All nine unwritten MEN1 partners trace to exactly two publications, neither of which mentions menin outside an antibody catalogue row
also called Q128 two-paper concentration · MEN1 35512704 33961781 · menin off-chromatin partners adjudication
Nine famous cancer genes appear in a database as menin's binding partners, and every one of them comes from just two big screens that were never about menin. Seven come from a screen that deliberately hunted for interactions made by MUTANT proteins, and that paper never discusses menin in its text at all.
The same thing, for a scientist
The dispatch brief named nine MEN1 IntAct partners with no entry in this graph: AKT1, BRAF, CD226, EGFR, FBXW7, PTEN, SMAD4, SPOP, ZC3HC1. A live IntAct REST pull on O00255 (2026-09-23, 137 records) returns 22 rows covering those nine, and **they trace to exactly two publications** [E1]. **Seven of the nine (AKT1, BRAF, EGFR, FBXW7, PTEN, SMAD4, SPOP) are PMID 35512704** - Mo et al., *Cell* 2022, a quantitative high-throughput differential BRET screen (qHT-dS) built to find interactions that a cancer **mutation creates** and the wild-type protein does not make [E2]. **Two (CD226, ZC3HC1) are PMID 33961781**, Huttlin et al. 2021, i.e. BioPlex 3.0 / HCT116 affinity-purification MS, spoke-expanded [E3]. **Every one of the seven Mo rows carries a named mutant allele on the PARTNER side**, and the full substitutions are resolved here for the first time in this corpus [E1]. In the 198,614-character PMC full text of Mo et al., the strings `menin` and `MEN1` occur **exactly twice, and both are the same key-resources row for a ThermoFisher anti-menin antibody** (PA5-79663, RRID:AB_2746778) [E2]. Menin appears in no sentence of that paper's narrative, results or discussion. A stated three-clause admission criterion (C2) **admitted zero of nine** [E5]. Separately, the brief's premise that these nine are *not mentioned anywhere in the corpus* is **false for 9 of 9** - the instrument's mention detector reads only ids, titles and aliases [E4].
The concentration, which is the finding
| source | partners | what the paper is |
|---|---|---|
| PMID 35512704 (Mo, Cell 2022) | AKT1, BRAF, EGFR, FBXW7, PTEN, SMAD4, SPOP - 7 of 9 | a BRET screen of 17,792 wild-type/mutant pairs designed to find mutation-created interactions |
| PMID 33961781 (Huttlin, Cell 2021) | CD226, ZC3HC1 - 2 of 9 | BioPlex 3.0: AP-MS of 10,128 baits in HEK293T plus 5,522 IPs in HCT116 |
| PMID 40205054 (Schaffer, Nature 2025) | CD226 (second row only) | a U2OS multimodal cell map: AP-MS + size-exclusion + immunofluorescence over ~5,100 proteins |
A paper count off IntAct is not an independence count, and here it is not even a large paper count. Nine partners, two papers. Not one of the three sources set out to study menin. The project already knew that two IntAct 'publications' for KMT2A were BioPlex 2.0 and BioPlex 3.0 - one platform, one collaboration. This is the same shape one level worse: the seven-partner block is one platform, one screen, one supplementary table.
The third row is a live correction to the instrument's own output. analyses/core-complex-interactome-gap/results.json records CD226 as n_publications: 1. IntAct REST returns two publications for MEN1-CD226 - 33961781 and 40205054. The difference is the PSICQUIC-versus-REST floor that rule 93 established, observed here on a real row rather than in the abstract.
