Oncofusion

start / graph / ana-q129-the-nucleolar-triple-survives-none-of-the-four-pre-registered-discriminators

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q129-the-never-discussed-partners-of-kmt2a-and-psip1

The ABT1/KRR1/UTP3 nucleolar triple: pre-registered, then adjudicated - it is not three proteins on two baits but a seven-member ribosomal block in one screen plus one hit in a neighbourhood that shows no nucleolar enrichment, and two of the four pre-registered discriminators were void exactly as the pre-registration predicted

also called nucleolar triple · ABT1 KRR1 UTP3 · Q129 pre-registration · nucleolar hypothesis adjudication

Three proteins that build ribosomes in the nucleolus showed up as partners of the two proteins this project studies, which could have meant those proteins do something in the nucleolus. Writing down the test before looking, then looking, the pattern turns out to be what the experiments themselves would produce by accident.

The same thing, for a scientist

**The hypothesis, as dispatched.** ABT1 and KRR1 (arriving on KMT2A) and UTP3 (arriving on PSIP1) are all nucleolar ribosome-biogenesis factors on two different baits. Three readings compete: **(a)** genuine nucleolar biology for KMT2A/PSIP1, which this graph tracks as an explicit PI dimension; **(b)** nucleolar and ribosomal proteins are classic AP-MS background; **(c)** coincidence at n=3. **Four discriminators were fixed in writing before any record was pulled** - D1 publication identity, D2 method orthogonality, D3 contaminant prior, D4 the source papers' own controls - together with an explicit statement of **what would make D1 and D2 uninformative**: if the whole thirteen-partner list traced to one or two publications and one method, then 'the nucleolar three share a screen' would be a fact about the list's composition and would say nothing about nucleolar specificity [E5]. **That clause fired, and reporting it is the point of writing a pre-registration.** The thirteen come from **four** publications and **four** methods, so the list is not fully homogeneous - but **BioPlex 3.0 alone supplies 6 of 13, including both ABT1 and KRR1**, so their sharing a screen is at the list's base rate and carries almost no information. **D1 is void as applied to ABT1 and KRR1, as predicted.** D2 is likewise weak: all thirteen are mass-spectrometry proximity or affinity methods and **not one has an orthogonal confirmation** anywhere in IntAct, so method is nearly constant across the list [E3]. **D3 and D4 carried the result, and both point to (b).** Inside BioPlex, the nucleolar pair is not a pair: **seven of the 24 baits that recover KMT2A are ribosomal or ribosome-biogenesis proteins and all seven sit at the 0.35 floor score, while the genuine WRAD core sits at 0.91-0.96** [E1]. Inside the K562 BioID2 map, **PSIP1's nine recovering baits contain two nucleolar ones against a panel base rate of nine in fifty - exact binomial p = 0.50, no enrichment** [E2]. **And one of the three is not securely nucleolar at all:** ABT1's human nucleolar annotation is **ECO:0000250, by similarity**, with no experimental code [E4] - which triggers the pre-registration's second uninformative condition, that a grouping supportable only from the annotation database that supplied it is an annotation artefact. **Verdict: reading (b), with the caveat that (a) is bounded rather than excluded.** No measurement here says KMT2A or PSIP1 has no nucleolar pool. It says these three records cannot be evidence that it does.

Go deeper
The pre-registration, quoted

Written to the session scratchpad as q129_prereg.md on 2026-09-23 before the first IntAct record was fetched 5:

D1. PUBLICATION IDENTITY. If all three trace to ONE publication, or to publications from one laboratory or one consortium running one method, reading (a) loses its 'two independent baits' support and (b) becomes the leading reading.

D2. METHOD ORTHOGONALITY. If every record is affinity-capture MS with no orthogonal method (two-hybrid, ITC, co-crystal, PCA, proximity labelling as a partial), (b) is favoured.

D3. CONTAMINANT PRIOR. Count each protein's total IntAct interaction record count. State the number, not a verdict word.

D4. WITHIN-PAPER CONTROL. Read each source paper's own stated controls and scoring. If the paper scores the interaction at or near its own threshold, or warns about ribosomal/nucleolar background, (b) is favoured.

