Oncofusion

start / graph / interface-menin-cd226-declined-proteome-scale-ap-ms

Interfaceweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q128-the-menin-interactome-that-is-not-chromatin

MEN1-CD226 is two proteome-scale AP-MS maps, both spoke-expanded, both in cell lines that do not express CD226 - declined

also called menin-CD226 · menin-DNAM-1 · EBI-54589265 · EBI-61191835

A database lists CD226, an immune-cell surface receptor, as a menin partner. Both supporting experiments overexpressed CD226 in cell lines that never normally make it - kidney, colon and bone cancer lines - and neither measured the menin pair directly.

The same thing, for a scientist

**Verdict: DECLINED, and it is the only one of the nine with two independent platforms behind it, which is why the decline is argued rather than asserted.** IntAct holds two MEN1-CD226 records. EBI-54589265 is BioPlex 3.0 (PMID 33961781): anti-tag co-IP, **spoke expansion**, HA-FLAG-CD226 as bait, MEN1 as prey, **HEK293T**, miscore 0.53 [E1]. EBI-61191835 is the Schaffer 2025 U2OS multimodal cell map (PMID 40205054): anti-tag co-IP, spoke expansion, C-terminally FLAG/HA-tagged CD226 bait, **U2OS**, miscore 0.53, Extended Data 1A [E4]. **Neither pair was measured as a binary interaction** - spoke expansion connects every prey in a bait pull-down to the bait. And the decisive point is expression: UniProt records CD226 as a **single-pass plasma-membrane** receptor expressed by **peripheral blood T-lymphocytes** [E3]. HEK293T is embryonic kidney, HCT116 is colon carcinoma and U2OS is osteosarcoma. **Two platforms agreeing on an over-expressed surface receptor in three non-expressing lines is agreement about a shared artefact hypothesis, not independent confirmation.** **Criterion C2: passes (a), fails (b); (c) does not apply.** **`confidence_overall` on this entry rates the INTERACTION, not the adjudication.** The adjudication is strong; the interaction is what is weak, and a reader scanning the field should see the weak thing.

Go deeper
Both records, and the instrument correction they contain
IntAct ACsourcecell lineroleexpansionmiscore
EBI-54589265PMID 33961781, BioPlex 3.0HEK293TCD226 bait, MEN1 preyspoke0.53
EBI-61191835PMID 40205054, Schaffer U2OS mapU2OSCD226 bait, MEN1 preyspoke0.53

analyses/core-complex-interactome-gap/results.json records CD226 with n_publications: 1 and publications: ['33961781']. The second row is real and live in IntAct REST today. This is a concrete instance of the PSICQUIC-versus-REST floor that rule 93 established in the abstract: the instrument's own output undercounts this partner's support, and it undercounts it in the direction that makes the partner look weaker than the database says. Reporting that is more important than the verdict, because it means every n_publications in that results file is a floor and nothing in the file says so at row level.

Why two platforms do not rescue it

This is the only partner in the beat that satisfies criterion C2's clause (a), and it is worth being explicit about why that is not enough.

1. Neither measured the pair. Both are spoke expansion. IntAct is honest about it in the row. A spoke-expanded pair is a bookkeeping device for representing an N-prey pull-down as binary edges; it is not a binary observation, and two spoke-expanded rows are two bookkeeping devices.

2. The bait is a surface receptor in cells that do not have it. CD226 (DNAM-1) is a lymphocyte/NK receptor. Introducing an epitope-tagged, single-pass membrane protein into HEK293T or U2OS puts a protein with an exposed cytoplasmic tail and a hydrophobic segment into the secretory pathway of a cell with no cognate partners for it. Abundant nuclear proteins are the classic background of such purifications.

3. Correlated, not independent. Both platforms use HA/FLAG tandem affinity from whole-cell lysate followed by MS. The failure modes - lysis-induced association, tag effects, abundant-protein carry-over - are shared. Two methods with the same failure mode are one method run twice. This is the same reasoning the project applied when it found that two IntAct 'publications' for KMT2A were BioPlex 2.0 and BioPlex 3.0.

What would overturn it: a co-IP of endogenous menin from a CD226-positive primary human T or NK cell, or the reciprocal MEN1-bait purification recovering CD226 in such a cell. Neither exists.

