Oncofusion

start / graph / ana-q99-read-queue-bounded-rejections

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q99-interactome-read-queue

Two of seven interactome read-queue partners should not get a gene entry, and the reasons are a pseudokinase with no nuclear pool and a membrane-skeleton protein whose two leukaemia hits are a gene-name artefact

also called CAMKV bounded rejection · EPB41L1 bounded rejection · Q99 read-queue adjudication

Seven proteins were nominated for their own entries because databases list them as touching the leukaemia machinery. Two of them should not get one, and this records exactly why, so nobody has to repeat the work.

The same thing, for a scientist

🟠 **ARM-COMPOSITION CAVEAT ADDED 2026-09-24 (cycle 140).** This entry's computation rests on `analyses/depmap-kmt2a-selective-scan-full`, and that scan **fixes both comparison arms from the first gene that returns screen data** (`run.py:170`, `if rear is None:`; line 174 prints *"Groups fixed from first gene with data"*). Because DepMap screens each gene in a gene-specific set of lines, **5 of 16 eligible KMT2A-rearranged lines are excluded from every gene comparison, including genes where they were screened** - 16 resolve under the scan's own query, 11 were used. **This does not show any conclusion here is wrong.** Dropping 5 of 16 costs power, and it biases a ranking only if the dropped lines differ systematically from the kept ones, **which nobody has tested.** The named one-line fix is to derive the arms from each gene's own screen list. Until the scan is re-run, selectivity ranks taken from it carry an unquantified arm-composition caveat. Found by `analyses/depmap-scan-line-identities` on 2026-09-21 and unreported until now; see `ana-the-depmap-selectivity-scan-fixes-its-arms-from-the-first-gene-and-drops-five-of-sixteen-lines-from-every-comparison`. --- Of the seven partners in the cycle-54 interactome read queue, five were written as entries and **two are rejected here with their reasoning**: `CAMKV` and `EPB41L1`. Both fail on the same four counts and for independently verified reasons. (1) **No nuclear pool.** CAMKV's only UniProt locations are cell membrane and cytoplasmic vesicle membrane, and both are `ECO:0000250` by similarity - not human measurements [E1]; EPB41L1's only location comment, cytoplasm-cytoskeleton, carries **no evidence code at all** [E2]. (2) **No catalytic or reader module, and for CAMKV no catalytic activity of any kind** - UniProt's function line reads "Does not appear to have detectable kinase activity" and the kinase domain at 24-286 is annotated as predicted catalytically inactive [E1]. (3) **Not a dependency in any line**: CAMKV mean DepMap gene effect -0.0215, EPB41L1 +0.0087 across 1,208 lines [E3]. (4) **The interactome evidence is bait-side overexpression AP-MS in non-haematopoietic lines, recovering members of one complex.** Both recover four or five subunits of the super elongation complex, and in every row the rejected protein is the bait [E4]. Zero literature connects either to KMT2A, to a histone mark or to leukaemia biology: `CAMKV[tiab] AND leukemia[tiab]` = 0, `CAMKV[tiab] AND histone[tiab]` = 0, `EPB41L1[tiab] AND histone[tiab]` = 0, `EPB41L1[tiab] AND nuclear[tiab]` = 0 [E5]. **Two identifier traps were found doing this and both are registry-grade**: CAMKV's IntAct record is split between the canonical accession and isoform-2, so an exact-accession filter sees half of it [E4]; and EPB41L1's two apparent leukaemia hits are a gastric-cancer array paper whose abstract contains the word leukaemia only inside the legacy name *"v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 4 (ERBB4)"*, and a 2003 microarray of integrin-ligated precursor-B leukaemia cells in which EPB41L1 is one of six validated adhesion genes [E5][E6].

