UniProt's menin 214-390 FANCD2 region is not a surface: it spans three spatial lobes and 29.9% of menin's solvent-accessible area, so its twenty-residue overlap with the gamma-H2A.X motif is what chance gives you
also called menin FANCD2 214-390 versus gamma-H2A.X 371-400 · does menin have one DNA-repair surface
A database says menin's middle third binds the repair protein FANCD2, and a 2026 paper says a thirty-residue stretch inside that middle third binds a different repair protein, so it looked as if menin might have one DNA-repair docking site with several tenants. Measuring the actual structure says no: the FANCD2 region is almost a third of menin's whole outer surface spread across three separate lumps of the protein, and a thirty-residue motif dropped anywhere at random would land inside it this deeply about three times in ten.
The same thing, for a scientist
The question was whether the UniProt region `214-390 Interaction with FANCD2` and the gamma-H2A.X motif `Val371-Gln400` describe **one** menin surface with two named occupants, which would make menin's DNA-repair interface a single druggable object, or two coarse annotations that happen to abut. **They are not the same surface, and the FANCD2 annotation is not a surface at all.** Three measurements settle it, all made here on verified coordinates. (1) **Provenance is singular, not convergent.** UniProt's region carries `ECO:0000269` pointing at PMID 12874027, Jin et al. Cancer Res 2003 - the same and only source IntAct curates for the pair, as `sufficient binding region 214-390` [E1]. The 2003 abstract states no region at all; the number exists only in that paper's deletion-construct figures, which are unreadable through all five routes [E1]. (2) **Geometry.** 173 of the region's 177 residues are observed in PDB 3U84 and they distribute 75 / 19 / 79 across menin's three spatial lobes; its radius of gyration is 17.4 A against the whole chain's 26.3 A; and it contributes **7,307 of menin's 24,438 A^2 of solvent-accessible surface, 29.9%**, at 52.0% exposed against a chain-wide 58.0% - i.e. it is slightly *less* surface-like than menin's average [E2]. The gamma-H2A.X motif, by contrast, lies **entirely** in one lobe. A 177-residue slab through the middle of a single compact domain is a construct boundary, not an epitope. (3) **The null.** Drop a 30-residue motif uniformly at random into a 610-residue chain: it overlaps a 177-residue region at all with probability 0.355, the expected overlap is 9.1 residues, and **the probability of an overlap of at least the observed 20 residues is 0.289** [E2]. (4) The two literatures are also not independent: Lian et al. 2026 **cite Jin 2003** as their reference 18, in one introductory sentence listing RPA, FANCD2 and Chk1 as prior menin-repair associations, and test none of them [E3]. **Verdict: coincidence of two coarse annotations.** What survives is narrower and still worth having - the only part of the FANCD2 region that could be the gamma-H2A.X surface is the ordered helix 371-385, 16 observed residues out of 173.
No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.
What the primary source behind the UniProt annotation actually is, and what it mapped
Fetched live from UniProt on 2026-09-22, O00255 carries exactly two Region features in its whole feature table:
| region | description | evidence |
|---|---|---|
| 214-390 | Interaction with FANCD2 | ECO:0000269 PMID 12874027 |
| 460-552 | Disordered | ECO:0000256 MobiDB-lite |
So the primary source is Jin S, Mao H, Schnepp RW, Sykes SM, Silva AC, D'Andrea AD, Hua X. Menin associates with FANCD2, a protein involved in repair of DNA damage. Cancer Res 2003; 63:4204-10. PMID 12874027 1.
Its abstract, read in full here, names no region. It reports four things: menin specifically interacts with FANCD2; the interaction is enhanced by gamma-irradiation; menin loss in mouse embryonic fibroblasts increases DNA-damage sensitivity; and menin is on chromatin and in the nuclear matrix, with the matrix fraction rising after irradiation. It also states that menin is "a protein of 610 amino acid residues", which fixes the numbering frame: 214-390 is in 610-residue numbering, and the 219-395 that appears in a 2020 review is the same region in 615-residue isoform-1 numbering.
So the number comes from the paper's figures, and the paper is closed: no PMCID, no DOI in PubMed, Crossref or OpenAlex, aacrjournals.org HTTP 403. interface-menin-fancd2 recorded that five-route failure on 2026-09-20 and this beat did not repeat it. Which deletion constructs produced 214-390 is therefore unknown, and that matters, because a boundary at 390 is a boundary inside the only internal stretch of menin that no crystal has ever resolved.