WHAT WOULD MAKE THIS CRITERION UNINFORMATIVE. D1 is informative only RELATIVE TO A BASE RATE. If ALL THIRTEEN partners trace to the same one or two publications, then 'the nucleolar three share a screen' is a fact about the composition of the whole list and says NOTHING about nucleolar specificity... Same for D2: if all 13 are AP-MS, method is constant across the list and cannot discriminate.

Second uninformative condition. 'Nucleolar ribosome-biogenesis factor' is itself an annotation-derived label. If I can only support it from the same database family that supplied the interaction, the grouping is a GO-annotation artefact.

Both uninformative conditions fired, partially or fully. That is the most useful thing this analysis produced, because it means the obvious version of this test - 'do they share a screen?' - would have returned an answer that felt decisive and was not.

The four discriminators, scored
testresultverdict
D1publication identityABT1 + KRR1 from BioPlex 3.0; UTP3 from Khoroshkin BioID2. Base rate: BioPlex supplies 6 of the 13 partners on this listVOID for ABT1/KRR1 - at base rate. Retains weak force for UTP3, which is a genuinely separate publication, laboratory and method
D2method orthogonalityanti-tag co-IP (ABT1, KRR1) vs BioID2 proximity (UTP3). No orthogonal non-MS method for any of the thirteenWEAK - two MS methods are not orthogonal in the sense that matters; both capture proximity in a lysate or a labelling radius, neither measures binding
D3contaminant priorABT1 761 IntAct rows, KRR1 464, UTP3 295 - against WDR5 639 and MEN1 137. ABT1, a 272-residue protein with a hedged function annotation, has more records than WDR5FAVOURS (b), weakly - a raw row count measures participation in large screens as much as stickiness
D4within-paper controlsBioPlex: CompPASS-Plus vs GFP, but reciprocal confirmation structurally unavailable because KMT2A is a prey-only node among its 10,128 baits, and no HCT116 replication for either. Khoroshkin: the authors Z-score every prey across baits before analysis, an explicit correction for promiscuous preys, and the IntAct rows are the raw layer beneath itFAVOURS (b) - both papers' own machinery is designed to remove exactly this class, and IntAct exposes the layer before the removal

Two post-hoc tests, labelled as post-hoc, did the real work.

P1 - the block, inside BioPlex 1. Of the 24 baits that recovered KMT2A, seven are ribosomal or ribosome-biogenesis proteins - ABT1, KRR1, RPL36, RPLP0, RPS9, Rpl13a and the RPL3/RPL3L group, with PIH1D1 of the R2TP snoRNP chaperone arguably an eighth - and all of them sit at intact-miscore 0.35 while WDR5, RBBP5 and ASH2L sit at 0.96, 0.95 and 0.91. The nucleolar 'pair' on KMT2A is two visible members of a seven-member block. In the same 0.35 band sit ANOS1 and FGFBP1, both experimentally annotated as secreted - pairs that cannot be physiological, at the identical score, in the identical experiment. The band is not uniformly false, and this analysis does not claim it is: it also contains SET, a genuine histone chaperone and nucleosome-assembly protein, and the three SRSF splicing factors. The claim is about what the seven ribosomal members support collectively; 0.35 is not a synonym for artefact.

P2 - the base-rate test, inside the BioID2 map 2. PSIP1 is recovered by 9 of the 50 baits. Two of those nine (UTP3, SBDS) are nucleolar or ribosome-biogenesis; the panel itself is 9 of 50 nucleolar. Observed 22%, expected 18%, exact binomial p = 0.50. A generous classification of 11 of 50 makes the match exact. With n = 9 the test would have needed 5 of 9 to reach p < 0.05, so it excludes a large enrichment and not a modest one.

The precedent this project already had, and did not connect

This graph has already found one false nucleolar KMT2A partner, and it was a curation error. ana-utp18-is-not-in-the-mll-complex-it-is-wdr5-miscurated establishes that UTP18 - a nucleolar SSU-processome protein, UniProt synonym WDR50 - appears in IntAct's curation of PMID 15199122 (Yokoyama 2004, the founding paper for this graph's WRAD position) in the slot where the paper itself reports WDR5, WD repeat-containing protein 5 against WD repeat-containing protein 50 6.

So before this beat began, the score on nucleolar SSU-processome proteins in KMT2A's IntAct interactome was: one candidate examined, one found to be a mis-curated WDR5. That is not evidence about ABT1, KRR1 or UTP3 - it is a different failure mode - but it is a prior that nobody carried forward, and it is the reason the dispatch brief's framing of a 'nucleolar signal' deserved a pre-registration rather than a look.

