Oncofusion

start / graph / interface-kmt2a-abt1-declined-bioplex-prey-row

Interfaceweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q129-the-never-discussed-partners-of-kmt2a-and-psip1

KMT2A-ABT1 is one spoke-expanded BioPlex 3.0 AP-MS row in HEK293T in which ABT1 is the bait and KMT2A the prey, and ABT1's nucleolus is ECO:0000250 only - declined

also called KMT2A-ABT1 · MLL-ABT1 · ESF2 · EBI-54995944 · EBI-54513736

A database lists ABT1, a small nucleolus protein, as a KMT2A partner. The experiment tagged ABT1 and pulled it out of kidney cells; KMT2A merely turned up in the catch, and nobody has ever done the reverse experiment.

The same thing, for a scientist

**Verdict: DECLINED. Detection method is anti-tag co-immunoprecipitation (AP-MS) in BioPlex 3.0, spoke-expanded, with ABT1 as bait and KMT2A as prey** [E1]. Two IntAct records (EBI-54995944, EBI-54513736) both from PMID 33961781, both HEK293T, both intact-miscore 0.35 - the floor, against 0.96/0.95/0.91 for WDR5/RBBP5/ASH2L in the same dataset [E1]. ABT1 is one of the three nucleolar ribosome-biogenesis factors this beat was dispatched to test as a possible real nucleolar signal for KMT2A. **It is the weakest of the three on its own annotation: UniProt Q9ULW3 gives Nucleus and Nucleus/nucleolus at ECO:0000250 - 'by similarity' - with no experimental human localisation at all** [E3], so 'ABT1 is a nucleolar protein' is itself an inference in this record, not a measurement. It is also a heavy IntAct node: **761 interaction rows, more than WDR5's 639**, for a 272-residue protein whose only UniProt function statement is hedged ('Could be a novel TATA-binding protein... By similarity') [E3, E4]. Spoke expansion means the binary pair was never observed.

Go deeper
What BioPlex 3.0 is, and which of its own validation criteria this row passes

BioPlex 3.0 affinity-purified 10,128 tagged human proteins in HEK293T (118,162 interactions among 14,586 proteins) plus 5,522 immunoprecipitations in HCT116, scored with CompPASS-Plus against a GFP negative control at 1% protein-level FDR with entropy-based dataset-level filtering 2.

The paper names three ways an interaction is confirmed: replication in the second cell line, reciprocal confirmation, and co-elution / co-fractionation. About 60% of interactions in both networks are confirmed by one of these, so about 40% are not 2.

Reciprocal confirmation is structurally unavailable for every KMT2A row in BioPlex, because KMT2A is a prey-only node there. All 31 BioPlex rows touching Q03164 have KMT2A as prey 1; the 3,969-residue protein is not among the 10,128 baits. The strongest check the authors offer was never run on this pair and cannot be run without new work.

This row also fails the first criterion: both records are HEK293T. It did not replicate in HCT116.

IntAct's own score stratifies the 24 baits that recovered KMT2A, and every partner in this beat sits at the floor

Twenty-four distinct BioPlex baits recovered KMT2A as prey. Their intact-miscores, pulled live 2026-09-23 1:

miscorebaits
0.91-0.96WDR5 0.96, RBBP5 0.95, ASH2L 0.91 - the WRAD core of the MLL1 complex
0.53-0.79CSNK2A2 0.79, CSNK2B 0.64, HCFC2 0.62, CSNK2A1 0.53
0.35ABT1, ANOS1, ASB7, FGFBP1, KRR1, OASL, PIH1D1, PRKRA, RPL36, RPLP0, RPS9, Rpl13a, SET, SRSF4, SRSF5, SRSF6, and the RPL3/RPL3L ambiguous group

The screen recovered the genuine MLL1 complex, which is a reason to take the dataset seriously rather than to dismiss AP-MS on principle.

The floor score is not an independent opinion, and saying so matters. intact-miscore is a pair-level score driven largely by how many publications and methods support the pair, so 0.35 is close to a restatement of 'this pair has one record of one kind'. It is quoted because it makes the stratification legible, not because it is a second measurement.

The specific problem with ABT1: its nucleolar identity is the same kind of inference as the interaction

This beat was dispatched with an explicit hypothesis that ABT1, KRR1 (on KMT2A) and UTP3 (on PSIP1) might be a real nucleolar signal. The pre-registered criterion included a second condition: if the 'nucleolar ribosome-biogenesis factor' label can only be supported from the same database family that supplied the interaction, the grouping is an annotation artefact.

ABT1 fails that condition. Its UniProt human record gives Nucleus and Nucleus/nucleolus at ECO:0000250 (by similarity) and its function comment is doubly hedged - "Could be a novel TATA-binding protein (TBP) which can function as a basal transcription activator... (By similarity)" 3. The nucleolar role is inherited from the yeast orthologue Esf2, not measured in a human cell in the record.