Go deeper
The rejection table
testCAMKVEPB41L1
accession, lengthQ8NCB2, 501 aaQ9H4G0, 881 aa
nuclear poolnone; cell membrane + cytoplasmic vesicle membrane, both ECO:0000250 (by similarity to rodent) 1none; cytoplasm-cytoskeleton with no evidence code at all 2
catalytic activitynone - "Does not appear to have detectable kinase activity"; kinase domain 24-286 predicted inactive 1none; FERM 97-378, spectrin-actin 483-541, CTD 746-881 2
chromatin modulenone 1none 2
DepMap gene effect, 1,208 lines-0.0215 (KMT2A-r -0.0514, other leukaemia -0.0479) 3+0.0087 (KMT2A-r +0.0270, other leukaemia +0.0144) 3
IntAct publications3 (28514442, 33961781, 40205054) 42 (33961781, 40205054) 4
IntAct baits recoveredAFF1, AFF4, ELL, MLLT1, MLLT3 - five SEC subunits 4AFF1, AFF4, MLLT1, MLLT3 - four SEC subunits 4
directionCAMKV is the bait in all 12 rows 4EPB41L1 is the bait in all 5 rows 4
hostsHEK293T, U2OS 4HEK293T, U2OS 4
X[tiab] AND leukemia[tiab]0 52, both artefactual or incidental 56
X[tiab] AND histone[tiab]0 50 5
total records30 526 5

What the rejection is not. It is not a claim that the IntAct rows are false. It is a claim that a protein with no nuclear pool, no enzymatic activity, no chromatin module, no dependency phenotype and no leukaemia literature, whose only connection to this project is bait-side co-purification with one complex in two non-haematopoietic lines, does not yet constitute a thing this graph should assert. The rows are recorded here and remain findable.

And the cost asymmetry that decided it. The PI's instruction is one entry per thing. Seven thin entries would each carry a shown_for_fusion: NO, a subnuclear: unpublished and an epigenetic_readout: not applicable that nobody could act on, and would raise every coverage line by seven while making the graph less true. Five substantive entries and two documented rejections is the honest split.

CAMKV: a pseudokinase, and the strongest single fact against it is in UniProt's function line

UniProt Q8NCB2's FUNCTION comment is one sentence: "Does not appear to have detectable kinase activity", and the DOMAIN comment states that "the protein kinase domain is predicted to be catalytically inactive" 1. CAMKV is a CAMK-family pseudokinase: a 501-residue protein with a kinase fold at 24-286 that does not catalyse, a disordered tail 327-501, calmodulin binding in the presence of calcium, and localisation "predominantly observed in association with the plasma membrane of soma and in neurites, both axons and dendrites" - explicitly by similarity 1.

So the protein is (a) neuronal, (b) peripheral-membrane, (c) enzymatically dead, and (d) its localisation is inferred from a rodent orthologue rather than measured in human cells. Its literature matches: 30 records total, 3 of which use the word pseudokinase, 3 dendritic, 9 cancer, 0 leukaemia, 0 histone, and exactly 1 containing the word nuclear, which was not opened 5.

The identifier trap, and it is registry-grade. IntAct curates CAMKV's BioPlex 2.0 rows under the isoform accession Q8NCB2-2 and its BioPlex 3.0 rows under the canonical Q8NCB2 4. A filter matching uniqueId == "Q8NCB2" - the obvious way to write it, and the way this beat wrote it first - returns 6 of the 12 rows and silently drops BioPlex 2.0 entirely. The first pass of this beat's own sweep reported CAMKV with two publications instead of three for exactly that reason, and it was caught only by cross-checking against the PSICQUIC rows, which show uniprotkb:Q8NCB2-2 in column 1. This is the same family as rule 93 - a processed chain or isoform is a different identifier from its parent - with an isoform rather than a processed chain, and it halves a count rather than zeroing it, which is why it is harder to notice.

What would reopen CAMKV. A reported nuclear pool in any human cell, a measured catalytic activity, or a reciprocal pulldown in which a SEC component is the bait and CAMKV is the prey. Any one of the three, and the rejection should be revisited.

EPB41L1: its two leukaemia records, read

EPB41L1[tiab] AND leukemia[tiab] returns 2, and this is the bounded-rejection part that a future cycle should not have to redo 5.