The alleles, resolved to substitution level
Cycle 72's ana-menin-neoppi-trap-ten-intact-partners-from-one-mutation-screen resolved these rows to residue only for EGFR and PTEN and said so honestly: 'asserting L858R and R173C would be recall dressed as a lookup.' The IntAct REST endpoint carries the substitution in featureShortLabels, which the MITAB 2.7 pull used in cycle 72 did not expose in that form. So this closes that entry's first open question, by lookup and not by recall 1.
| partner | allele(s) on the IntAct feature | IntAct feature type | method(s) | source panel |
|---|---|---|---|---|
| AKT1 | p.Glu17Lys | mutation causing an interaction | BRET | Table S2(B) |
| BRAF | p.Val600Glu | mutation causing an interaction; and mutation increasing interaction strength on the pull-down | BRET, NanoPCA split-luciferase, GST pull-down, anti-tag co-IP | S2(B), S4, Data S1 fig i, Fig 4D |
| EGFR | p.Leu858Arg | mutation causing an interaction | BRET | Table S6 |
| PTEN | p.Arg173Cys | mutation causing an interaction | BRET | Table S6 |
| FBXW7 | p.Arg465Cys, p.Arg465His, p.Arg479Gln, p.Arg505Cys | mutation with no effect | BRET | Table S2(B) |
| SMAD4 | p.Asp351His, p.Arg361Cys, p.Arg361His, p.Gly386Asp | mutation with no effect | BRET | Table S2(B) |
| SPOP | p.Tyr87Cys, p.Phe102Cys, p.Phe133Leu, p.Phe133Val | mutation with no effect | BRET | Table S2(B) |
Read the source panel column, because it is the whole argument. Table S2(B) is the CARINA output of the primary screen - the raw candidate list. Table S6 is the same thing for the extended driver-gene set. Those are screen hits. Only BRAF leaves the supplementary tables.
BRAF V600E is the one that is real, and it is still not a menin partner for this project
Mo et al. Figure 4D was read directly from the PMC figure image (nihms-1802312-f0004.jpg, 1050x666) 2. Panel D is a semi-IP of FLAG-BRAF V600E against endogenous, untagged binding partners in H1299 cells, ctrl versus V600E, IP and WCL lanes. MEN1 is legible as one of the blotted rows, in the left column between MAPK6 and POLE. That is consistent with the anti-menin antibody sitting in the key-resources table, and it makes MEN1-BRAF V600E the only one of the nine with (a) more than one method, (b) an endogenous untagged partner on one side, and (c) a main-text figure panel rather than a supplementary table.
And it still fails admission, on the clause that matters here: it is a property of the V600E allele. The screen's entire design is wild-type versus mutant, IntAct annotates the feature mutation causing an interaction, and the panel's control lane is empty vector rather than BRAF wild type. Whatever this complex is, the graph cannot use it to argue that menin has a non-chromatin role, because the thing that changed was BRAF.
BRAF V600E is also not a lesion of KMT2A-rearranged leukaemia. Where it is a lesion - hairy cell leukaemia, Langerhans cell histiocytosis, melanoma, a minority of lung adenocarcinoma - nobody has asked the menin question, and cycle 72 already logged that as an open question. This entry does not close it; it sharpens it, because the endogenous-menin lane in Fig 4D is better evidence than cycle 72 recorded.
C2: the admission criterion, and what would have made it uninformative
Stated before any verdict was assigned, and after the IntAct method columns and Mo Fig 4 had been read. This is NOT a blind pre-registration and saying otherwise would be false. What is honest about it is that the three clauses were taken from the project's standing quality bar - 'a single high-throughput AP-MS or Y2H hit is not evidence of a physiological complex' - and not invented to fit the rows.
C2. A partner P is ADMITTED as a menin interactor for this graph only if all three hold: (a) at least one detection method other than the single screen platform that first reported it; (b) at least one side of the pair is the endogenous, untagged protein; (c) the interaction is reported for the wild-type partner, not only for a mutant allele.
What would make C2 uninformative: if all nine failed on clause (a) alone, C2 would have done no work beyond counting papers and clauses (b) and (c) would be decoration. Equally, if all nine passed, C2 would be too weak to be a filter.
