Oncofusion

start / graph / ana-q34-menin-surface-geometry

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q34-menin-surface-map

ADJUDICATION: the SvID and the gamma-H2A.X motif are OVERLAPPING BUT NOT THE SAME SURFACE - and the premise that made the question urgent, that the SvID is distinct from the MLL cavity, is FALSE at the SvID's own boundary

also called SvID versus gammaH2AX adjudication · menin surface geometry

Two papers described two jobs for nearly the same stretch of menin, and neither knew about the other's result. Measuring the actual three-dimensional positions shows the two jobs live at opposite ends of that stretch, about 16 to 26 angstroms apart - but the stretch itself starts inside the pocket the drugs occupy, which is the opposite of what the graph had been assuming.

The same thing, for a scientist

The question: are the SUV39H1 interaction domain (menin 360-445, PMID 23579270) and the gamma-H2A.X recognition motif (menin Val371-Gln400, PMID 42578372) the same surface, overlapping surfaces, or coincidentally nested ranges? **Verdict: overlapping ranges with disjoint functional cores, and the nesting is an artefact of how coarsely the SvID was bracketed.** Evidence for sameness: 371-400 is entirely inside 360-445; both are placed by their authors in the Palm/Fingers region; both are mapped by co-IP of menin truncations in overexpressing cells. Evidence against: the only SvID residues whose substitution costs SUV39H1 binding, D418N and W436R, lie 18 and 36 residues C-terminal to the motif and 15.7 A and 10.2 A from Val371 in menin's fold, on helices 407-424 and 434-445, while two SvID residues inside the motif window (A411P, L413P) do NOT cost binding; the gamma-H2A.X motif is necessary AND sufficient as an isolated 30-mer peptide, which a 86-residue truncation bracket is not; and W436 sits 7.5 A from histone H3 on the nucleosome face in 8GPN while V371 sits 14 A away. **The finding that changes the graph** is separate from the verdict: the SvID's N-terminal boundary residues 363, 366, 367 and 371 line the MBM cavity at 2.3-4.3 A from ziftomenib, revumenib, DS-1594b, M-89 and the JUND peptide, and UniProt curates menin E366A/D370A as reducing KMT2A binding - so the 2013 paper's inference that the SvID is a distinct domain from the MLL cavity is wrong at its first twelve residues, and the entry this graph uses to protect the menin-interface hypothesis rests on that inference.

Go deeper
The verdict, with the evidence for each reading set out separately

Reading A - the same surface.

  • 371-400 is 100% contained in 360-445; 30 of 86 residues, 34.9% of the window.
  • Both are assigned by their authors to the Palm/Fingers region of menin.
  • Both rest on co-IP of overexpressed menin truncations; neither has a structure, an affinity or a competition assay.
  • Menin has documented conformational plasticity and all five inhibitor co-crystals are of free or peptide-bound menin, so a distance measured in those states does not bound a rearrangement.

Reading B - different surfaces that happen to overlap in sequence.

  • The SUV39H1-critical residues are D418 and W436, not anything in 371-400. Within the same window, A411P and L413P do not impair binding, so the effect is position-specific and localised to 418-436.
  • In menin's fold, V371 is 15.7 A from D418 and 10.2 A from W436 (4GQ6, min heavy-atom, single chain). The centroid of 371-400 is 17.4 A from the centroid of 401-445.
  • The gamma-H2A.X motif is sufficient: the isolated 30-mer Pep-MH pulls down H2A.X. The SvID has never been shown sufficient; no 360-445 fragment has been expressed.
  • They sit on different faces. In 8GPN, W436 is 7.5 A from histone H3 and 8.1 A from H4 - on the nucleosome grip - while V371 is 14.0 A from H3 and 371-385 points at the cavity rim.
  • Residues 389-397 of the gamma-H2A.X motif are observed in 0 of 69 menin deposits and absent from 42 of the constructs; D418 and W436 are observed in 69 of 69. One is a disordered loop, the other is structured helix.

Verdict: B, with a caveat that is not decoration. The two functions are carried by residues that are 10-26 A apart and structurally different in kind. But 'different surfaces' is not 'independent': the LEDGF IBD, a genuinely unrelated surface on the same protein, is 22.8-37.9 A from all of these. On that calibration a 10-16 A separation is near. An allosteric or steric effect of a pocket ligand on helices 10-16 A away is not excluded by geometry, and nothing in the literature tests it.