PMID 25368240 - a high-density array comparative genomic hybridisation study of intestinal-type gastric adenocarcinoma, 22 samples on a CytoScan HD array, which reports loss of heterozygosity of EPB41L1 at 20q11 in samples with peritoneal invasion 6. The word "leukemia" in that abstract comes from a different gene's legacy name: the abstract lists "v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 4 (ERBB4)". This is a contaminated-token finding worth the registry in its own right: leukemia[tiab] matches the legacy full names of the ERBB family and of other viral-oncogene-homolog genes, so any GENE[tiab] AND leukemia[tiab] query can return pure epithelial-cancer papers with no haematological content at all.

PMID 12393420 - a temporal gene-expression profile of human precursor-B leukaemia cells after beta-1 integrin ligation with fibronectin versus poly-L-lysine adhesion, in which EPB41L1 is one of six differentially expressed genes validated by real-time quantitative PCR, grouped with VAV2 and CORO1A under "adhesion" 7. This one is genuinely in the right disease and the right lineage, and it is the reason this rejection is bounded rather than flat: EPB41L1 is an integrin-adhesion-responsive transcript in precursor-B leukaemia cells. It is still one gene in a 38-gene microarray list from 2003, with no protein-level, localisation or mechanism follow-up, and no connection to KMT2A.

The localisation gap is the decisive one. Q9H4G0's single location comment is "Cytoplasm, cytoskeleton" with an empty evidence list - neither ECO:0000269 nor ECO:0000250, nothing 2. And EPB41L1[tiab] AND nucleus[tiab] = 0, AND nuclear[tiab] = 0 5. The alias 4.1N does return five records containing "nuclear" 5, which were not opened and are recorded as not opened - so the alias control fired and was not followed through, which is stated rather than hidden.

What would reopen EPB41L1. Any of: those five 4.1N-and-nuclear records establishing a nuclear pool; a follow-up on the precursor-B integrin result at protein level; or a reciprocal pulldown with a SEC component as bait.

What was read and not used, so the search is not repeated blind

Recorded per the standing instruction that a bounded rejection leaves a trace.

Judged off-beat after reading the abstract:

  • PMID 25368240 - gastric adenocarcinoma aCGH. On-beat only through a gene-name artefact (above) 6.
  • PMID 27462403 - mature-adipocyte nuclei isolation (MA-INTACT mice). Returned by ILF3 AND histone; a method paper about adipose epigenomes.
  • PMID 23856356 - review of spermatozoal RNAs. Returned by ILF3 AND histone.
  • PMID 34012635 - gastric-cancer chemoresistance microarray. Returned by ILF3 AND histone.
  • PMID 31747963 - ERp57/STAT3/ILF3 loop in clear-cell renal carcinoma. Returned by ILF3 AND chromatin; a different disease and a different mechanism.
  • PMID 20054697 - yeast med19/Set2 and phospholipid biosynthesis. One of only two MED19 AND histone hits; cross-species and about inositol biosynthesis, so not used as a human epigenetic claim.

Found, on topic, and NOT cited because it is retracted:

  • PMID 34141456 - lncRNA ILF3-AS1 promotes proliferation and metastasis of colorectal cancer cells by recruiting histone methylase EZH2, Mol Ther Nucleic Acids 2021. PubMed publication types include Retracted Publication. It is the ONLY record returned by ILF3[tiab] AND H3K27[tiab], and it is about the antisense lncRNA at the locus rather than the protein. Named so no later cycle cites it 5.

Adjudicated by the dispatch brief and re-verified rather than re-litigated: H3C1 (a histone, which belongs in this graph as marks and interfaces), AFG2A/AFG2B (SPATA5/SPATA5L1 out of PMID 38554706) and YWHAG (covered by a six-entry 14-3-3 axis). No independent work was done on these three and none is claimed.