KRR1 and UTP3, by contrast, carry ECO:0000269 nucleolar locations and explicit SSU-processome function text. So the nucleolar 'triple' is, at the level of annotation quality, a pair plus an inference.

What would change this. A human immunofluorescence or fractionation experiment placing ABT1 in the nucleolus, cited in the record, would restore the third member - and would not by itself make the KMT2A row true.

Facts and figures
partner agene-kmt2a
partner a uniprotQ03164
partner a length AA3969
partner bABT1
partner b uniprotQ9ULW3
partner b length AA272
verdictDECLINED
adjudicated byQ129, cycle 92, 2026-09-23
source publication pmid33961781
source datasetBioPlex 3.0
detection methodanti-tag co-immunoprecipitation (AP-MS)
experimental role partnerbait
experimental role kmt2aprey
kmt2a was never the baittrue
expansion methodspoke expansion
intact miscore0.35
replicated in second bioplex linefalse
reciprocally confirmedfalse
reciprocal confirmation possiblefalse
n publications1
n methods1
orthogonal method existsfalse
partner nucleolar evidence codeECO:0000250 (by similarity) only
intact row count partner761
druggabilityNot assessable. No residue, domain or region is mapped on either side in any record, and no structure of the pair exists.
absence classnever-measured
absence class noteNo experiment has ever tested KMT2A against ABT1 with KMT2A as the bait, in any cell, and none has tested the pair in a haematopoietic cell of any kind. Closing this needs new wet work - a KMT2A or KMT2A-fusion pull-down in a leukaemia line - not a re-analysis. A cheaper secondary gap is measured-not-searched: BioPlex Supplementary Table 1 and the BioPlex web portal hold the CompPASS-Plus score for this bait-prey pair and were not opened here.
shown for fusionNO - nobody has looked in any KMT2A fusion. The record is full-length KMT2A as a prey in a non-leukaemic embryonic kidney line; no fusion protein was present.
epigenetic readoutnone - no chromatin, histone-mark or occupancy measurement exists for this pair.
nucleolar hypothesis roleOne of the three nucleolar ribosome-biogenesis factors tested by ana-q129-the-nucleolar-triple-survives-none-of-the-four-pre-registered-discriminators; the weakest of the three because its nucleolar annotation is itself an inference.
host cell lines
[
  "HEK293T embryonic kidney"
]
records
[
  "EBI-54995944",
  "EBI-54513736"
]
interface residues a
[]
structures
[]
ligands
[
  "drug-revumenib",
  "drug-ziftomenib"
]
How this could bite us back
  • If KMT2A does have a nucleolar pool, a menin inhibitor tested only on chromatin readouts would not see the compartment where part of the protein sits, and a resistance mechanism living there would be invisible to every assay now used in the menin-inhibitor trials.
  • Declining this row is not evidence that KMT2A has no nucleolar biology. The correct statement is that nobody has looked with KMT2A as the bait, so the compartment is unmeasured rather than excluded.
  • ABT1 is a general basal-transcription and ribosome-biogenesis factor. If it were a real KMT2A partner it would be a poor intervention point: ribosome biogenesis is required by every proliferating cell, so the therapeutic window against normal haematopoiesis would be narrow.
Still open
  1. Does endogenous KMT2A co-purify with endogenous ABT1 in a KMT2A-rearranged leukaemia line? No experiment in any system has used KMT2A as the bait against ABT1.
  2. Is human ABT1 actually nucleolar? UniProt carries only ECO:0000250. A single published human immunofluorescence or nucleolar fractionation would settle it, and nothing in this graph currently depends on the answer.
  3. What CompPASS-Plus probability did BioPlex assign to the ABT1-KMT2A pair? It is published on the BioPlex portal and would convert this decline from an argument about metadata into an argument about the score the authors themselves computed.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    IntAct curates two records of a KMT2A-ABT1 association from PMID 33961781 (BioPlex 3.0): EBI-54995944 (annotated Table S2-B) and EBI-54513736 (annotated Supplementary table 1). Detection method anti-tag co-immunoprecipitation, interaction type 'association', expansion method SPOKE EXPANSION, host Homo sapiens HEK293T embryonic kidney cell in both. In both records ABT1 (Q9ULW3) is the BAIT and KMT2A (Q03164) is the PREY. intact-miscore 0.35. No mutation is recorded on either side. Across the whole of IntAct, all 31 BioPlex rows touching Q03164 have KMT2A as prey.

    What this evidence cannot show

    Read live from IntAct REST on 2026-09-23, not from PSICQUIC (rule 93: PSICQUIC returns about 64% of IntAct's rows for these baits). The source publication's own supplementary table was NOT opened, so the row rests on IntAct's curation of it. Spoke expansion is the decisive field: the binary KMT2A-ABT1 pair was never measured, it was inferred by connecting every prey in an ABT1 pull-down to the bait.