What would settle it - and it is cheap

The decisive experiment is a reciprocal deletion co-IP and it uses reagents both source laboratories already have:

  1. Express menin delta-371-400 - the construct PMID 42578372 already built and validated as gamma-H2A.X-null - and ask whether it still co-precipitates SUV39H1. If it does, the two surfaces are separable and the verdict holds. If SUV39H1 binding is also lost, the SvID's requirement is not localised to 418-436 and the two readings collapse.
  2. Express menin D418N and W436R - the constructs PMID 23579270 already built - and ask whether they still co-precipitate gamma-H2A.X after irradiation. If they do, separable.
  3. The controls that make it interpretable: menin A411P and L413P (SvID mutants that do NOT affect SUV39H1) as internal negatives, and the KMT2A MBM co-IP on every construct, because residues 366-371 are in the drug pocket and a construct that has lost KMT2A binding has lost its fold, not a contact.

This is written up as exp-reciprocal-deletion-test-of-the-svid-and-the-gammah2ax-motif. It is two co-IPs in 293T with existing plasmids.

And the structural version, if anyone wants to spend a year: a cryo-EM structure of menin on a gamma-H2A.X nucleosome. 8GPN shows menin contacting canonical H2A at exactly one residue, so the damage-induced mode is either different or requires the H2A.X tail, and no coordinate set answers that.

The entry and field this threatens, named

gene-suv39h1, the section headed "The menin connection, stated exactly, because it touches the interface hypothesis", and evidence item E3's claim field.

Current text: "It does not follow that a menin inhibitor competes SUV39H1 off menin: on this mapping the inhibitor's pocket and the SvID are different surfaces." And E3: "Menin interacts with SUV39H1 through a domain distinct from the one that binds MLL." The entry's TITLE also says "a menin effector at a surface distinct from the MLL-binding cavity".

What is wrong: the pocket and the SvID are not different surfaces. Menin 363, 366, 367 and 371 are inside the SvID as drawn AND within 4.5 A of bound revumenib, ziftomenib, DS-1594b, M-89, the JUND peptide and the KMT2A chain; UniProt curates menin E366A/D370A, from PMID 22327296, as reducing KMT2A binding.

What it should say instead: the two SvID residues whose mutation is known to cost SUV39H1 binding, D418 and W436, are 9.9-16.6 A from every clinical menin inhibitor and are not contacted in any of the five inhibitor co-crystals - so a pocket ligand is not expected to compete SUV39H1 off menin, although the SvID's own N-terminal boundary is pocket wall and the separation has never been tested experimentally.

The citations that settle it: PMID 22327296 (Huang et al., Nature 2012; full text read 2026-09-22, and the E366/D370 assignment via UniProt's curated mutagenesis features on O00255 from that paper's Supplementary Fig. 6, which was NOT readable); and the coordinate measurement in analyses/q34-menin-surface-geometry/results.json over PDB 7UJ4, 8VA6, 8IG0, 6E1A, 9WKU, 4GQ6, 3U85, 3U86, 3U88 and 8GPN.

This is reported, not edited. gene-suv39h1 belongs to the Q110 beat.

Also owed, to interface-menin-kmt2a-mbm2 (king folder): data.structures_absent says "no separate MBM2 peptide co-crystal exists" and "Arg24/Arg25 are reported disordered in the menin-MBM crystal", and open question 3 asks whether MBM2 can be crystallised at all. Measured from 3U88's coordinates today, the KMT2A chain is modelled at 6-15, 23-39 and 107-135, and Arg24 is 2.9-3.2 A from menin Asp370 and 3.0-3.5 A from Glu366 in both crystallographic copies. The disorder statement is true of 3U85 (KMT2A 5-13 modelled) and not of 3U88. See interface-menin-acidic-shelf-e366-d370-v371.

And a smaller one, to interface-psip1-ibd-kmt2a: it records 3U88 as containing 'KMT2A 103-153' while interface-menin-kmt2a-mbm2 records '5-15 + 23-39 + 109-153'. The MODELLED spans are KMT2A 6-15, 23-39, 107-135 and PSIP1 348-430.