Facts and figures
scriptnone (database queries: IntAct REST + PSICQUIC, UniProt REST, HGNC REST, NCBI E-utilities, plus analyses/depmap-kmt2a-selective-scan-full/results.json)
statusrun
queue size7
rejection is nota claim that the IntAct rows are false. Both rows now have TWO platforms behind them (BioPlex plus the 2025 U2OS multimodal map), which is better replication than several partners that received entries in earlier cycles. The claim is that a protein failing all four tests above does not yet constitute a thing this graph should assert.
result summaryFive of seven read-queue partners written as entries; CAMKV and EPB41L1 rejected on four independently verified counts each, with named reopening conditions. Four identifier traps found, three of them registry-grade: an isoform-split IntAct record that halves a count, leukemia[tiab] matching legacy viral-oncogene-homolog gene names, IntAct accession search matching by substring, and the 4.1 paralogue alias family.
absence classmeasured-not-curated
arm composition caveat cycle 140source scan fixes both arms from the first gene with data (run.py:170); 5 of 16 eligible KMT2A-rearranged lines are excluded from every comparison. Effect on this entry's conclusion untested.
inputs
[
  "IntAct REST findInteractions, retrieved 2026-09-22",
  "IntAct PSICQUIC MITAB tab25, retrieved 2026-09-22",
  "UniProt REST, retrieved 2026-09-22",
  "HGNC rest.genenames.org, retrieved 2026-09-22",
  "NCBI E-utilities esearch and efetch, run 2026-09-22",
  "analyses/depmap-kmt2a-selective-scan-full/results.json, computed 2026-09-19"
]
written as entries
[
  "gene-med19",
  "gene-pip4k2a",
  "gene-ilf3",
  "gene-bud13",
  "gene-epb41l3"
]
rejected
[
  "CAMKV",
  "EPB41L1"
]
rejection criteria all four failed
[
  "no nuclear pool reported (CAMKV: two ECO:0000250 membrane locations only; EPB41L1: one uncoded cytoplasm-cytoskeleton comment)",
  "no catalytic or chromatin-reader module (CAMKV additionally has no catalytic activity at all - UniProt: 'Does not appear to have detectable kinase activity')",
  "not a dependency in any DepMap line (CAMKV -0.0215, EPB41L1 +0.0087 over 1,208 lines)",
  "interactome evidence is bait-side overexpression AP-MS in non-haematopoietic lines recovering subunits of one complex"
]
reopening conditions
{
  "CAMKV": "a reported nuclear pool in any human cell; a measured catalytic activity; or a reciprocal pulldown with a SEC component as bait and CAMKV as prey",
  "EPB41L1": "a nuclear pool established from the five unopened 4.1N-and-nuclear records; protein-level follow-up of the precursor-B integrin result (PMID 12393420); or a reciprocal pulldown with a SEC component as bait"
}
identifier traps found
[
  {
    "token": "Q8NCB2 / Q8NCB2-2 (CAMKV)",
    "class": "one protein's interaction record split between a canonical accession and an ISOFORM accession",
    "measured_2026_09_22": {
      "findInteractions/Q8NCB2 rows matching a core bait, exact-accession filter": 6,
      "same with isoform-tolerant filter (strip trailing -N)": 12,
      "publications seen by exact filter": 2,
      "publications seen by isoform-tolerant filter": 3
    },
    "control": "PSICQUIC MITAB column 1 reads uniprotkb:Q8NCB2-2 for the BioPlex 2.0 rows and uniprotkb:Q8NCB2 for BioPlex 3.0, which is how the split was diagnosed rather than assumed.",
    "note": "Same family as the project's rule 93 (a processed chain is a different identifier from its parent) but with an ISOFORM, and it HALVES a count rather than zeroing it, which is why it survives casual inspection. This beat's own first sweep hit it."