Outcome: 0 of 9 admitted, and C2 did split the set, so it was informative. BRAF passes (a) and (b) and fails only (c). The other six Mo partners fail (a), (b) and (c) together. CD226 passes (a) - two independent proteome-scale maps, BioPlex 293T and Schaffer U2OS - and fails (b), and (c) does not apply. ZC3HC1 fails (a) and (b). So the failure mode is not uniform, and reporting '0 of 9' without the decomposition would hide that one row has an endogenous-partner blot behind it and one row has two independent platforms behind it.
Two identifier problems found in passing, in files this beat does not own
1. RL3_RL3L_HUMAN is not a parser artefact of this project's script. The brief asked what it is. It is IntAct interactor EBI-25396188, whose description field is the literal string RPL3 and RPL3L and whose only alias is rl3_rl3l_human, typed MI:0301 (gene name). It carries no UniProt accession at all - altIds contains only the IntAct AC. It reaches KMT2A through one BioPlex 3.0 anti-tag co-IP row (PMID 33961781), and it is an ambiguous protein group: mass-spectrometry peptides that cannot be assigned between RPL3 (P39023) and RPL3L (Q92901), which IntAct has modelled as a single interactor and labelled with a composite UniProt-entry-name-style string. So analyses/core-complex-interactome-gap/run.py read IntAct's gene-name field faithfully, and IntAct's gene-name field is not a gene name. Neither RPL3 nor RPL3L appears anywhere in this corpus (0 files, word-boundary search over all 2,412 node files), so no entry is being shadowed. No entry has been written for it, per the brief.
2. IntAct writes the Mo et al. BRAF co-IP host cell as H199 lung lymph node carcinoma cell. The paper's Figure 4D legend says H1299. NCI-H1299 is a non-small-cell lung carcinoma line derived from a lymph node metastasis, which is exactly what IntAct's descriptor describes, so this is a one-character typo in the IntAct host-organism string and not a different cell line. Reported, not edited - and recorded because a cell-line name is an identifier and this project has been burned by identifiers that were almost right.
| script | no committed script; live IntAct REST pull, PMC efetch, PMC figure raster read, and a word-boundary corpus scan, all 2026-09-23 |
|---|---|
| status | run |
| result summary | Nine unwritten MEN1 partners, two source publications. 7/9 from PMID 35512704, a mutation-directed BRET neoPPI screen in which menin appears only as an antibody catalogue row; 2/9 from PMID 33961781 (BioPlex 3.0 / HCT116). CD226 has a third-platform row (PMID 40205054) the project's PSICQUIC-based instrument does not see. All seven Mo rows carry a mutant partner allele, now resolved to substitution level, closing an open question in ana-menin-neoppi-trap-ten-intact-partners-from-one-mutation-screen. Criterion C2 admitted 0 of 9. The brief's premise that the nine are unmentioned in the corpus is false for 9 of 9. |
| absence class | measured-not-reported |
| absence class note | The wild-type-versus-mutant BRET values for MEN1 against each of the seven Mo partners were MEASURED - the qHT-dS platform screens both counterparts by design - and they sit in the published supplementary tables S2(A)/S2(B) and S6 of PMC9597701. Nobody in this project has opened them. Closing this is an afternoon with a supplement, not wet work, and it would convert seven 'unsupported' verdicts into seven measured numbers with a direction. |
inputs