And yet this is the one that would matter most if it were true

CD226/DNAM-1 is an activating receptor on NK and T cells and a live immuno-oncology target. Menin inhibitors are given to patients with acute leukaemia whose disease control plausibly involves NK-cell surveillance. If menin sat downstream of CD226 in a lymphocyte, a menin inhibitor would have an immune-cell-intrinsic effect that no trial is measuring.

That is not a reason to believe the row. It is a reason to note that the experiment which would settle it - menin in a CD226-positive primary lymphocyte - has never been done, and that its absence is never-measured rather than measured-not-analysed: no archive holds it.

Facts and figures
partner agene-men1
partner bCD226
partner b uniprotQ15762
partner b length AA336
partner a uniprotO00255
partner a length AA610
verdictDECLINED
adjudicated byQ128, cycle 91, 2026-09-23
druggabilityNot assessable. No region, residue or domain is mapped on either protein in any record, so it is not known whether the menin-inhibitor MBM pocket is involved, and no structure exists.
n methods1
n platforms2
expansion methodspoke expansion (both records)
intact miscore0.53
absence classnever-measured
absence class noteNo experiment has tested menin against CD226 in any cell that natively expresses CD226. Both existing records are tagged over-expression in non-expressing lines. Closing this needs a primary human T or NK cell co-IP - wet work, not a supplement, and no deposited dataset was found that would substitute for it.
interface residues a
[]
structures
[]
ligands
[
  "drug-revumenib",
  "drug-ziftomenib"
]
records
[
  "EBI-54589265 (anti tag coip, HEK293T, BioPlex 3.0)",
  "EBI-61191835 (anti tag coip, U2OS, Schaffer 2025)"
]
intact feature types
[
  "tandem tag",
  "flag tag",
  "ha tag"
]
expression mismatch
{
  "partner_tissue": "peripheral blood T-lymphocytes (UniProt Q15762)",
  "partner_compartment": "cell membrane, single-pass",
  "host_lines_used": [
    "HEK293T",
    "U2OS"
  ],
  "none_expresses_the_bait_natively": true
}
instrument correction
{
  "file": "analyses/core-complex-interactome-gap/results.json",
  "field": "all_missing[gene=CD226].n_publications",
  "recorded": 1,
  "live_intact_rest": 2,
  "missing_pmid": "40205054",
  "cause": "PSICQUIC does not return the record; rule 93 floor, observed at row level"
}
c2
{
  "a_second_method": true,
  "b_endogenous_side": false,
  "c_wild_type": null,
  "admitted": false,
  "failing_clause": "b"
}
How this could bite us back
  • If menin has any function in lymphocytes, a menin inhibitor is an immunomodulator that nobody is monitoring as one. Differentiation syndrome and QTc are the monitored toxicities; immune-cell-intrinsic effects would be invisible in that monitoring.
  • Declining this row must not be read as evidence that menin inhibitors are immunologically inert. The correct statement is that the question has never been asked in a cell that expresses the receptor.
Still open
  1. Does endogenous menin co-precipitate with endogenous CD226 in a primary human T or NK cell? Nothing like it exists, and it is the only experiment that would settle this row.
  2. How many other rows in analyses/core-complex-interactome-gap/results.json undercount n_publications? CD226 is one confirmed case; the fix is to re-derive the whole file from IntAct REST rather than PSICQUIC, which rule 93 already recommended and which nobody has done.
  3. Does a menin inhibitor alter NK or T-cell function at clinically achieved concentrations? Independently of this row, that is an unmeasured safety and efficacy question for a drug given in acute leukaemia.
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 record EBI-54589265 curates a MEN1-CD226 interaction from PMID 33961781 (BioPlex 3.0). Detection method anti-tag co-immunoprecipitation, interaction type 'association', expansion method SPOKE EXPANSION, host Homo sapiens HEK293T embryonic kidney cell, CD226 as HA-FLAG tandem-tagged BAIT and MEN1 as PREY, annotated to Supplementary Table 1. intact-miscore 0.53. No mutation is recorded on either side.