And the larger precedent. ana-kmt2a-proximity-promiscuity-is-explained-by-size-97th-percentile-becomes-91st-among-proteins-its-own-length concluded, from the same BioID2 census, that KMT2A's apparent promiscuity is explained by its size and that the census cannot support individual proximity partners 2. This analysis is consistent with that and extends it to the PSIP1 side of the same dataset, where the size argument does not apply and the base-rate argument does.

What this does NOT establish, stated plainly

It does not establish that KMT2A or PSIP1 has no nucleolar pool. Nothing in this graph has ever measured that. UniProt gives KMT2A one coded Nucleus comment and two uncoded fragment comments and no nucleolar line; that is an absence in an annotation, not a measurement.

It does not establish that KRR1 or UTP3 is absent from a chromatin complex. Both are experimentally nucleolar, UTP3's own UniProt function comment opens with a chromatin-silencing role, and neither has ever been tested against KMT2A or PSIP1 with those proteins as bait.

It does not rule out a modest nucleolar enrichment in PSIP1's neighbourhood - n = 9 bounds only a large one.

The class of experiment that would settle it is nucleolar-fraction proteomics or imaging of KMT2A and PSIP1 in a KMT2A-rearranged leukaemia line. No archive query for such a dataset was run in this beat, so whether one already exists in PRIDE or in an imaging resource is unknown, and that is the cheapest next step - not wet work.