[ "https://www.ebi.ac.uk/intact/ws/interaction/findInteractions/O00255?page=0&pageSize=1000", "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pmc&retmode=xml&id=9597701", "https://pmc.ncbi.nlm.nih.gov/articles/instance/9597701/bin/nihms-1802312-f0004.jpg", "data/graph/nodes/*/*.json" ]
mo allele table — the numbers behind the chart
[
{
"partner": "AKT1",
"acc": "P31749",
"allele": "p.Glu17Lys",
"intact_feature": "mutation causing an interaction",
"method": "bret",
"source_table": "Table S2(B)",
"intact_ac": "EBI-63961482",
"miscore": 0.27
},
{
"partner": "BRAF",
"acc": "P15056",
"allele": "p.Val600Glu",
"intact_feature": "mutation causing an interaction / mutation increasing interaction strength",
"method": "bret + split luciferase (NanoPCA) + GST pull down + anti tag coip",
"source_table": "Table S2(B), Table S4, Data S1 Figure i, Figure 4D",
"intact_ac": "EBI-63963639, EBI-52381059, EBI-63973327, EBI-52380398",
"miscore": 0.6
},
{
"partner": "EGFR",
"acc": "P00533",
"allele": "p.Leu858Arg",
"intact_feature": "mutation causing an interaction",
"method": "bret",
"source_table": "Table S6",
"intact_ac": "EBI-63960130",
"miscore": 0.27
},
{
"partner": "PTEN",
"acc": "P60484",
"allele": "p.Arg173Cys",
"intact_feature": "mutation causing an interaction",
"method": "bret",
"source_table": "Table S6",
"intact_ac": "EBI-63958506",
"miscore": 0.27
},
{
"partner": "FBXW7",
"acc": "Q969H0",
"allele": "p.Arg465Cys; p.Arg465His; p.Arg479Gln; p.Arg505Cys",
"intact_feature": "mutation with no effect",
"method": "bret",
"source_table": "Table S2(B)",
"intact_ac": "EBI-63685884, EBI-63690742, EBI-63695122, EBI-63699348",
"miscore": 0.27
},
{
"partner": "SMAD4",
"acc": "Q13485",
"allele": "p.Asp351His; p.Arg361Cys; p.Arg361His; p.Gly386Asp",
"intact_feature": "mutation with no effect",
"method": "bret",
"source_table": "Table S2(B)",
"intact_ac": "EBI-63703597, EBI-63707938, EBI-63712107, EBI-63716028",
"miscore": 0.27
},
{
"partner": "SPOP",
"acc": "O43791",
"allele": "p.Tyr87Cys; p.Phe102Cys; p.Phe133Leu; p.Phe133Val",
"intact_feature": "mutation with no effect",
"method": "bret",
"source_table": "Table S2(B)",
"intact_ac": "EBI-63732205, EBI-63735878, EBI-63743193, EBI-63944378",
"miscore": 0.27
}
]source concentration
{
"n_partners": 9,
"n_publications": 2,
"largest_block": {
"pmid": "35512704",
"n_partners": 7,
"pct": 77.8
},
"third_platform_row": {
"pmid": "40205054",
"partner": "CD226",
"seen_by_psicquic": false
}
}intact rest pull
{
"accession": "O00255",
"total_records": 137,
"rows_covering_the_nine": 22,
"retrieved": "2026-09-23"
}rl3 rl3l human
{
"what_it_is": "IntAct ambiguous protein group, not a gene symbol",
"intact_ac": "EBI-25396188",
"description_field": "RPL3 and RPL3L",
"alias_type": "MI:0301 (gene name)",
"uniprot_accession": null,
"candidate_members": [
"RPL3 (P39023)",
"RPL3L (Q92901)"
],
"reached_via": "KMT2A (Q03164), PMID 33961781, anti tag coip, 1 row",
"corpus_presence_of_RPL3_or_RPL3L": 0,
"entry_written": false
}intact curation typo
{
"field": "hostOrganism",
"intact_value": "H199 lung lymph node carcinoma cell",
"paper_value": "H1299",
"source": "Mo et al. Figure 4D legend",
"severity": "cosmetic; same line"
}criterion c2
{
"clauses": [
"(a) a second detection method beyond the originating platform",
"(b) at least one endogenous untagged side",
"(c) reported for the wild-type partner, not only a mutant allele"
],
"admitted": 0,
"n": 9,
"uninformative_if": "all nine fail on clause (a) alone, or all nine pass",
"was_informative": true,
"decomposition": {
"BRAF": "passes a,b; fails c",
"CD226": "passes a; fails b; c n/a",
"ZC3HC1": "fails a,b; c n/a",
"AKT1": "fails a,b,c",
"EGFR": "fails a,b,c",