Which of the dispatch brief's premises were wrong
premisestatus
PMID 23579270 maps SUV39H1 to menin 360-445 by truncation, with D418N and W436R losing bindingcorrect, verbatim
PMID 42578372 places a gamma-H2A.X motif at menin Val371-Gln400, DOI 10.1093/nar/gkag790correct, verbatim, DOI confirmed from PubmedData/ArticleIdList
Val371-Gln400 falls inside 360-445correct
"neither source paper mentions the other"HALF WRONG. PMID 42578372 cites PMID 23579270 as reference 7, in the bundle [6-8] supporting a sentence about menin and H3K4me3/H3K9me3/H3K36me3. It cites it and discusses nothing about it, and never uses the string SUV39H1 anywhere.
the graph uses the SUV39H1 result to protect the menin-interface hypothesiscorrect, and localised: it is one section and one evidence claim in gene-suv39h1, which is the only entry in the corpus citing PMID 23579270 or containing the string SvID
registry row 178: a global regex for <ArticleId IdType="doi"> over PMID 42578372's record returns 67 DOIsWRONG NUMBER, right trap. Measured live 2026-09-22: the global regex returns 40; <ArticleId of all types returns 114; the record carries 40 <Reference> elements. The anchored PubmedData/ArticleIdList path returns exactly one DOI, 10.1093/nar/gkag790. The trap and the remedy are real; 67 is not the count for this record.
Facts and figures
scriptanalyses/q34-menin-surface-geometry/run.py
statusrun
run date2026-09-22
complete runtrue
verdictOVERLAPPING RANGES, DISJOINT FUNCTIONAL CORES. 371-400 is nested in 360-445 by sequence, but the SUV39H1-critical residues (418, 436) and the gamma-H2A.X motif are 10-26 A apart, sit on different faces, and differ in kind (structured helix versus never-observed loop). The nesting is an artefact of the SvID being an 86-residue truncation bracket rather than a mapped interface.
second and larger findingThe premise that made the question urgent is false. The SvID's N-terminal boundary (360-371) IS the MLL/drug cavity: menin 363, 366, 367 and 371 are within 4.5 A of bound revumenib, ziftomenib, DS-1594b, M-89, the JUND peptide and the KMT2A chain, and UniProt curates menin E366A/D370A as reducing KMT2A binding. The 2013 'distinct domains' inference fails at the SvID's own edge.
result summary0 of 69 menin deposits observe residues 389-397; 69 of 69 observe D418 and W436. V371-D418 15.7 A, V371-W436 10.2 A. V371 is 2.3-4.3 A from 3 of 4 clinical menin inhibitors and 10.7 A from the fourth. D418 and W436 are 9.9-19.5 A from all of them. The LEDGF control surface is 22.8-37.9 A from all four positions.
what would settle the adjudicationA reciprocal deletion co-IP: (1) does menin delta-371-400 still bind SUV39H1? (2) do menin D418N and W436R still bind gamma-H2A.X after irradiation? With menin A411P/L413P as internal negatives and a KMT2A MBM co-IP on every construct as a fold control, because residues 366-371 are pocket wall. Written up as exp-reciprocal-deletion-test-of-the-svid-and-the-gammah2ax-motif.
inputs
[
  "UniProt O00255, fetched live",
  "all 69 RCSB entries whose polymer entity references O00255",
  "PMID 23579270 full text (51,918 chars, Europe PMC fullTextXML)",
  "PMID 42578372 full text (96,530 chars, Europe PMC fullTextXML)",
  "PMID 22327296 full text (41,215 chars, efetch db=pmc after EPMC 500)"
]
key numbers
{
  "n_pdb_deposits_swept": 69,
  "residues_371_400_never_observed": [
    389,
    390,
    391,
    392,
    393,
    394,
    395,
    396,
    397
  ],
  "n_constructs_deleting_387_398": 42,
  "svid_length_aa": 86,
  "gamma_motif_length_aa": 30,
  "fraction_of_svid_covered_by_gamma_motif": 0.349,
  "v371_to_d418_angstrom": 15.7,
  "v371_to_w436_angstrom": 10.2,
  "centroid_371_400_to_401_445_angstrom": 17.4,
  "ledgf_ibd_control_distance_range_angstrom": "22.8 to 37.9"
}
entries this threatens
[
  {
    "entry": "gene-suv39h1",