The boundary at 390 sits inside menin's only internal invisible stretch
Measured over all 69 menin deposits with per-residue numbering verified: residues 389-397 are observed in zero of them, and 387-398 are absent from the construct altogether in 41 2. The next never-observed run does not start until 462.
So the FANCD2 region's C-terminal boundary, 390, is the seventh residue of menin's single internal disordered segment, and the gamma-H2A.X motif's own never-observed segment (389-397) begins one residue after it.
Two readings, and the honest position is that this analysis cannot choose between them.
- Deflationary and most likely: a flexible loop is where a molecular biologist puts a construct boundary, because it is the place a deletion is least likely to unfold the protein. Two groups twenty-three years apart, cutting at the only convenient place, is not a shared epitope.
- Substantive and not excluded: a flexible loop is also where a linear motif for a partner protein would sit, and both papers are about DNA-damage partners. The 2026 work does place a functional, deletion-sensitive, peptide-sufficient element at 371-400, which straddles exactly this loop.
Neither reading is testable from the 2003 paper, because its constructs cannot be read. Recorded as an inference, labelled as one, and not as a finding.
Geometry: the two windows are not the same kind of object
From analyses/q36-menin-residue-axis/run.py, all on PDB 3U84 with every residue number verified against O00255 2:
| FANCD2 region 214-390 | gamma-H2A.X motif 371-400 | SvID 360-445 | |
|---|---|---|---|
| width | 177 | 30 | 86 |
| observed residues | 173 | 16 | 71 |
| spatial lobes it occupies | 3 (75 / 19 / 79) | 1 | 1 |
| radius of gyration | 17.4 A | 7.6 A | 13.4 A |
| total solvent-accessible area | 7,307 A^2 | 952 A^2 | 2,810 A^2 |
| share of menin's 24,438 A^2 | 29.9% | 3.9% | 11.5% |
| residues exposed >25 A^2 | 52.0% | 62.5% | 46.5% |
Chain-wide, 58.0% of menin's residues are exposed above 25 A^2. The FANCD2 region is below that. It is not a patch on the outside of the protein; it is a slab through it, carrying roughly a third of the surface and a comparable share of the core.
The SvID and the gamma-H2A.X motif are both wholly inside the fingers lobe. The FANCD2 region is 43% in the NTD/palm lobe, 11% in the thumb and 46% in the fingers. Two objects can only be "the same surface" if both of them are a surface.
The null, and why the overlap is not evidence
Rule 84. Before reporting that a 30-residue motif overlaps a 177-residue region by 20 residues, write down what a random 30-residue motif would do.
Uniform placement of the motif's start over all 581 positions that keep it inside the 610-residue chain:
| P(any overlap with 214-390) | 0.355 |
| expected overlap | 9.1 residues |
| observed overlap | 20 residues |
| P(overlap >= 20) | 0.289 |
Roughly three random placements in ten would produce an overlap at least this deep. That is not a signal, and it is the whole of the sequence-level case for the two being one surface.
The null is deliberately crude - it ignores structure, and a structure-aware null would have to condition on the fact that both windows are in the modelled, ordered part of the protein, which would push the probability up, not down. Stated as a floor.
What would have made the answer yes, and what it would have meant
This beat was dispatched on the reading that if the two are the same surface, menin has one DNA-repair-partner surface with three claimed occupants, which is a much stronger statement than three separate findings. That reading is right, and the evidence does not support it. For the record, what would have supported it:
- A narrower FANCD2 region. Had Jin et al. mapped a 20-40 residue window rather than 177, an overlap would have carried information. A region covering 29% of the protein's surface cannot.
- Co-localisation in one lobe. The SvID and the gamma-H2A.X motif genuinely are one lobe, which is why
exp-reciprocal-deletion-test-of-the-svid-and-the-gammah2ax-motifis a sensible experiment. The FANCD2 region is not. - Competition data. None exists. Nobody has asked whether FANCD2 and H2A.X compete for menin, and the reagent to ask with - the Pep-MH 30-mer from the 2026 paper - now exists and is a published, synthesisable sequence.