"PTEN": "fails a,b,c",
"FBXW7": "fails a,b,c",
"SMAD4": "fails a,b,c",
"SPOP": "fails a,b,c"
},
"pre_registration_honesty": "fixed after the IntAct method columns and Mo Fig 4 were read; clauses taken from the standing quality bar, not fitted to the rows"
}adjudicated register update
{
"target": "analyses/core-complex-interactome-gap/run.py ADJUDICATED",
"proposed_rows": {
"AKT1": "cycle 91 Q128; DECLINED - single BRET row, PMID 35512704 Table S2(B), allele p.Glu17Lys annotated 'mutation causing an interaction'; interface-menin-akt1-declined-single-bret-neoppi",
"BRAF": "cycle 91 Q128; DECLINED as a menin partner but the BEST of the nine - four methods incl. endogenous-menin semi-IP in Mo Fig 4D; a property of BRAF V600E, absent from KMT2A-r; interface-menin-braf-v600e-the-only-one-with-an-endogenous-menin-lane",
"CD226": "cycle 91 Q128; DECLINED - two proteome-scale AP-MS maps (BioPlex 3.0 HEK293T, Schaffer U2OS), both in lines that do not express this T-lymphocyte surface receptor; interface-menin-cd226-declined-proteome-scale-ap-ms",
"EGFR": "cycle 91 Q128; DECLINED - single BRET row, Table S6, allele p.Leu858Arg 'mutation causing'; a single-pass plasma-membrane RTK against nuclear menin; interface-menin-egfr-declined-single-bret-neoppi",
"FBXW7": "cycle 91 Q128; DECLINED - four BRET rows, one assay, all 'mutation with no effect'; interface-menin-fbxw7-declined-single-bret-no-allele-effect",
"PTEN": "cycle 91 Q128; DECLINED - single BRET row, Table S6, allele p.Arg173Cys 'mutation causing'; interface-menin-pten-declined-single-bret-neoppi",
"SMAD4": "cycle 91 Q128; DECLINED - four BRET rows, one assay, all 'mutation with no effect'; interface-menin-smad4-declined-single-bret-no-allele-effect",
"SPOP": "cycle 91 Q128; DECLINED - four BRET rows, one assay, all 'mutation with no effect'; interface-menin-spop-declined-single-bret-no-allele-effect",
"ZC3HC1": "cycle 91 Q128; DECLINED - one BioPlex HCT116 spoke-expanded prey row; UniProt records ZC3HC1 weakly or not expressed in leukocytes; interface-menin-zc3hc1-declined-proteome-scale-ap-ms"
}
}- The commercial and clinical temptation here is a combination rationale: menin inhibitor plus a BRAF, EGFR, AKT or PI3K inhibitor, justified by a database row. Every one of those rows is a mutant allele that KMT2A-rearranged leukaemia does not carry, so the combination would be built on a lesion the patient population does not have.
- The mirror risk is the one this entry itself could create. Declining nine partners makes 'menin is only chromatin' feel established, and it is not established - it is unrefuted. If menin does have a cytoplasmic or signalling pool, a menin inhibitor would have an off-chromatin effect nobody is monitoring in the trials, and the reason nobody is monitoring it is that the only candidate partners on offer were bad ones.
- The 'mutation with no effect' rows are the more dangerous half for a curator, because they carry no warning label and read like ordinary wild-type interactions. Three of this beat's nine (FBXW7, SMAD4, SPOP) are of that kind, and a future beat pulling IntAct without reading the featureType column would write all three in.
- What is the wild-type MEN1 BRET value against each of the seven partners in Mo et al. Table S2(A)? The screen measured it. If wild-type menin gives a signal indistinguishable from the mutant for FBXW7, SMAD4 and SPOP, the honest reading of 'mutation with no effect' is a constitutive over-expression proximity signal; if wild-type gives nothing, those three become gain-of-interaction rows that IntAct has mis-typed.