  },
  {
    "token": "leukemia[tiab]",
    "class": "the query word matches the LEGACY FULL NAME of unrelated genes",
    "measured_2026_09_22": {
      "EPB41L1[tiab] AND leukemia[tiab]": 2,
      "of which purely epithelial-cancer papers matching only via a gene's legacy name": 1
    },
    "example": "PMID 25368240, gastric adenocarcinoma aCGH, matches because its abstract writes 'v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 4 (ERBB4)'.",
    "note": "Affects every ERBB-family and viral-oncogene-homolog legacy name (erythroblastic leukemia viral oncogene homolog, myeloblastosis viral oncogene homolog). A GENE[tiab] AND leukemia[tiab] query can return pure solid-tumour papers with no haematological content, which inflates the apparent leukaemia literature of any gene co-mentioned with one of them."
  },
  {
    "token": "IntAct interactor accession search",
    "class": "matches by SUBSTRING, so a shorter AC returns a longer AC's interactions",
    "measured_2026_09_22": {
      "findInteractions/EBI-2620048 rows returned": 7,
      "IntAct's own interactionCount for EBI-2620048": 6,
      "spurious row": "S100B-NEDD8 from PMID 32814053, whose interaction AC is EBI-26200486 and contains the queried string"
    },
    "control": "The spurious rows are identifiable because NEITHER participant is the queried interactor; the same collision adds a SPATS1-SARS1 row to a query on EBI-2620075.",
    "note": "Registry-worthy for any beat querying IntAct by accession rather than by UniProt id."
  },
  {
    "token": "EPB41L1 / EPB41L3 and the 4.1 paralogue family",
    "class": "six FERM paralogues with overlapping numeric aliases (4.1R, 4.1N, 4.1G, 4.1B)",
    "accessions": {
      "EPB41L1": "Q9H4G0 (881 aa)",
      "EPB41L3": "Q9Y2J2 (1,087 aa)"
    },
    "note": "This beat's first IntAct sweep used Q9Y2J2 for EPB41L1 from recall, which silently returned EPB41L3's record. The 206-residue length gap would be caught by a length check; the ALIAS confusion (4.1N versus 4.1B) would not. Accessions were then resolved programmatically from UniProt gene_exact plus HGNC."
  }
]
read and not used
[
  {
    "pmid": "25368240",
    "reason": "gastric adenocarcinoma aCGH; on-beat only through the ERBB4 legacy-name artefact",
    "what_would_change_it": "an EPB41L1 copy-number or LOH finding in a leukaemia cohort"
  },
  {
    "pmid": "27462403",
    "reason": "mature-adipocyte nuclei isolation method; returned by ILF3 AND histone",
    "what_would_change_it": "nothing - a method paper in a different tissue"
  },
  {
    "pmid": "23856356",
    "reason": "review of spermatozoal RNAs; returned by ILF3 AND histone",
    "what_would_change_it": "nothing"
  },
  {
    "pmid": "34012635",
    "reason": "gastric-cancer chemoresistance microarray; returned by ILF3 AND histone",
    "what_would_change_it": "nothing"
  },
  {
    "pmid": "31747963",
    "reason": "ERp57/STAT3/ILF3 loop in clear-cell renal carcinoma; different disease and mechanism",
    "what_would_change_it": "an ILF3 transcriptional-loop result in a haematopoietic cell"
  },
  {
    "pmid": "20054697",
    "reason": "yeast med19/Set2 and phospholipid biosynthesis; cross-species, about inositol biosynthesis, so not used as a human epigenetic claim",
    "what_would_change_it": "a mammalian MED19-SET2/SETD2 result"
  }
]
found on topic and not cited because retracted
[
  {
    "pmid": "34141456",
    "title_as_indexed": "lncRNA ILF3-AS1 promotes proliferation and metastasis of colorectal cancer cells by recruiting histone methylase EZH2.",