    Evidence that argues against this (1)
    • IntAct's curation is not the whole record. BioPlex publishes per-interaction CompPASS-Plus probabilities that IntAct does not carry, and a reader wanting to overturn this decline should pull the score for this bait-prey pair from bioplex.hms.harvard.edu rather than re-read IntAct. That number was not retrieved here. Huttlin EL et al. 2021
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In human HEK293T and HCT116 epithelial cell lines, BioPlex 3.0 affinity-purified 10,128 tagged human proteins giving 118,162 interactions among 14,586 proteins, plus 5,522 immunoprecipitations in HCT116, scored by CompPASS-Plus against a GFP control at 1% protein-level FDR with additional entropy-based dataset-level filtering. The paper states that about 60% of interactions in both networks are confirmed by replication, reciprocal recovery or co-fractionation, that proteins average just 8 partners, and that its headline finding is extensive cell-line specificity.

    Dual proteome-scale networks reveal cell-specific remodeling of the human interactome. · Huttlin EL, Bruckner RJ, Navarrete-Perea J, Cannon JR, Baltier K, Gebreab F, et al. (28 authors), Gygi SP · Cell 2021
    PMID 33961781DOI
    What this evidence cannot show

    Full text read 2026-09-23 through the PMC HTML route (PMC8165030, 221,127 stripped characters) after Europe PMC fullTextXML returned HTTP 500. The string 'KMT2A' does not appear anywhere in the paper's text and 'nucleolar' appears zero times. These rows exist only in the supplementary tables, which were not opened - so the paper does not REPORT this interaction in any readable sense; it deposited a table from which IntAct curated it.

    Evidence that argues against this (1)
    • The same objection would discard most of the structured human interactome. BioPlex recovered WDR5, RBBP5 and ASH2L on KMT2A at miscores of 0.96, 0.95 and 0.91 - the real MLL1 complex - so the dataset demonstrably sees true KMT2A biology. The argument here is specific to rows at the floor score with no reciprocal arm, no second cell line, and an annotation problem on the partner. IntAct (EMBL-EBI) 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB Q9ULW3 is 272 aa, primary gene name ABT1, recommended name 'Activator of basal transcription 1'. Its only subcellular-location comments are Nucleus [ECO:0000250] and Nucleus, nucleolus [ECO:0000250] - by similarity, with no experimental code. Its only function comment is 'Could be a novel TATA-binding protein (TBP) which can function as a basal transcription activator. Can act as a regulator of basal transcription for class II genes (By similarity)'. There is no tissue-specificity comment. KMT2A (Q03164) is 3,969 aa with one ECO:0000269 Nucleus comment plus two uncoded location comments for the taspase1 cleavage products N320 and C180.

    What this evidence cannot show

    UniProt subcellular-location comments are curated pointers, not measurements. Rule 91 established that ECO:0000269 means a curator read AN experiment, not necessarily THE experiment, and that uncoded comments carry the condition-scoped statements. Here the relevant observation is the opposite and stronger: there is no experimental code at all on ABT1's human localisation. The yeast Esf2 literature was not searched, so 'no human experiment exists' is not asserted - only that UniProt cites none.

    Evidence that argues against this (1)
    • An ECO:0000250 code is a statement about UniProt's curation, not about the literature. ABT1 has a real published history as an RNA-processing factor and a human nucleolar experiment may well exist uncited - rule 91 records four cases in this project where UniProt's cited experimental paper was not the founding or best observation. This entry therefore declines the INTERACTION and flags the annotation, rather than asserting ABT1 is not nucleolar. UniProt Consortium 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    ABT1 (Q9ULW3) carries 761 interaction records in IntAct, more than WDR5 (P61964, 639) and more than five times MEN1 (O00255, 137), despite being a 272-residue protein with a hedged function annotation. KRR1 (Q13601) carries 464 and UTP3 (Q9NQZ2) carries 295.

    What this evidence cannot show

    A raw IntAct row count is not a promiscuity metric. It counts records, not distinct partners, and it is dominated by how many large screens a protein happened to appear in - so it measures participation in high-throughput proteomics as much as biochemical stickiness. This number is stated as a prior worth checking, not as a verdict; a proper frequent-flier statistic would be a CRAPome-style frequency across matched negative controls, and the CRAPome was not queried in this beat.

    Evidence that argues against this (1)
    • WDR5 at 639 rows is a genuine hub with dozens of real partners, so a high row count plainly does not imply artefact. The comparison is offered only to show that ABT1's count is not small, which is the opposite of what its sparse functional annotation would lead a reader to expect. IntAct (EMBL-EBI) 2026
Connected to