- Is MEN1 a labelled node in the Mo Figure 4A BRAF V600E spoke diagram, and if so is its edge drawn as Go-PPI (red) or comPPI (cyan)? The raster read here was legible for panel 4D and not decisively for 4A. The answer decides whether the authors themselves counted menin as a gain.
- How many other baits in this graph carry IntAct partner lists contaminated by 35512704 rows? The screen covered 17,792 pairs and this beat has only looked at MEN1; KMT2A, AFF1, AFF4, DOT1L and PSIP1 have not been checked against that PMID.
- Should analyses/core-complex-interactome-gap/run.py flag the publication concentration of its own absent list? A nine-item gap list that is two papers is a different object from a nine-item gap list that is nine papers, and nothing in the current output distinguishes them.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
A live IntAct REST pull on O00255 (MEN1) on 2026-09-23 returns 137 interaction records. Filtering to the nine partners named in the dispatch brief returns 22 rows. Seven partners (AKT1, BRAF, EGFR, FBXW7, PTEN, SMAD4, SPOP) are sourced solely to PMID 35512704; CD226 is sourced to PMID 33961781 and PMID 40205054; ZC3HC1 is sourced to PMID 33961781. Every 35512704 row carries a mutant allele on the partner side in featureShortLabels, with featureType 'mutation causing' for AKT1 p.Glu17Lys, BRAF p.Val600Glu, EGFR p.Leu858Arg and PTEN p.Arg173Cys, and 'mutation with no effect' for four FBXW7, four SMAD4 and four SPOP substitutions. The BRET rows carry intact-miscore 0.27; the BRAF rows carry 0.6.
IntAct molecular interaction database, REST findInteractions on O00255, retrieved 2026-09-23 by Q128 · IntAct (EMBL-EBI) 2026What this evidence cannot show
This is one endpoint at one moment. IntAct gains depositions and the record count will move. The row-level fields quoted here (featureShortLabels, featureTypes, detectionMethod, hostOrganism, annotations) are the REST JSON's own field names and were read directly, not inferred. The 22-row figure counts interaction records, not distinct experiments: four are FBXW7 alleles of one BRET assay and four are SPOP alleles of one BRET assay, so the number of independent measurements is far smaller than 22.
Evidence that argues against this (1)
- A REST count is not the interactome. This endpoint is broader than PSICQUIC but it is still only IntAct: any interaction curated elsewhere, or published and never curated, is invisible to it. This project has already found an AP-MS interactome of the index fusion (PMID 29777171) that IntAct has never entered, so a menin partner list built from IntAct is a floor in both directions. IntAct (EMBL-EBI) 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
PMID 35512704 is a quantitative high-throughput differential screening platform coupling BRET biosensors with miniaturised co-expression in live HEK293T cells, screening 17,792 wild-type/mutant interaction pairs across 2,172,864 data points. In its PMC full text (PMC9597701, 198,614 characters retrieved 2026-09-23), the case-insensitive tokens 'menin' and 'MEN1' occur exactly twice, and both occurrences are the same key-resources-table row for a ThermoFisher rabbit polyclonal anti-Menin (MEN1) antibody, catalogue PA5-79663, RRID:AB_2746778. Figure 4D is a semi-immunoprecipitation of FLAG-BRAF V600E against endogenous binding partners in H1299 cells, and MEN1 is one of the blotted rows in that panel.