Facts and figures
scriptad hoc, not scripted: IntAct REST findInteractions paginated to exhaustion for Q03164, O75475, 39251607 and 30021884, with intactMiscore, experimentalRole, expansionMethod and detectionMethod read per row; UniProt REST for each partner; binomial computed analytically. Raw pulls are in the session scratchpad (q129_rest_*.json, q129_pub_*.json, q129_uniprot.json, q129_prereg.md) and are NOT persisted to the repository.
statusrun
hypothesis testedABT1, KRR1 (on KMT2A) and UTP3 (on PSIP1) indicate genuine nucleolar biology (a), AP-MS background (b), or coincidence (c)
verdict(b) favoured; (a) bounded, not excluded
absence classnever-measured
absence class noteWhether KMT2A or PSIP1 has any nucleolar pool has never been measured in this graph or found in any source read here. Closing it needs nucleolar-fraction proteomics or imaging in a KMT2A-rearranged leukaemia line. A cheaper first step that was NOT taken: query PRIDE and ProteomeXchange for deposited nucleolar-fraction proteomics from any leukaemia line, which would tell whether this is never-measured or measured-not-analysed. Until that query is run, this class is asserted rather than established.
inputs
[
  "IntAct REST findInteractions, 2026-09-23",
  "UniProt REST, 2026-09-23",
  "PMID 33961781 full text via PMC HTML (PMC8165030)",
  "PMID 39251607 full text via Europe PMC fullTextXML (PMC11385195)",
  "PMID 30021884 full text via PMC HTML (PMC6166682)",
  "q129_prereg.md, written before data collection"
]
prespecified discriminators
{
  "D1_publication_identity": "VOID for ABT1/KRR1 - at the list's base rate, exactly as the pre-registration predicted; weak force retained for UTP3",
  "D2_method_orthogonality": "WEAK - four MS methods across the thirteen, no orthogonal non-MS confirmation for any",
  "D3_contaminant_prior": "FAVOURS (b) weakly - ABT1 761 IntAct rows, KRR1 464, UTP3 295, against WDR5 639 and MEN1 137",
  "D4_within_paper_controls": "FAVOURS (b) - BioPlex's reciprocal check structurally unavailable and no HCT116 replication; Khoroshkin Z-scores every prey across baits and IntAct exposes the layer beneath that"
}
post hoc tests
{
  "P1_ribosomal_block": "7 of 24 BioPlex baits recovering KMT2A are ribosomal/ribosome-biogenesis, all at miscore 0.35, against WRAD at 0.91-0.96",
  "P2_nucleolar_base_rate": "PSIP1's 9 recovering baits contain 2 nucleolar against a panel base rate of 9/50; expected 1.62, observed 2, exact binomial p = 0.50; P<0.05 would need 5 of 9"
}
uninformative conditions that fired
[
  "D1's base-rate condition: BioPlex supplies 6 of 13, so ABT1 and KRR1 sharing a screen is at base rate",
  "the annotation-artefact condition, for ABT1 only: its nucleolar label is ECO:0000250 with no experimental code"
]
How this could bite us back
  • If KMT2A or PSIP1 does have a nucleolar arm, every readout used in the menin-inhibitor trials - HOXA9/MEIS1 expression, H3K79me2, chromatin occupancy at protein-coding targets - is blind to it, and a resistance mechanism living in the nucleolus would present as treatment failure with no assay pointing anywhere.
  • The intervention implied by reading (a) has no therapeutic window. Ribosome biogenesis is required by every proliferating cell, SSU-processome inhibition would hit normal haematopoietic progenitors at least as hard as leukaemic blasts, and this is a disease of infants. So even the favourable reading does not yield a target - which is a reason to demand strong evidence before pursuing it, not a reason to stop asking whether the compartment matters.
  • The method risk cuts back at this analysis: 'ribosomal proteins are contaminants' is a heuristic with a long history of being right and a real record of discarding true nucleolar biology. If this project hardens it into a filter, it will delete genuine nucleolar findings about a chromatin protein whose nucleolar status has never been measured.
Still open
  1. Compute PSIP1's baits-per-prey percentile WITHIN its own length decile in the K562 BioID2 map. KMT2A's 97th percentile fell to the 91st once size was held constant; PSIP1 is 530 aa and the same correction would push its 9 baits UP. This is the single calculation that could reopen the UTP3 row and it was not run.
  2. Query PRIDE and ProteomeXchange for deposited nucleolar-fraction proteomics from any leukaemia line. The absence_class on this entry is asserted as never-measured without that query having been run, which is precisely the gap this project's absence-class rule exists to close.
  3. Retrieve BioPlex's own CompPASS-Plus probabilities for the seven ribosomal KMT2A rows and for WDR5, RBBP5 and ASH2L. If the authors' own score does not separate a ribosomal protein from the WRAD core on the same prey, that is a finding about the dataset that matters far beyond this beat.
  4. Is there any published human experiment placing ABT1 in the nucleolus? UniProt cites none, its annotation is ECO:0000250, and the yeast Esf2 literature was not searched - so the third member of the 'triple' is currently a protein this project cannot confirm is nucleolar at all.
  5. Does KMT2A occupy rDNA? The nucleolar question for a chromatin methyltransferase is ultimately an occupancy question, and no ChIP or CUT&RUN dataset in this graph has been examined at rDNA repeats.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Of the 24 distinct BioPlex 3.0 baits that recover KMT2A as prey in IntAct, the intact-miscores are WDR5 0.96, RBBP5 0.95, ASH2L 0.91, CSNK2A2 0.79, CSNK2B 0.64, HCFC2 0.62, CSNK2A1 0.53, and 0.35 for the remaining seventeen. Seven of those seventeen are ribosomal or ribosome-biogenesis proteins - ABT1, KRR1, RPL36, RPLP0, RPS9, Rpl13a and the RPL3/RPL3L ambiguous group - with PIH1D1 of the R2TP snoRNP-assembly chaperone as a possible eighth. The same 0.35 band contains ANOS1 and FGFBP1, both UniProt-annotated Secreted at ECO:0000269. All 31 BioPlex rows touching Q03164 have KMT2A as prey, so KMT2A is not among BioPlex's 10,128 baits and reciprocal confirmation is structurally unavailable.

    What this evidence cannot show

    The ribosomal/nucleolar classification of the seven baits is THIS ANALYSIS'S, made from UniProt function and location comments, and another reader could draw the boundary differently - PIH1D1 is counted separately for that reason. intact-miscore is a pair-level score driven mainly by how many publications and methods support a pair, so the 0.35 floor largely restates 'one record of one kind' and is NOT an independent grading of these experiments; BioPlex's own per-pair CompPASS-Plus probabilities were not retrieved and would be the better instrument. The BioPlex supplementary tables were not opened.