And one thing does survive, which is a cheaper experiment than the one the brief imagined. The only part of the FANCD2 region that could be the gamma-H2A.X surface is the ordered helix 371-385. A menin delta-371-400 construct already exists, published, and is shown to abolish gamma-H2A.X binding. Nobody has run a FANCD2 co-immunoprecipitation against it. If delta-371-400 keeps FANCD2, the two contacts are separable and this entry's verdict is confirmed by experiment rather than by geometry; if it loses FANCD2, the 177-residue region collapses onto a 30-residue one and the brief's stronger statement is back.
A correction to the framing, not to the entry: the graph had already found the source
The dispatch brief for this beat described the FANCD2 overlap as "one overlap nobody has connected". Partly true and partly not, and the difference matters for how the coordinator reads its own inventory.
interface-menin-fancd2was written in cycle 72 (2026-09-20). It already names PMID 12874027 as the sole source, already records the 214-390 sufficient binding region and its 177-of-610 width, already records the five-route unreadability, and already carries the 2008 negative follow-up as counter-evidence. The primary source behind the UniProt annotation was found two cycles ago.interface-menin-h2ax-val371-gln400already states the overlap question in itsopen_questions: "Does UniProt's curated menin 214-390 FANCD2 region describe the same contact seen again? Both are coarse truncation maps in a DNA-damage context on overlapping menin residues, from papers 23 years apart that do not cite each other."- The last clause of that sentence is wrong and is corrected here. Lian et al. 2026 cite Jin et al. 2003 as their reference 18 3. They do cite each other - in one direction, in one sentence, without comparing regions.
So the beat's contribution is the adjudication, not the discovery of the question. The inventory generator did not surface either entry: it truncated its interface- listing at 22 of 44 with the note "and 28 more, run the script to see them all", so interface-menin-fancd2 never appeared, and its ALREADY ANSWERED section carried one item.
| script | analyses/q36-menin-residue-axis/run.py |
|---|---|
| status | run |
| result summary | Coincidence of two coarse annotations. The FANCD2 region is 29.9% of menin's solvent-accessible surface across three spatial lobes; the gamma-H2A.X motif is 3.9% in one lobe; P(overlap >= 20 residues) under random placement is 0.289. The two literatures are also not independent - the 2026 paper cites the 2003 one. |
| figure | |
| verdict | NOT the same surface. The UniProt FANCD2 region is not a surface. |
| shown for fusion | NO - nobody has looked in any fusion. Neither the 2003 FANCD2 co-immunoprecipitation nor the 2026 gamma-H2A.X mapping was done in a KMT2A-rearranged cell; the systems are mouse embryonic fibroblasts and A549 lung adenocarcinoma respectively. |
| absence class | measured-not-reported |
| absence class note | The deletion-construct figures behind 214-390 were measured in 2003 and printed; they are simply behind a paywall this project cannot pass. One library request closes it, which is why this is `measured-not-reported` and not `never-measured`. |
inputs
[ "UniProt O00255 feature table, fetched live 2026-09-22", "PDB 3U84 coordinates, numbering verified residue-for-residue", "PMID 12874027 abstract; PMID 42578372 full text from PMC13458390" ]
primary source behind the uniprot annotation
{
"pmid": "12874027",
"evidence_code": "ECO:0000269",
"what_it_mapped": "UNKNOWN - the abstract names no region; the 214-390 boundary is in figures that are unreadable through all five routes plus the publisher site",
"numbering_frame": "610-residue isoform 2, fixed by the abstract's own statement that menin is 610 residues"
}measurements — the numbers behind the chart
{
"fancd2_region": {
"start": 214,
"end": 390,
"width": 177,
"observed_residues": 173,
"lobes": {
"lobe1_ntd_palm": 75,
"lobe2_thumb": 19,
"lobe3_fingers": 79
},
"sasa_A2": 7307.3,
"pct_of_chain_sasa": 29.9,
"pct_residues_exposed_gt_25A2": 52,
"radius_of_gyration_A": 17.36
},
"gamma_h2ax_motif": {
"start": 371,
"end": 400,
"width": 30,
"observed_residues": 16,
"lobes": {
"lobe3_fingers": 16
},
"sasa_A2": 952.4,
"pct_of_chain_sasa": 3.9,
"pct_residues_exposed_gt_25A2": 62.5,
"radius_of_gyration_A": 7.59
},
"chain": {
"total_sasa_A2": 24437.5,
"pct_residues_exposed_gt_25A2": 58,
"radius_of_gyration_A": 26.31
},
"overlap_null": {
"motif_width": 30,
"region_width": 177,
"chain_length": 610,
"observed_overlap": 20,
"P_any_overlap": 0.3546,
"expected_overlap_residues": 9.14,
"P_overlap_at_least_20": 0.2892,
"n_placements": 581
}
}domains — the numbers behind the chart
[
{
"name": "FANCD2 interaction region (UniProt, curated from PMID 12874027)",
"start": 214,
"end": 390
},
{
"name": "gammaH2A.X motif (menin 371-400)",
"start": 371,
"end": 400
},
{
"name": "overlap of the two windows",
"start": 371,
"end": 390
},
{
"name": "the only ordered, modelled part of that overlap",
"start": 371,
"end": 385
}
]- This entry removes a rationale rather than adding one. If menin has no single DNA-repair surface, then the proposal to treat menin inhibition as a synthetic-lethal partner for DNA-damaging agents loses its tidiest structural story and has to rest on the HR phenotype alone.