Systematic discovery of mutation-directed neo-protein-protein interactions in cancer. · Mo X, Niu Q, Ivanov AA, Tsang YH, Tang C, Shu C, Li Q, Qian K, Wahafu A, Doyle SP, Cicka D, Yang X, Fan D, Reyna MA, Cooper LAD, Moreno CS, Zhou W, Owonikoko TK, Lonial S, Khuri FR, Du Y, Ramalingam SS, Mills GB, Fu H · Cell 2022What this evidence cannot show
The full text was retrieved via efetch db=pmc and the token search run over the tag-stripped, entity-unescaped body, so a menin mention inside an image would not be found by it - which is precisely what Figure 4D turned out to be. The Fig 4D reading is from the 1050x666 PMC raster (nihms-1802312-f0004.jpg) and is a visual identification of a band label, not a reading of the blot's intensity: it establishes that menin was blotted, not how much co-precipitated. The supplementary tables S2, S4 and S6, which hold every other menin datum in this paper, were NOT parsed here; the allele assignments come from IntAct's curation of them.
Evidence that argues against this (1)
- A proximity or affinity signal measured between two tagged, co-overexpressed proteins in a heterologous cell line is a real measurement and is not a demonstration that the complex forms at endogenous stoichiometry in any tissue. The Mo et al. platform is explicitly a co-expression BRET screen in live HEK293T cells, and its own validation arm exists because the primary screen output is a candidate list, not a partner list. Mo X et al. 2022
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
PMID 33961781 reports BioPlex 3.0: affinity-purification mass spectrometry of 10,128 human proteins in 293T cells yielding 118,162 interactions among 14,586 proteins, plus a second network from 5,522 immunoprecipitations in HCT116 cells. The IntAct rows for MEN1-CD226 (host HEK293T) and MEN1-ZC3HC1 (host HCT116) are spoke-expanded anti-tag co-immunoprecipitation rows from those two networks, with the partner as bait and MEN1 as prey.
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 2021What this evidence cannot show
Abstract read in full; the underlying supplementary interaction tables were not opened. Spoke expansion means the binary MEN1-partner pair was INFERRED from a bait pull-down that recovered many preys; it is not a measured binary interaction. The two host lines match the two networks the abstract describes, which is an internal consistency check on the curation and not independent confirmation of the pair.
Evidence that argues against this (1)
- BioPlex's own design argument cuts the other way for a specific pair: the paper's headline is that thousands of interactions are cell-line-specific and that networks 'rewire' between 293T and HCT116. A partner seen once, in one line, as a spoke-expanded prey is exactly the class the paper itself says is least transferable. Huttlin EL et al. 2021
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
The instrument's 'not mentioned anywhere in the corpus' flag is computed by load_graph() in analyses/core-complex-interactome-gap/run.py, which tokenises only each node's id, title and aliases. It does not read summary_lay, summary_expert, sections[].body_md, evidence or data. A word-boundary search of the full serialised text of all 2,412 node files finds all nine symbols present: AKT1 in 3 files, BRAF in 18, CD226 in 8, EGFR in 13, FBXW7 in 4, PTEN in 16, SMAD4 in 4, SPOP in 3, ZC3HC1 in 5. All nine appear specifically in MEN1-interaction context in ana-menin-neoppi-trap-ten-intact-partners-from-one-mutation-screen, ana-menin-interactome-enumeration-and-bounded-rejections and ana-the-interactome-read-queue-is-empty-and-the-instrument-was-discarding-the-only-list-anyone-would-act-on.
analyses/core-complex-interactome-gap/run.py, function load_graph(), read together with a word-boundary full-text scan of data/graph/nodes/*/*.json run 2026-09-23 · oncofusion-knowledge-graph (this repository) 2026What this evidence cannot show
The scan is a regex over the raw serialised JSON, so it counts a symbol appearing anywhere in a file including in an unrelated context. That is why the claim distinguishes the raw file counts from the three entries where the symbol appears in explicit MEN1-interaction context; those three were opened and read. BRAF at 18 files and PTEN at 16 are mostly unrelated oncology context, and quoting those counts as 'menin coverage' would be the mirror error. The citation URL points at the IntAct endpoint because the repository has no public URL; the source is the named file in this repo.