    Evidence that argues against this (1)
    • The 0.35 band also contains SET, a genuine histone chaperone and nucleosome-assembly protein, and the three SRSF splicing factors. So the band is not uniformly false and 'floor score' is not a synonym for artefact. The block argument is about which reading the seven ribosomal members collectively support, not a claim that each of the seventeen is wrong. IntAct (EMBL-EBI) 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    In the Khoroshkin 50-RBP BioID2 map in K562 (PMID 39251607), IntAct REST serves 19,380 rows resolving to 50 baits and 2,682 preys, median 5 baits per prey, mean 7.23, maximum 40. PSIP1 is recovered by 9 baits (AGGF1, FASTKD2, HNRNPC, ILF3, PPIL4, RBM15, SBDS, U2AF1, UTP3), the 68th percentile of preys; KMT2A by 25, the top 2.9%. Nine of the 50 baits are nucleolar or ribosome-biogenesis proteins (DDX21, DDX52, GNL3, NOLC1, NPM1, RPS3, RPS11, SBDS, UTP3) = 18% of the panel; two of PSIP1's nine are in that set = 22%. Exact binomial with n=9, p=0.18 gives expectation 1.62 and P(X>=2) = 0.50; P<0.05 would require 5 of 9. The authors convert every bait-prey target score to a Z-score across baits before enrichment analysis, an explicit correction for promiscuous preys, and IntAct exposes the raw bait-prey layer beneath that step.

    What this evidence cannot show

    The nucleolar classification of 9 of 50 baits is this analysis's own; a generous version adding DDX55 and EXOSC5 gives 11 of 50 = 22%, against which the observed 22% is an exact match, so the conclusion is insensitive to the boundary. The test is POST-HOC and labelled so. The 50 baits are not independent draws - they are all RBPs and several nucleolar ones are in shared machinery - so the effective n is below 9 and the bound is weaker than stated. Source full text read 2026-09-23 via Europe PMC fullTextXML (PMC11385195); the strings UTP3, PSIP1, LEDGF and KMT2A appear zero times in it, and Data file S6 was not opened.

    Evidence that argues against this (1)
    • PSIP1 is 530 aa. The size correction that dissolved KMT2A's apparent promiscuity would run the OTHER WAY for a small protein: if large proteins are recovered by roughly twice as many baits as typical ones, a 530 aa protein recovered by 9 baits may be more promiscuous than its size predicts. That within-length-decile computation was NOT performed here, and it is the single calculation that could reopen the UTP3 row. Oncofusion Knowledge Graph (internal) 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The thirteen IntAct-curated human partners of KMT2A or PSIP1 never discussed anywhere in this corpus come from four publications and four detection methods: PMID 33961781 (BioPlex 3.0, anti-tag co-immunoprecipitation) supplies six - ABT1, ANOS1, FGFBP1, KRR1, OASL, PRKRA; PMID 30021884 (Fasci, DSSO crosslink) supplies five - GDF5, IGHMBP2, PPP1R13B, TMEM214 on PSIP1 and TMCO1 on KMT2A; PMID 30833792 (Hubel, pull down) supplies IRF2; PMID 39251607 (Khoroshkin, BioID2 proximity) supplies UTP3. Each of the thirteen has exactly one publication and one detection method in IntAct; none has an orthogonal non-mass-spectrometry confirmation.

    What this evidence cannot show

    The thirteen-partner list derives from analyses/core-complex-interactome-gap/results.json, which parses PSICQUIC and is therefore a floor (rule 93) - a partner visible only through REST would not be on it. All thirteen symbols plus nine common aliases were independently verified absent from the corpus by case-insensitive word-boundary grep over every file in data/graph/nodes on 2026-09-23, returning zero matches, so the 'never discussed' premise checks out even though the denominator may be incomplete. The count of 'one publication each' is IntAct's; a partner supported by another database and not by IntAct would be missed.

    Evidence that argues against this (1)
    • Four publications and four methods is more heterogeneity than the pre-registration's worst case, so D1 and D2 are weakened rather than fully void. In particular UTP3 arrives from a different laboratory, a different cell line, a different method and a different bait protein than ABT1 and KRR1, which is real independence and is the strongest structural argument the nucleolar hypothesis has. IntAct (EMBL-EBI) 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB Q9ULW3 (ABT1, 272 aa) gives Nucleus and Nucleus/nucleolus at ECO:0000250 - by similarity, with no experimental code - and its sole function comment is 'Could be a novel TATA-binding protein (TBP) which can function as a basal transcription activator... (By similarity)'. By contrast Q13601 (KRR1, 381 aa) gives Nucleus/nucleolus, Nucleus and Cytoplasm all at ECO:0000269 with an explicit SSU-processome function, and Q9NQZ2 (UTP3, 479 aa) gives Nucleus/nucleolus at ECO:0000269 with an SSU-processome function that opens 'Essential for gene silencing: has a role in the structure of silenced chromatin'. IntAct interaction record counts: ABT1 761, KRR1 464, UTP3 295, against WDR5 (P61964) 639 and MEN1 (O00255) 137.