- The deflationary direction has its own failure mode: concluding 'the FANCD2 annotation is uninformative' is one step from concluding 'menin does not bind FANCD2', which this analysis does not show and cannot. A coarse annotation is a bad map, not a refutation.
- If a clinical menin inhibitor does perturb HR through the fingers lobe, combining it with a PARP inhibitor or a crosslinker in a child is a combination whose toxicity ceiling nobody has measured, and the absence of a clean surface makes it harder, not easier, to predict which compounds do it.
- Does menin delta-371-400 - a published, existing construct - still co-immunoprecipitate FANCD2? This is the direct experiment, it needs no new reagent, and it converts this entry's geometric verdict into a measured one. FALSIFIER: if delta-371-400 loses FANCD2 binding in a side-by-side co-IP with wild-type menin in the same cell line at matched expression, with a >=70% reduction in recovered FANCD2 by densitometry across three biological replicates, this entry's verdict is wrong and the 177-residue region collapses onto the 30-residue motif.
- Do FANCD2 and H2A.X compete for menin? The Pep-MH 30-mer (menin 371-400, sequence published) blocks the menin-gamma-H2A.X interaction; adding it to a menin-FANCD2 co-IP is a one-plate experiment. FALSIFIER: if 10 uM Pep-MH reduces co-immunoprecipitated FANCD2 by >=50% relative to the alanine-substituted Pep-Mut control in the same lysate, the two contacts share a surface after all.
- Can anyone obtain the Jin 2003 figures? Everything uncertain here turns on which deletion constructs produced 214-390, and that is a library request rather than an experiment.
- Is FANCD2 recoverable in a modern menin affinity-purification at all? IntAct holds one source publication for the pair and two decades of menin AP-MS have not recovered it, so the honest prior on the interaction itself is lower than the UniProt feature table implies.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
UniProt O00255 carries exactly two Region features, one of which is 214-390 'Interaction with FANCD2' with evidence code ECO:0000269 citing PMID 12874027; that paper's abstract reports a specific menin-FANCD2 interaction enhanced by gamma-irradiation, increased DNA-damage sensitivity in menin-null mouse embryonic fibroblasts, and menin on chromatin and nuclear matrix, and states menin's length as 610 residues - but names no region, so the 214-390 boundary exists only in figures that cannot be read.
Menin associates with FANCD2, a protein involved in repair of DNA damage. · Jin S, Mao H, Schnepp RW, Sykes SM, Silva AC, D'Andrea AD, Hua X · Cancer Research 2003What this evidence cannot show
Abstract only. The paper has no DOI in PubMed's ArticleIdList, no PMCID and no open version; `interface-menin-fancd2` records all five retrieval routes failing plus HTTP 403 from the publisher on 2026-09-20 and that was not re-attempted here. So the claim 'the abstract names no region' is bounded to the abstract, and the claim about which constructs produced 214-390 is an absence of access, not an absence of data - the figures exist and somebody with a library subscription can read them in an afternoon. The system is mouse embryonic fibroblasts with largely FLAG-tagged constructs; nothing in it is human or haematopoietic.