    "fields": [
      "title (the clause 'a menin effector at a surface distinct from the MLL-binding cavity')",
      "sections[] heading 'The menin connection, stated exactly, because it touches the interface hypothesis', second paragraph first sentence",
      "evidence E3 claim ('through a domain distinct from the one that binds MLL')"
    ],
    "what_is_wrong": "the SvID is not a different surface from the MLL cavity; residues 363-371 of the SvID line the cavity and E366A/D370A reduce KMT2A binding",
    "what_it_should_say": "the two SvID residues known to be required for SUV39H1 binding, D418 and W436, are 9.9-16.6 A from every clinical menin inhibitor and are contacted by none of them, although the SvID's N-terminal boundary is pocket wall and the separation has never been tested experimentally",
    "citation_that_settles_it": "PMID 22327296 (read in full 2026-09-22) plus UniProt O00255 curated Mutagenesis features E366A/D370A, plus the coordinate measurement in analyses/q34-menin-surface-geometry/results.json",
    "read_the_paper_or_only_the_abstract": "full text read; the specific E366/D370 assignment is from UniProt's curation of that paper's Supplementary Fig. 6, which was NOT retrievable"
  },
  {
    "entry": "interface-menin-kmt2a-mbm2",
    "fields": [
      "data.structures_absent (both the 'no menin-MBM2 complex' and the 'Arg24/Arg25 disordered' items)",
      "open_questions[2] ('Can MBM2 be crystallised at all')",
      "data.structure_details[0].kmt2a_construct_maps_to"
    ],
    "what_is_wrong": "MBM2 is ORDERED in 3U88: the KMT2A chain is modelled at 6-15, 23-39 and 107-135, and Arg24 contacts menin Asp370 (2.9-3.2 A) and Glu366 (3.0-3.5 A) in both crystallographic copies",
    "what_it_should_say": "no deposit is ANNOTATED as a menin-MBM2 complex and Arg24/Arg25 are disordered in the 21-mer peptide structure 3U85, but they are resolved in 3U88 where MBM2 contacts the menin acidic shelf",
    "citation_that_settles_it": "PDB 3U88 coordinates, read 2026-09-22; see interface-menin-acidic-shelf-e366-d370-v371",
    "read_the_paper_or_only_the_abstract": "coordinates read directly; Huang et al. full text read"
  },
  {
    "entry": "interface-psip1-ibd-kmt2a",
    "fields": [
      "data.structure_details, the 3U88 'contents' string"
    ],
    "what_is_wrong": "it says 'MEN1 2-550 + KMT2A 103-153 + PSIP1 347-435', which describes the constructs; the MODELLED spans are menin 2-550 (with gaps), KMT2A 6-15/23-39/107-135 and PSIP1 348-430, and a sibling entry gives a third, different KMT2A span",
    "what_it_should_say": "distinguish construct from model, and give both",
    "citation_that_settles_it": "PDB 3U88 coordinates and _pdbx_poly_seq_scheme, read 2026-09-22",
    "read_the_paper_or_only_the_abstract": "coordinates read directly"
  }
]
bounded absences
{
  "no_structure_of_menin_with_suv39h1": "0 of 69 O00255 deposits contain SUV39H1; RCSB polymer-entity search on O00255, 2026-09-22",
  "no_structure_of_menin_with_h2ax": "0 of 69; the only nucleosome complex, 8GPN, contains canonical H2A type 1 (P06897)",
  "no_measurement_of_any_menin_partner_under_a_menin_inhibitor_for_these_two_contacts": "neither SUV39H1 nor gamma-H2A.X co-precipitation has been reported under revumenib, ziftomenib, bleximenib or any tool compound, in any system",
  "absence_class": "never-measured"
}
falsifiers
[
  "THE ADJUDICATION FAILS if menin delta-371-400 - the internal-deletion construct of PMID 42578372, which spans exactly menin residues 371-400 - loses SUV39H1 co-immunoprecipitation in HEK-293T overexpression, defined as co-IP band intensity at or below 30% of wild-type menin in the same blot across three biological replicates, while the same construct retains KMT2A MBM co-IP at or above 70% of wild type as the fold control. That result would mean the 86-residue SvID bracket is one functional unit and the two contacts are the same surface.",