    "venue": "Mol Ther Nucleic Acids 2021",
    "pubmed_publication_type": "Retracted Publication",
    "why_it_matters": "It is the ONLY record returned by ILF3[tiab] AND H3K27[tiab], and it concerns the antisense lncRNA at the ILF3 locus rather than the ILF3 protein. Named so no later cycle cites it."
  }
]
adjudicated by the brief and not re litigated
[
  "H3C1 - a histone, which belongs in this graph as marks and interfaces rather than as a gene entry",
  "AFG2A / AFG2B - SPATA5 / SPATA5L1, from a replisome-proteostasis TAP preparation (PMID 38554706)",
  "YWHAG - already covered by a six-entry 14-3-3 axis"
]
How this could bite us back
  • A documented rejection is easier to quote than to re-open, and this one will be read as 'CAMKV and EPB41L1 are not relevant' rather than as 'four tests failed and here are the three things that would change it'. The reopening conditions are the load-bearing part of this entry.
  • Both rejected proteins have two independent platforms behind their super-elongation-complex recovery, which is better replication than several partners that DID receive entries in earlier cycles. If the bar applied here were applied retrospectively, some existing entries would not clear it - and this entry does not propose applying it retrospectively.
  • The strongest argument used here against both proteins is the absence of a nuclear pool, and for EPB41L1 that absence is an UNCODED UniProt comment with no traceable experiment. A rejection resting on a curator's statement is weaker than one resting on a measurement, and that asymmetry is stated rather than hidden.
  • Rejecting two of seven raises no coverage line and closes no gap, so nothing in this project's reporting will register that this work was done. That is the specific reason it is written as an entry rather than left in a memo.
Still open
  1. Do the five unopened 4.1N-and-nuclear records establish a nuclear pool for EPB41L1? If any of them does, the EPB41L1 rejection should be revisited, and the cost of checking is five abstracts.
  2. Is CAMKV expressed at all in the DepMap leukaemia panel? Its near-zero gene effect is uninformative if the transcript is absent, and a single expression lookup would tell us whether that test carried any weight in the rejection.
  3. How many other partners in the 128-row all_missing list have their IntAct record split across isoform accessions? The CAMKV case halves a publication count silently, and nothing in this project's interactome analyses is isoform-tolerant.
  4. How much of this graph's apparent leukaemia literature for any gene comes from the ERBB legacy-name artefact? The measurement is a corpus-wide query and has never been run.
  5. Should a rejection like this be re-adjudicated automatically when a new platform deposits? Both rejected proteins gained a third and second publication respectively between the queue's PSICQUIC snapshot and this REST query, without anybody noticing.
Every source behind this entry (7)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt Q8NCB2 is the single reviewed human CAMKV record at 501 aa (54,354 Da) whose FUNCTION comment states it does not appear to have detectable kinase activity and whose DOMAIN comment states the protein kinase domain (24-286) is predicted to be catalytically inactive; its only two locations - cell membrane and cytoplasmic vesicle membrane, both peripheral - carry ECO:0000250 (by similarity) and no experimental code, and there is no nucleus comment.