    What this evidence cannot show

    Spoke expansion is the decisive caveat and it is in the row itself: the binary MEN1-CD226 pair was never measured. It was INFERRED by connecting every prey in a CD226 bait pull-down to the bait. IntAct records this honestly and a consumer that ignores the expansionMethod column converts an inference into a measurement. Read live from IntAct REST 2026-09-23; the BioPlex supplementary table was not opened.

    Evidence that argues against this (1)
    • BioPlex's own headline is that interactions are extensively cell-line-specific and that networks rewire between 293T and HCT116. A partner recovered once, in one line, as a spoke-expanded prey is precisely the class the paper says transfers least well - so the source argues against its own row being general. Huttlin EL et al. 2021
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PMID 33961781 reports BioPlex 3.0: AP-MS of 10,128 human proteins in 293T cells giving 118,162 interactions among 14,586 proteins, plus a second network from 5,522 immunoprecipitations in HCT116 cells. Its stated headline is that interactions are extensively cell-line-specific and that subnetworks rewire between the two lines.

    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, Gygi MP, Thornock A, et al. (28 authors), Gygi SP · Cell 2021
    PMID 33961781DOI
    What this evidence cannot show

    Abstract read in full; the supplementary interaction tables were not opened, so the MEN1 row is taken from IntAct's curation of them rather than from the source table. BioPlex is a platform and a collaboration, not an independent observer: BioPlex 1.0, 2.0 and 3.0 share methods, reagents and personnel, and this project has already been caught counting two BioPlex releases as two publications.

    Evidence that argues against this (1)
    • BioPlex recovers an enormous number of real complexes and dismissing an AP-MS row on principle would discard most of the structured interactome. The objection here is specific - one line, one pull-down, spoke expansion, a bait that the line does not natively express - and it would not apply to a BioPlex row recovered reciprocally or in both cell lines. Huttlin EL et al. 2021
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB Q15762 (CD226) is 336 aa. Subcellular location: Cell membrane, single-pass (ECO:0000269). Tissue specificity: Expressed by peripheral blood T-lymphocytes MEN1 (O00255, 610 aa) carries a single subcellular location, Nucleus, with experimental evidence (ECO:0000269), and tissue specificity 'Ubiquitous'.

    What this evidence cannot show

    UniProt subcellular-location comments are curated pointers, and rule 91 in this project's brief established that an ECO:0000269 code means a curator read AN experiment, not necessarily THE experiment or the best one. The cited papers behind these location lines were not opened here. Uncoded lines are recorded as uncoded and not discarded.

    Evidence that argues against this (1)
    • A compartment mismatch is a prior, not a refutation. Proteins move: menin's own UniProt record annotates only Nucleus, yet the founding localisation paper (PMID 9465067) maps at least two NLSs and UniProt does not cite it, which this project recorded as rule 91. An argument from UniProt compartments is therefore an argument from an incomplete record on both sides of the pair. UniProt Consortium 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PMID 40205054 (Schaffer et al., Nature 2025) is a U2OS multimodal cell map integrating biophysical interaction measurement with immunofluorescence over more than 5,100 proteins and resolving 275 molecular assemblies. IntAct record EBI-61191835 curates a CD226-bait, MEN1-prey anti-tag co-immunoprecipitation from it, spoke-expanded, in U2OS osteosarcoma cells, annotated to Extended Data 1A, intact-miscore 0.53.

    Multimodal cell maps as a foundation for structural and functional genomics. · Schaffer LV, Hu M, Qian G, Moon KM, Pal A, Soni N, Latham AP, Pontano Vaites L, et al. (34 authors), Ideker T · Nature 2025
    PMID 40205054DOI
    What this evidence cannot show

    Abstract read in full; Extended Data 1A itself was not opened, so the MEN1 row rests on IntAct's curation. U2OS does not express CD226, so this is a third non-expressing line rather than an independent biological context.

    Evidence that argues against this (1)
    • This is a genuinely different platform from BioPlex - different laboratory, different cell line, size-exclusion and imaging arms - so treating it as a repeat of BioPlex is an interpretation. If the two datasets were fully independent, a repeated recovery would be meaningful evidence, and the argument here rests on their shared reliance on tagged over-expression AP-MS. Schaffer LV et al. 2025
Connected to