  "THE ADJUDICATION ALSO FAILS if menin D418N or W436R - single substitutions on helices 407-424 and 434-445 - lose gamma-H2A.X co-immunoprecipitation in irradiated HEK-293T or A549 cells (10 Gy, harvest 1 h), defined as at or below 30% of wild type across three biological replicates, with the KMT2A MBM fold control at or above 70%. The internal negatives A411P and L413P must stay at or above 70% for SUV39H1 binding in the same experiment or the assay is uninterpretable.",
  "THE 'DRUG IS INNOCENT OF THE SvID' READING FAILS if endogenous SUV39H1 co-immunoprecipitation with menin falls by more than 2-fold in a KMT2A-rearranged human line (MV4-11 or SEM) after 4 hours at 1 uM revumenib, at a concentration where the menin-KMT2A co-IP is demonstrably reduced. D418 and W436 are 9.9-16.6 A from the compound, so geometry predicts no change; a 2-fold fall would show that distance does not bound perturbation and would retire the geometric argument in this graph.",
  "THE STRUCTURAL CLAIM ABOUT MBM2 FAILS if a re-refinement or a validation report for PDB 3U88 shows the KMT2A 23-39 segment modelled into density below a real-space correlation coefficient of 0.7, i.e. placed rather than observed. The claim rests on coordinates read from the deposit and on agreement between the two crystallographic copies; it does not rest on a density inspection, which this beat did not do."
]
How this could bite us back
  • The direction of the correction is unhelpful for the interface hypothesis. The SUV39H1 result was carried as protective; the protection is now narrower, conditional on two residues, and accompanied by the finding that three of four clinical compounds touch Val371 of a motif required for accurate DNA repair.
  • If a pocket ligand does perturb menin's DNA-damage scaffolding, the population at risk is children receiving a menin inhibitor concurrently with TOP2 poisons and anthracyclines - which is the standard backbone in this disease and is already in this graph as fifteen drug entries.
  • The cheapest experiments here are also the ones that could most embarrass the current position, which is a reason to run them rather than a reason to defer.
  • A degrader removes every surface adjudicated here at once, so the whole adjudication is moot for that modality and the risk is uniformly larger - see risk-a-menin-degrader-is-not-a-pocket-inhibitor-for-the-hr-scaffold.
Still open
  1. Does menin delta-371-400 still bind SUV39H1? This single co-IP decides the adjudication and both plasmids exist in published laboratories.
  2. Do D418N and W436R retain gamma-H2A.X binding after irradiation? The mirror-image experiment, equally cheap.
  3. Is a 10-16 A separation enough? The LEDGF calibration says a genuinely independent surface is 23-38 A away, so this graph should stop treating 'not in the pocket' as 'not perturbed by a pocket ligand' until somebody runs a co-IP under drug.
  4. How many other 'distinct domain' claims in this corpus rest on a truncation bracket rather than on residues? The SvID took fourteen years and one coordinate calculation to come apart, and nothing flagged it in the interim.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    Over all 69 RCSB deposits referencing O00255, menin residues 389-397 are observed in zero and D418/W436 in 69; Val371 lies 15.7 A from D418 and 10.2 A from W436 within a single menin chain in 4GQ6; the centroid of residues 371-400 is 17.4 A from the centroid of 401-445; Val371 is 2.3-4.3 A from ziftomenib, DS-1594b, M-89 and revumenib but 10.7 A from bleximenib; D418 and W436 are 15.5-19.5 A and 9.9-14.3 A from all five compounds; and the LEDGF IBD, a control surface, is 22.8-37.9 A from all of I360, V371, D418 and W436.