    What this evidence cannot show

    Retrieved 2026-09-22. Three isoforms are annotated (Q8NCB2-1, -2, -3) and 501 aa is the canonical. 'Does not appear to have detectable kinase activity' is a curator summary of the primary literature, not an assay this entry read, and 'predicted to be catalytically inactive' is explicitly a prediction. Absence of a nucleus comment is absence of an annotation. The alias control on VACAMKL returns 0 PubMed records, so it is uninformative rather than reassuring; HGNC records no previous official symbol for this gene.

    Evidence that argues against this (1)
    • A pseudokinase can be functionally important without catalysis - pseudokinases act as allosteric regulators, scaffolds and nucleotide sensors - so 'no detectable kinase activity' is not an argument that a protein does nothing, only that its name misdescribes it. UniProt Consortium 2025
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt Q9H4G0 is the single reviewed human EPB41L1 record at 881 aa (98,503 Da) with a FERM domain (97-378), a spectrin-actin-binding region (483-541) and a C-terminal domain (746-881), whose ONLY subcellular-location comment - Cytoplasm, cytoskeleton - carries an EMPTY evidence list with no ECO code of any kind, with no nucleus comment, highest expression in brain, and one annotated interaction partner (AGAP2).

    What this evidence cannot show

    Retrieved 2026-09-22. Four isoforms are annotated and 881 aa is the canonical. An uncoded UniProt location comment is not a measurement and this graph has already measured that 10.7% of its entries rest on such comments - so 'cytoplasm, cytoskeleton' here is a curator statement with no traceable experiment, which is a weaker basis for a rejection than a positive experiment would be, and is stated as such. EPB41L1 is Q9H4G0 and EPB41L3 is Q9Y2J2; this beat's first IntAct sweep used Q9Y2J2 for EPB41L1 from recall and the accessions were then resolved programmatically from UniProt gene_exact and HGNC.

    Evidence that argues against this (1)
    • The alias control fired and was not followed through: 4.1N[tiab] AND nuclear[tiab] returns five records, which were not opened. A rejection that rests partly on the absence of nuclear literature under the approved symbol is therefore weaker than it looks, and this is exactly the alias-recall failure that changed the epigenetic call for the paralogue EPB41L3 in the same tranche. NCBI PubMed 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Neither protein is a dependency in any cell line: CAMKV mean DepMap gene effect -0.0215 across 1,208 lines (-0.0514 in 11 KMT2A-rearranged versus -0.0479 in 43 other leukaemia lines), EPB41L1 +0.0087 across 1,208 lines (+0.0270 versus +0.0144) - all values within 0.06 of zero.

    What this evidence cannot show

    These are floor-compressed measurements: when every group mean is within 0.06 of zero the between-group difference is a difference between two nulls, so the scan RANKS (CAMKV 8,698; EPB41L1 10,787 of 17,728) carry no information and are deliberately not quoted as findings - roughly half of any gene set lies above the median by construction and this project's base rate is 45.3%. A knockout screen also cannot see a gene not expressed in the panel, and CAMKV is a neuronal protein. The stored scan records a compute date and NOT a DepMap release.

    Evidence that argues against this (1)
    • A near-zero CRISPR gene effect is uninformative for a gene that is not expressed in the screened lines, and CAMKV's documented expression is neuronal - so the DepMap null for CAMKV may be measuring absence of the transcript rather than dispensability of the protein, which means this test carries less weight in the rejection than the localisation and activity arguments do. UniProt Consortium 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    IntAct REST serves twelve CAMKV rows over three publications (28514442, 33961781, 40205054) and five EPB41L1 rows over two (33961781, 40205054), all MI:0007 anti-tag co-immunoprecipitation typed MI:0914 association and spoke-expanded in HEK293T or U2OS, recovering five and four super-elongation-complex subunits respectively, with the rejected protein the BAIT in every row - and CAMKV's record is split between the canonical accession Q8NCB2 and the isoform accession Q8NCB2-2, so an exact-accession filter returns six of the twelve rows and silently drops BioPlex 2.0.

    What this evidence cannot show

    Retrieved 2026-09-22 and the counts will move. The isoform split was found because this beat's own first sweep matched on exact accession and reported CAMKV with two publications instead of three; it was caught by cross-checking PSICQUIC MITAB column 1, not by the REST query. Note this cuts BOTH ways for the rejection: the corrected count is HIGHER, so CAMKV has three publications rather than two, which strengthens the row while leaving every other test failed. Counting four or five SEC subunits as four or five partners double-counts one complex.

    Evidence that argues against this (1)
    • A third independent platform is a real strengthening of these rows, not noise: PMID 40205054 is a 5,100-protein U2OS multimodal AP-MS and immunofluorescence map from a different laboratory than BioPlex, so CAMKV's and EPB41L1's SEC recovery has been seen on two platforms in two cell lines - which is better replication than several partners that DID receive entries in earlier cycles. Schaffer LV et al. 2025
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Neither protein has any leukaemia, KMT2A or histone literature: CAMKV[tiab] AND leukemia[tiab] = 0, AND histone[tiab] = 0, AND nuclear[tiab] = 1, AND pseudokinase[tiab] = 3, total 30 records, VACAMKL[tiab] = 0; EPB41L1[tiab] AND histone[tiab] = 0, AND nucleus[tiab] = 0, AND nuclear[tiab] = 0, AND MLL[tiab] = 0, total 26 records, with the alias 4.1N[tiab] AND nuclear[tiab] = 5 not opened; and the only ILF3-AND-H3K27 record in PubMed, PMID 34141456, carries the PubMed publication type Retracted Publication.