Evidence that argues against this (1)
- The designed genetic test of the functional claim failed: MEN1 and FANCD2 mutants have different mutation spectra and no genetic interaction for crosslink sensitivity, so MEN1's role in crosslink repair is distinct from the Fanconi genes'. A physical association with no genetic interaction is also consistent with the association being real and non-functional. Marek LR et al. 2008
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
On PDB 3U84 with every residue number verified against O00255, menin residues 214-390 contribute 7,307 of the chain's 24,438 A^2 of solvent-accessible surface (29.9%) at 52.0% of residues exposed above 25 A^2 against a chain-wide 58.0%, distribute 75/19/79 across three spatial lobes with a radius of gyration of 17.4 A against the chain's 26.3 A, and contain the 389-397 stretch that is observed in none of the 69 menin deposits; while 371-400 lies entirely in one lobe, and a 30-residue motif placed uniformly at random in a 610-residue chain overlaps a 177-residue region by at least 20 residues with probability 0.289.
analyses/q36-menin-residue-axis/run.py - Shrake-Rupley accessibility, contact-graph lobe assignment and a uniform-placement overlap null over all 69 RCSB deposits referencing UniProt O00255, run 2026-09-22 · RCSB PDB and UniProt, computed 2026What this evidence cannot show
Accessibility is computed on the isolated menin chain from one crystal form; a region's exposure changes when partners bind and 3U84 has a peptide in the pocket. Four of the region's 177 residues are unmodelled and contribute nothing to the area, so 29.9% is a slight underestimate. The lobe assignment is a contact-graph clustering, not a thermodynamic domain definition - see `ana-q36-menin-residue-axis` for its stated uncertainties, the largest of which is +/- 11 residues at the NTD|thumb boundary. The overlap null assumes uniform placement with no structural constraint and is stated as a floor: conditioning on both windows being in the ordered, modelled part of the chain would raise the probability, not lower it.
Evidence that argues against this (1)
- A large curated region can still contain a small real interface, and 29.9% of the surface is an argument that the ANNOTATION is uninformative rather than that the interaction is not real. The 2003 co-immunoprecipitation and the diepoxybutane hypersensitivity of MEN1 patient lymphocytes are both real observations that this geometry says nothing about. Itakura Y et al. 2000
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The 2026 paper that maps menin's gamma-H2A.X motif cites the 2003 menin-FANCD2 paper as its reference 18, in a single introductory sentence grouping RPA, FANCD2 and Chk1 as prior menin associations linked to homology-directed repair; it performs no FANCD2 experiment, compares no regions, and its own mapping rests on truncations of the Palm/Fingers domains followed by an internal deletion of 371-400 and a synthetic 30-mer covering the same residues.
The tumor suppressor menin is a key scaffold mediator for homologous recombination repair. · Lian A, Zhu X, Jin B, Gu G, Xu B, Zhang L, et al. · Nucleic Acids Research 2026What this evidence cannot show
Read in full here from PMC13458390 (96,530 characters). The truncation constructs are in supplementary figures S2M and S2N, which were not retrieved, so the boundaries of their 'Palm' and 'Fingers' constructs are unknown and it cannot be checked whether either of them resembles 214-390. Their system is A549 lung adenocarcinoma and Men1-knockout MEFs; there is no leukaemia experiment and no haematopoietic cell in the paper.
Evidence that argues against this (1)
- That the 2026 authors cite the 2003 paper only in a motivational list is weak evidence that they judged the two contacts distinct - it is at least as likely that they simply did not consider the question. An author's silence is not an author's verdict, and this entry should not be read as saying the field has adjudicated anything. Jin S et al. 2003
the curated 214-390 region is not a surface: it spans three spatial lobes and 29.9% of menin's total solvent-accessible area
answers that entry's stated open question: the 20-residue overlap is what chance gives you
- related to →FANCD2 — the FA effector menin binds, and the weakest FA gene in this graph's own selectivity scan
the whole protein-level menin-FANCD2 record is one 2003 co-immunoprecipitation with no mapped interface
- related to →H2AX - filed as a gene and not as a mark, and the reason that decision changes what the graph can say
the gamma-H2A.X motif is the other occupant of the region the overlap question was about
the lobe assignment and the accessibility numbers come from that parse
weakens the structural half of that hypothesis: there is no mapped menin surface for FANCD2 to occupy
a third truncation bracket, and the union of the three covers 66% of menin