Oncofusion

start / graph / ana-the-index-fusion-has-an-ap-ms-interactome-and-this-graph-never-cited-it

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

KMT2A::AFF1 has had a published affinity-purification interactome since 2018 — in a paper titled for SETD2, uncurated by IntAct, and cited zero times here

also called Skucha 2018 MLL-fusion interactome · PXD009338 · the AP-MS survey we missed

A 2018 paper mapped every protein touched by seven versions of this fusion, including the one we study, and deposited the raw measurements publicly - and we had never cited it. It stayed invisible because its title is about a different protein, and because the interaction database we draw from never entered it.

The same thing, for a scientist

Skucha 2018 (PMID 29777171, Nat Commun) built isogenic Jurkat lines by FRT/Flp cassette exchange carrying Dox-inducible, SINGLE-COPY, Strep-HA-tagged variants of seven MLL fusions — MLL-AF1p, **MLL-AF4**, MLL-AF9, MLL-CBP, MLL-EEN, MLL-ENL, MLL-GAS7 — and affinity-purified each from NUCLEAR lysate for LC-MS/MS on an LTQ Orbitrap Velos, with a generalised linear model for bait-specific enrichment. Its abstract states verbatim: "We present the first comprehensive survey of protein-protein interactions of seven distantly related MLL-fusion proteins. Functional investigation of 128 conserved MLL-fusion-interactors identifies a specific role for the lysine methyltransferase SETD2 in MLL-leukemia." MLL-AF4 IS KMT2A::AFF1, the index fusion. **This graph cited the paper zero times before this entry** (grep over all node files for the PMID, the DOI and the first author). **IntAct has not curated it**: `pubid:29777171` returns 0 rows against a working positive control of 1,000 rows on `pubid:33961781`, and the PMID is absent from all 26 publications behind KMT2A's own 161 IntAct rows. Raw data are public at ProteomeXchange **PXD009338**, titled "Protein Interactions of 7 MLL-fusion Proteins". This graph's interactome is IntAct-derived, so it could not have contained this dataset and cannot until the per-bait table is imported by hand.