    What this evidence cannot show

    All queries are one tag per term and never a parenthesised group; each querytranslation was read back and all were clean. Positive controls through the identical code path: CAMKV[tiab] = 30, CAMKV[tiab] AND cancer[tiab] = 9, EPB41L1[tiab] = 26, H3K79me2[tiab] = 224 - so the zeros are not instrument failures. VACAMKL[tiab] = 0 means the alias is absent from PubMed entirely, which makes that control uninformative rather than negative. These zeros bound titles and abstracts only; a supplementary mass-spectrometry table naming either protein in a KMT2A model would be invisible to all of them. The one CAMKV-and-nuclear record and the five 4.1N-and-nuclear records were NOT opened, and that is recorded rather than glossed.

    Evidence that argues against this (1)
    • A gene-symbol literature count is a poor instrument for a protein whose founding papers predate its approved symbol, as this very tranche demonstrated: the two papers establishing that the paralogue EPB41L3 binds and inhibits an arginine methyltransferase are wholly unreachable from EPB41L3[tiab] AND PRMT[tiab], which returns 0 while the alias query returns the paper. Singh V et al. 2004
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    One of EPB41L1's two apparent leukaemia records is a gene-name artefact: PMID 25368240 is a high-density array comparative genomic hybridisation study of 22 intestinal-type gastric adenocarcinomas reporting loss of heterozygosity of EPB41L1, and the word leukaemia appears in its abstract only inside the legacy gene name 'v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 4 (ERBB4)'.

    High-density array comparative genomic hybridization detects novel copy number alterations in gastric adenocarcinoma. · Seabra AD, Araujo TM, Mello Junior FA, et al. (11 authors; senior author Khayat AS) · Anticancer Res 2014
    PMID 25368240
    What this evidence cannot show

    Read as abstract and title. The EPB41L1 finding in this paper is an LOH call from a 22-sample array study in a different disease, reported among 22 alterations described as novel for gastric cancer, with no functional follow-up. The generalisable point is the query artefact: leukemia[tiab] matches the legacy full names of ERBB-family and other viral-oncogene-homolog genes, so a GENE[tiab] AND leukemia[tiab] query can return pure epithelial-cancer papers with no haematological content.

    Evidence that argues against this (1)
    • A word-level artefact in one abstract does not make a symbol's whole leukaemia count artefactual, and in this case the second of the two records is genuinely in precursor-B leukaemia cells - so the correct reading is that this symbol's leukaemia literature is one real record and one artefact, not zero. Astier AL et al. 2003
  7. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    EPB41L1's one genuine leukaemia record places it as an integrin-adhesion-responsive transcript in the right lineage: in serum-starved human precursor-B leukaemia cells, temporal expression profiling after beta-1 integrin ligation with fibronectin versus poly-L-lysine adhesion identified 38 differentially expressed genes, of which six were validated by real-time quantitative PCR - EPB41L1 among them, grouped with VAV2 and CORO1A under adhesion.

    PMID 12393420DOI
    What this evidence cannot show

    A 2003 microarray study; EPB41L1 is one transcript of 38 selected and one of six validated by RT-qPCR, with no protein-level, localisation or mechanistic follow-up in that paper or since, and no connection to KMT2A or to any fusion. The system is precursor-B leukaemia cells adhering to fibronectin, which is the right lineage but a stromal-adhesion question rather than a transcription question. Read as abstract and title. This record is the reason the EPB41L1 rejection is written as bounded and reopenable rather than flat.

    Evidence that argues against this (1)
    • A transcript changing on integrin ligation is a plausible consequence of cytoskeletal remodelling in a membrane-skeleton gene and does not imply a nuclear or transcriptional role - which is what the rejection turns on. EPB41L1's only annotated location is cytoplasm-cytoskeleton, with no evidence code, and no nuclear pool is reported anywhere. UniProt Consortium 2026
Connected to