: cell-line names of this kind are not reliably searchable in PubMed at all, so no line-name query can establish how much these lines have been used. That has to come from full texts, where line names live in methods and tables rather than abstracts — which is also why query B returns 1.
What the absence does and does not license
Licensed: 'No genome-scale dependency screen in infant-derived KMT2A::AFF1 material has been published and indexed in PubMed with these terms.' The DepMap side is independently consistent: ALL-PO (ACH-002209) is in the collection and has no gene-effect value 2.
Not licensed: any statement about what such a screen would find. This graph's standing rule holds that an absence of testing is not a negative result, and it applies here with full force — the infant population has never been asked the question that the eleven-line panel was asked.
Also not licensed: the claim that these lines are unused. They have been drug-screened against 62 agents, RNA-sequenced, cytogenetically characterised and xenografted. What has not happened is a loss-of-function screen.
| run date | 2026-09-19 |
|---|---|
| conclusion | no published genome-scale loss-of-function screen in infant-derived KMT2A::AFF1 material |
| absence class | never-measured |
| bound | abstract-level PubMed indexing plus one DepMap API endpoint; line names are not reliably searchable in PubMed |
| absence class note | RECLASSIFIED 2026-09-23 (cycle 76). This field held free text, which `scripts/check-absence-class.py` rejects; the prose is preserved here VERBATIM and unsummarised because it is more informative than the token that replaced it. Token chosen: never-measured - the entry's own words are 'absence of testing, not a negative result' - no screen was ever run in infant KMT2A::AFF1 material, which is the definition of never-measured. ORIGINAL absence_class, verbatim: «absence of testing, not a negative result» |
queries — the numbers behind the chart
[
{
"id": "A",
"query": "(CRISPR screen[tiab] OR CRISPR-Cas9 screen[tiab] OR genome-wide screen[tiab] OR dependency screen[tiab]) AND (infant leukemia[tiab] OR infant leukaemia[tiab] OR infant ALL[tiab] OR infant acute lymphoblastic leukemia[tiab])",
"count": 0
},
{
"id": "B",
"query": "(PER-490[tiab] OR PER-485[tiab] OR PER-494[tiab] OR PER-785[tiab] OR PER-703[tiab] OR PER-910[tiab] OR PER-784[tiab] OR PER-826[tiab])",
"count": 1
},
{
"id": "C",
"query": "(KOCL-45[tiab] OR KOCL-48[tiab] OR KOCL-58[tiab] OR KOCL-69[tiab] OR SCMC-L1[tiab] OR BLIN-3[tiab] OR ALL-PO[tiab])",
"count": 41,
"usable": false,
"reason": "hyphenated short-token parser artefacts; hits include SARS-CoV-2 spike, animal genomic kinship, coriander powder fish feed, pulse oximeter bench tests"
},
{
"id": "D",
"query": "(CRISPR[tiab]) AND (PER-490[tiab] OR PER-485[tiab] OR PER-785[tiab] OR ALL-PO[tiab] OR SCMC-L1[tiab] OR KOCL[tiab])",
"count": 0
},
{
"id": "H",
"query": "(patient-derived xenograft[tiab] OR patient derived xenograft[tiab] OR PDX[tiab]) AND infant[tiab] AND (KMT2A[tiab] OR MLL[tiab])",
"count": 10
}
]- Because the absence is an absence of testing, it cannot be cited against any dependency claim. Citing it that way would repeat the error this project logged when an unmeasured variable was read as a measured null.
- If a screen is eventually run and reported only for the genes the investigators cared about, the absence will persist for every other gene while appearing to have been closed.
- Does a full-text search engine (Europe PMC full text, or Google Scholar) return a screen in any of these lines that a [tiab] query cannot see? That is the obvious next check and would either close or strengthen this absence.
- Has any of the Perth, KOCL or Hopkins material been used in an unpublished screen? A negative in the literature is not a negative in the field.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
PubMed esearch on 2026-09-19 returns zero records for a CRISPR/genome-wide/dependency-screen term crossed with infant-leukaemia terms, and zero records for a CRISPR term crossed with the identifiers of the infant KMT2A::AFF1 cell lines found in this beat; a query on the eight Perth line names alone returns one record, and a query on the Japanese and other infant line names returns 41 records dominated by parser artefacts unrelated to leukaemia.
PubMed esearch counts for four stated queries on infant leukaemia dependency screening, run 2026-09-19 · Oncofusion knowledge graph, Q21 index-fusion-models beat · NCBI Entrez eutils (esearch, db=pubmed) 2026What this evidence cannot show
A [tiab] query cannot see a screen described only in a paper's methods, and cell-line names overwhelmingly appear in methods rather than in abstracts — which query B demonstrates by returning a single record for eight lines that have at least four published preclinical programmes between them. So this absence is bounded by abstract-level indexing, and the honest strength of it comes from the DepMap check in E2, not from the PubMed counts alone. Aliases searched: PER-xxx line names, KOCL-xx names, SCMC-L1, BLIN-3, ALL-PO; NOT searched: institutional catalogue numbers or the lines' synonyms (PER-490A, PER-485A, PER-785A, SCMC-L-1, SCMCL-L1).
Evidence that argues against this (1)
- The same beat found a near-miss that a slightly different query would have caught differently: a 2023 report establishing a new t(11;19) KMT2A-rearranged B-ALL cell line for preclinical evaluation, and a 2023 report generating an inducible KMT2A-AFF1 iPSC line, both surfaced only from a cell-line-establishment query and neither from any CRISPR query. Resource papers are indexed by what they made, not by what can be done with them, so a capability-keyed absence systematically misses resource-keyed literature. Takahashi M 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
The DepMap side of the same absence is independently checkable and agrees: a query of DepMap CRISPR gene-effect screens for MEN1 through the Open Targets Platform returned 1,258 screened cell lines on 2026-09-19, including all five expected KMT2A-rearranged lines, and did not include ALL-PO (ACH-002209), the one infant-derived KMT2A::AFF1 line in the DepMap collection.
DepMap CRISPR (Chronos) gene-effect screens for MEN1 (ENSG00000133895), queried via the Open Targets Platform GraphQL API · Broad Institute Cancer Dependency Map; Open Targets Platform · Open Targets Platform / DepMap 2026What this evidence cannot show
Absence from one gene's screen list is not the same as absence from the screen collection, and the Open Targets release lags the DepMap portal. Five positive controls returning values in the same call are what make this negative interpretable. No other infant KMT2A::AFF1 line is in the DepMap collection at all, so this check covers one of the ten infant lines and simply does not apply to the other nine.
Evidence that argues against this (1)
- A negative built on one API endpoint is the exact exposure this project's own load-bearing-citation audit flagged: 219 evidence items across 167 entries already run through this single Open Targets endpoint, and the audit's first keying collapsed sixteen internal analyses sharing that URL into one source. The correct check is the DepMap portal's own model list, which this beat did not query. Bairoch A et al. 2025
bounded absence underpinning the verdict's 'screening gap, not derivation gap'
no genome-scale loss-of-function data exists in infant index-fusion material
bounded rejections and unreadables alongside the bounded absence