Go deeper
How a paper this relevant stayed invisible for eight years
What this does and does not do to the saturation priority
Facts and figures
pmid29777171
doi10.1038/s41467-018-04329-y
proteomexchangePXD009338
shown for fusionYES for MLL-AF4 = KMT2A::AFF1 specifically, as one of seven tagged baits — but in a T-ALL host cell (Jurkat) and as a single-copy inducible transgene, not the endogenous fusion in a B-lineage blast.
design
{
  "cells": "isogenic Jurkat, FRT/Flp locus-specific cassette exchange",
  "expression": "doxycycline-inducible, single-copy",
  "tag": "Strep-HA (SH), biotin elution",
  "input": "nuclear lysate",
  "instrument": "LTQ Orbitrap Velos",
  "specificity_model": "generalised linear model over protein spectral counts and summed peptide scores; DDX5 and DDX17 used as ubiquitous nuclear-background normalisers",
  "baits_verbatim": [
    "MLL-AF1p",
    "MLL-AF4",
    "MLL-AF9",
    "MLL-CBP",
    "MLL-EEN",
    "MLL-ENL",
    "MLL-GAS7"
  ]
}
network size reported
{
  "shared_by_more_than_one_fusion": "406 proteins (42.3%)",
  "shared_by_at_least_five_of_seven": "128 proteins (13.3%)",
  "interacting_with_all_seven": "42 core proteins, of which SETD2 is one",
  "total_network_computed_by_king": "CORRECTED 2026-09-21 (cycle 47), found by Q84 and verified by me against PMC5959866. This field previously read 'The paper does not state a total - this is my arithmetic from its own two percentages'. THAT WAS FALSE. The paper states 960 five times, including Results: \"p-value-based filtering for the 300 most significant interactions per MLL-fusion resulted in a network of 960 high-confidence cellular proteins\", and Methods: \"Detailed information about all 960 identified interactors ... is provided in Supplementary Data 1\". My arithmetic (959.8 and 962.4) was right and unnecessary. The figure is the paper's own.",
  "upstream_numbers_also_stated": "4,600 proteins recovered in total, engaging in 15,094 putative interactions, filtered to the 960."
}
why we missed it
{
  "title_names_a_different_protein": "The title is 'MLL-fusion-driven leukemia requires SETD2 to safeguard genomic integrity'. The interactome is the paper's method of discovery, not its claim. This is the 'measured but not featured' shape found earlier this cycle in the IDH supplement — but one level worse, because here the measurement is not in a supplement, it is Figure 1.",
  "intact_never_curated_it": "pubid:29777171 = 0 rows; positive control pubid:33961781 = 1,000 rows on the same endpoint in the same session. This graph's interactome inherits IntAct's curation frontier.",
  "our_own_absence_entry_counted_it_without_opening_it": "exp-no-proximity-labelling-has-ever-been-done-on-a-kmt2a-fusion records a control of 56 records for 'KMT2A/MLL interactome or interaction proteomics in leukaemia'. This paper is inside that 56. The control number was reported and the records behind it were never read."
}
How this could bite us back
  • This entry supports no clinical action. It records that a published dataset exists and that we had not read it; it makes no claim that any interactor in it is a target.
  • SETD2 appears here only as the paper's own headline dependency, validated by its authors in MV4-11 and SEM. Nothing here supports SETD2 inhibition as a therapeutic proposition, and the paper's finding is a LOSS-of-function dependency combined with a DOT1L inhibitor — a combination whose toxicity in infants is entirely unstudied.
  • The interactome was generated in a T-ALL line from a tagged transgene. Treating any partner on that list as an established KMT2A::AFF1 interactor in infant B-ALL, without orthogonal confirmation in a genuine fusion line, would repeat the error this project already made with atlas-derived partners from HEK293.
Still open
  1. Open Supplementary Table 1 and the per-bait interactor lists of PMID 29777171 and extract the MLL-AF4 column specifically. This is the highest-value desk job available to the project: the PI's first priority is every interactor of KMT2A::AFF1 with evidence, and this is a published, bait-specific, nuclear-lysate list nobody here has read.
  2. Re-search PXD009338 against a fusion-augmented database including the MLL-AF4 junction sequence. It closes absence-no-kmt2a-fusion-junction-peptide-reported-in-any-mass-spectrometry-dataset, whose own data block already says 'deposited raw data not swept' and names no accession. This is the accession.
  3. Cross the 128 conserved interactors and the 42 pan-fusion core against this graph's existing interactome: how many are already entries, and how many are genes this project has never written? That number is the honest current state of the saturation priority.
  4. Why is SETD2 named in 17 entries with no gene entry, when the paper that made it an MLL-leukaemia dependency is uncited? Check whether those 17 reached SETD2 solely through the menin–gammaH2A.X–SETD2–H3K36me3–LEDGF chain, which would mean the graph knows SETD2 only as a downstream node and not as a dependency.
  5. Does the Jurkat T-ALL host invalidate the list for a B-lineage disease, or is the nuclear fusion interactome largely lineage-independent? The paper's own MV4-11 and SEM validation of one hit is a single data point on that question.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A 2018 affinity-purification mass-spectrometry survey covered seven MLL fusion proteins including MLL-AF4, the index fusion, expressed as Strep-HA-tagged single-copy doxycycline-inducible transgenes in isogenic Jurkat cells and purified from nuclear lysate. 406 proteins (42.3% of the network) co-purified with more than one fusion, 128 (13.3%) with at least five of seven, and 42 interacted with all seven, SETD2 among them.