    What this evidence cannot show

    Rule 91: an ECO:0000250 code is a statement about UniProt's curation, not about the literature - a human nucleolar experiment for ABT1 may exist uncited, and the ABT1 and yeast Esf2 primary literature was NOT searched in this beat. A raw IntAct row count is not a promiscuity metric: it counts records rather than distinct partners and is dominated by how many large screens a protein appeared in. A proper frequent-flier statistic would be a CRAPome-style frequency across matched negative controls; the CRAPome was not queried.

    Evidence that argues against this (1)
    • WDR5 carries 639 IntAct rows and is a genuine hub with dozens of real partners, so a high row count plainly does not imply artefact and D3 is the weakest of the four discriminators. It is reported as a number rather than a verdict for exactly that reason. IntAct (EMBL-EBI) 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    The four discriminators D1-D4, and two explicit conditions under which they would be uninformative, were written to the session scratchpad file q129_prereg.md on 2026-09-23 before the first IntAct record for any of the thirteen partners was fetched. Both uninformative conditions subsequently fired: BioPlex 3.0 supplies 6 of the 13 partners so D1's shared-screen observation for ABT1 and KRR1 is at the list's base rate, and ABT1's nucleolar label is supported only by an ECO:0000250 UniProt comment, which is the annotation-artefact condition.

    What this evidence cannot show

    THE PRE-REGISTRATION IS NOT INDEPENDENTLY VERIFIABLE. It was written to a session scratchpad directory that is not committed to the repository and is not timestamped by any third party, so a reader has only this entry's word that it preceded the data. The text is quoted in full above so the claim can at least be judged on its content, and the honest status is 'self-reported pre-registration' - confidence is set to moderate for that reason and not higher. A future beat wanting this to carry weight should commit the pre-registration to the repository at the start of the beat, which costs nothing and would make the claim checkable.

    Evidence that argues against this (1)
    • A pre-registration that only the author can see provides no protection against the bias it exists to prevent. The two post-hoc tests (P1 and P2) that actually produced this analysis's conclusion are labelled post-hoc precisely because the pre-registration did not cover them - which means the headline result rests on unregistered analysis, and the pre-registration's main demonstrated value here was predicting its OWN failure modes rather than constraining the result. Oncofusion Knowledge Graph (internal) 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    This graph already established that UTP18 (Q9Y5J1), a nucleolar SSU-processome protein whose UniProt synonym is WDR50, appears in IntAct's curation of PMID 15199122 in the slot where that paper reports WDR5, and that the correct reading of the KMT2A-bait gap it produced is zero. The entry ana-utp18-is-not-in-the-mll-complex-it-is-wdr5-miscurated records that the full text of PMID 15199122 contains no occurrence of UTP18, WDR50, U3, snoRNA, nucleolar or nucleolus, and that IntAct's curation of that paper lists eight participants including UTP18 and no WDR5 at all.

    What this evidence cannot show

    This is an internal entry cited to establish what this project already concluded, not as independent evidence - an internal citation cannot corroborate an internal claim. Its underlying sources (PMID 15199122 and IntAct's curation of it) were NOT re-verified in this beat, so it is taken on the earlier beat's reading. The exact file path was read from disk on 2026-09-23; the URL given is constructed from the repository layout and is not a resolvable public link, which is a limitation of how this project cites itself.

    Evidence that argues against this (1)
    • A mis-curation is a different failure mode from background co-purification, so the UTP18 precedent does not transfer as evidence about ABT1, KRR1 or UTP3 - it transfers only as a prior that nucleolar SSU-processome proteins appearing in KMT2A's IntAct interactome have already been wrong once for an unrelated reason. Treating it as more than that would be exactly the kind of pattern-matching this analysis is arguing against. Oncofusion Knowledge Graph (internal) 2026
Connected to