Evidence that argues against this (1)
- A mention is not an entry, and the instrument is measuring something real. Its union_absent list is still the correct list of partners with no node of their own; only the words 'and are not mentioned anywhere in the corpus' are false. Over-correcting - treating a mention inside a summary table as coverage - would recreate the failure rule 89 documents, where a link near a thing was counted as the thing. IntAct (EMBL-EBI) 2026
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
Applying criterion C2 to the nine partners admits none of them. BRAF satisfies clause (a) multi-method and clause (b) endogenous untagged partner and fails only clause (c) wild-type; CD226 satisfies (a) via two independent proteome-scale platforms and fails (b); the remaining seven fail two or three clauses. The criterion therefore separated the set rather than rejecting it uniformly, which is the condition under which it carries information.
IntAct molecular interaction database, REST findInteractions on O00255, retrieved 2026-09-23 by Q128 · IntAct (EMBL-EBI) 2026What this evidence cannot show
C2 was fixed after the IntAct method columns and Mo Figure 4 had been read, so it is a stated criterion and not a blind pre-registration; this is recorded rather than glossed. Clause (b) is judged from IntAct's featureShortLabels and the source figure legends, not from reading every method section. Clause (c) does not apply to the BioPlex rows, so CD226 and ZC3HC1 are scored on two clauses and the seven Mo partners on three, which means the scores are not on one scale.
Evidence that argues against this (1)
- Declining a row for weak evidence is not the same as showing the interaction does not occur. No published experiment has tested any of these pairs at endogenous levels in a haematopoietic cell, so the honest state is 'unsupported', not 'refuted'. A single well-powered endogenous co-IP could overturn any of these adjudications. IntAct (EMBL-EBI) 2026
- E6strong 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 combining biophysical interaction measurement and immunofluorescence over more than 5,100 proteins, resolving 275 molecular assemblies. IntAct record EBI-61191835 curates a CD226-bait, MEN1-prey anti-tag co-immunoprecipitation from it, annotated to Extended Data 1A, spoke-expanded, in U2OS osteosarcoma cells. This publication is absent from the MEN1-CD226 row in analyses/core-complex-interactome-gap/results.json, which records n_publications 1.
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 2025What this evidence cannot show
Abstract read in full; the Extended Data table itself was not opened, so the claim that MEN1 is a CD226 prey there rests on IntAct's curation of it. Osteosarcoma is a third cell line that does not express CD226, so a second platform agreeing does not make the pair more physiological - it makes the shared-artefact hypothesis (tagged overexpression of a surface receptor in a non-expressing line) more likely, not less.
Evidence that argues against this (1)
- Two platforms agreeing is normally evidence of reproducibility, and the counter-reading offered here - that both are overexpression artefacts in non-expressing lines - is an interpretation, not a measurement, and is unfalsifiable from the database alone. It would be settled by one endogenous co-IP in a CD226-positive primary T or NK cell, which nobody has run. Schaffer LV et al. 2025
adjudicates nine IntAct-curated MEN1 partners; 0 of 9 admitted
extends the cycle-72 finding and closes its residue-level open question with substitution-level alleles
same partner list, re-pulled 2026-09-23 through IntAct REST rather than PSICQUIC
the instrument flags these nine as unmentioned in the corpus; all nine are mentioned, including in this entry
- related to →KMT2A / MLL1 — the gene that breaks
RL3_RL3L_HUMAN, the non-symbol in the KMT2A absent list, resolved to IntAct ambiguous protein group EBI-25396188
per-partner adjudication of BRAF
per-partner adjudication of AKT1
per-partner adjudication of EGFR
per-partner adjudication of PTEN
per-partner adjudication of FBXW7
per-partner adjudication of SMAD4
per-partner adjudication of SPOP
per-partner adjudication of CD226
per-partner adjudication of ZC3HC1
the instrument finding behind this beat's premise correction
pattern noticed while adjudicating the nine
the system-level bound under the per-partner adjudications