    MLL-fusion-driven leukemia requires SETD2 to safeguard genomic integrity. · Skucha A, Ebner J, Schmoellerl J, Roth M, Eder T, Cesar-Razquin A, Stukalov A, Vittori S, Muhar M, Lu B, Aichinger M, Jude J, Mueller AC · Nature Communications 2018
    PMID 29777171DOI
    What this evidence cannot show

    Full text read from PMC5959866 on 2026-09-21; the quoted design and percentages are verbatim from its Results and Methods. The AP-MS host is JURKAT, a T-ALL line, whereas KMT2A::AFF1 disease is B-lineage — interactors depending on a B-cell or infant proteome cannot appear, and T-lineage-specific partners may appear spuriously. The bait is a tagged, single-copy, doxycycline-induced transgene, not the endogenous fusion. I did NOT open Supplementary Table 1 or the per-bait interactor lists, so this entry does not state WHICH proteins were MLL-AF4-specific; that is the next step and it is a desk job.

    Evidence that argues against this (1)
    • A tagged fusion in a T-ALL line is a heterologous system, and this project has repeatedly found that a partner list from one cell type does not transfer. The paper's own validation of the SETD2 hit was done in MV4-11 and SEM — both genuine KMT2A::AFF1 lines — which shows the authors treated the Jurkat screen as a discovery layer needing orthogonal confirmation, not as the fusion's interactome in its own disease. Further, the 128-protein 'conserved' core is defined by sharing across seven fusions, so it is by construction the set LEAST specific to the index fusion; the interactors most relevant to KMT2A::AFF1 biology may sit in the unique fraction the paper does not emphasise. Nature Communications 2018
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    IntAct has not curated this publication. A PSICQUIC query pubid:29777171 against the IntAct endpoint returns 0 rows, while the same endpoint in the same session returns 1,000 rows for pubid:33961781, establishing that the query form works and that the zero is a real absence rather than a wrong field name. The PMID is also absent from all 26 distinct publications appearing in the 161 IntAct rows for Q03164 (KMT2A).

    What this evidence cannot show

    Bounded to IntAct via PSICQUIC. BioGRID, MINT, ComplexPortal and the wider IMEx set were NOT queried, so this is not a claim that no database holds it. The first three query forms tried — publication:, pubmed: and the bare PMID — returned 0 for the positive control as well and were discarded as uninformative before the working form was found. Recorded because a zero from an untested query form is exactly the error this project keeps making.

    Evidence that argues against this (1)
    • Absence from IntAct is a statement about curation throughput, not about data quality or availability, and IntAct's backlog is long. Reading non-curation as any judgement on the dataset would be wrong. The actionable consequence is narrower: any interactome coverage figure this graph computes from IntAct has this paper outside its denominator, so our coverage percentages are computed against a frontier that excludes the most directly relevant published experiment. PRIDE / ProteomeXchange (public; submitted 2018-04-23, released 2018-05-25; Homo sapiens; LTQ Orbitrap Velos) 2018
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The raw mass-spectrometry data are public. ProteomeXchange PXD009338, titled 'Protein Interactions of 7 MLL-fusion Proteins', was submitted 2018-04-23 and released 2018-05-25, Homo sapiens, LTQ Orbitrap Velos, with a description naming the same seven fusions. Because the fusion is the BAIT rather than an incidental prey, these files are the single most likely public location of a KMT2A fusion junction peptide.

    Protein Interactions of 7 MLL-fusion Proteins · PRIDE Archive / ProteomeXchange 2018
    What this evidence cannot show

    Metadata fetched from the PRIDE v3 API on 2026-09-21. I did NOT download the raw files and have not confirmed that MLL-AF4 pulldown runs are individually identifiable within the submission, nor what search database was used. Whether a tryptic junction peptide is DETECTABLE depends on the junction sequence, cleavage spacing and tag position, none of which I have checked.

    Evidence that argues against this (1)
    • A bait pulldown maximises the chance of junction coverage but does not guarantee it: the original search used a standard human database, in which a fusion-spanning peptide has no entry and would go unassigned rather than being reported as absent. The absence of a junction peptide in the published output is therefore uninformative — which is the point, and also the reason this needs a re-search rather than a re-read. Nature Communications 2018
Connected to