    What this evidence cannot show

    Computed 2026-09-22 against live RCSB mmCIF and live UniProt; minimum heavy-atom distances per chain pair; auth numbering proved equal to UniProt numbering by residue-identity readback at eight positions per deposit. Hydrogens absent. Every structure measured is free menin, menin plus a short peptide, or menin plus a small molecule - none contains SUV39H1, H2A.X, or a nucleosome other than 8GPN's canonical one, so all distances describe states that are not the states in dispute. Two errors made and corrected during the analysis are recorded in the script docstring: menin is entity 7 not entity 1 in 8GPN, and an all-chain minimisation produced a spurious 1.3 A between menin 386 and 399 in 4GQ6, which is the signature of that construct's engineered 387-398 deletion.

    Evidence that argues against this (1)
    • A distance measured in a crystal of free or peptide-bound menin is not a statement about a ligand's ability to perturb a contact 10-16 A away. The strongest version of the opposing case is that menin's cavity ligands are large and buried, that helix 407-424 packs against the cavity helices, and that no experiment has ever asked whether a pocket inhibitor changes SUV39H1 or H2A.X co-precipitation - so the geometric argument is the only argument, and geometry is not thermodynamics. Huang J et al. 2012
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PMID 42578372 cites PMID 23579270 as its reference 7, inside the bundle [6-8] supporting the sentence that menin interacts with epigenetic writers to orchestrate H3K4me3, H3K9me3 and H3K36me3, and its full text contains no occurrence of SUV39H1, SvID, MBM, cavity, pocket, revumenib or 'menin inhibitor'.

    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, Lin X, Zhu X, Jin G · Nucleic Acids Res 2026
    PMID 42578372DOI
    What this evidence cannot show

    Case-insensitive regex over the 96,530-character Europe PMC fullTextXML flattened text, fetched 2026-09-22. This bounds the body, figure legends and reference titles; it does not reach supplementary material or text inside figure images (rule 80). A bundle citation with no discussion is weak evidence of engagement, but it is not zero and the brief's premise said 'neither paper mentions the other'.

    Evidence that argues against this (1)
    • The 2013 paper cannot cite the 2026 one, so 'neither mentions the other' was never symmetrically testable. The asymmetric version - that the 2026 authors had the 2013 result in hand and did not engage with it - is the one that matters, and it is confirmed. Yang YJ et al. 2013
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    A global regex for the string <ArticleId IdType="doi"> over the live efetch XML of PMID 42578372 returns 40 matches, not 67; the record contains 114 <ArticleId elements of all types and 40 <Reference> elements, and the single DOI under PubmedData/ArticleIdList is 10.1093/nar/gkag790.

    NCBI E-utilities efetch, db=pubmed, id=42578372, retmode=xml · National Center for Biotechnology Information · NCBI Entrez 2026
    What this evidence cannot show

    Measured 2026-09-22. PubMed records accrue reference lists over time, so a count taken on a different date can differ - which is itself an argument for never quoting a reference-list-derived count as a fixed property. The registry row's remedy (take the DOI only from PubmedData/ArticleIdList) is correct and is what this beat used for every citation.

    Evidence that argues against this (1)
    • The discrepancy could be a difference of method rather than of fact: 114 ArticleId elements of all types are present, so a regex counting all ArticleIds on a subset, or a count taken against esummary rather than efetch, could plausibly land near 67. The registry row should record the exact command as well as the number - which is the general lesson, and it is the same one rule 94 states about pasting querytranslation beside a count. National Center for Biotechnology Information 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Within the SvID, the two patient mutants that lose SUV39H1 binding (D418N, W436R) and the two that do not (A411P, L413P) are all C-terminal to residue 400, so every residue-level functional datum about the SUV39H1 contact lies outside the gamma-H2A.X motif; conversely the gamma-H2A.X motif is necessary (delta-371-400 abolishes binding) and sufficient (the isolated 30-mer precipitates H2A.X), a standard the SvID has never been held to.

    Menin mediates epigenetic regulation via histone H3 lysine 9 methylation. · Yang YJ, Song TY, Park J, Lee J, Lim J, Jang H, Kim YN, Yang JH, Song Y, Choi A, Lee HY, Jo CH, Han JW, Kim ST, Youn HD, Cho EJ · Cell Death Dis 2013
    PMID 23579270DOI
    What this evidence cannot show

    Both source papers used overexpression co-IP, and a truncation or deletion in a compact alpha-helical protein can report folding rather than contact. The four patient mutants were tested by in-vitro-translation pulldown with partially purified FLAG-SUV39H1, with no fold control. Neither paper tested the other's construct, which is exactly the gap the proposed experiment closes.

    Evidence that argues against this (1)
    • The strongest case for Reading A is that a menin delta-371-400 protein has never been tested for SUV39H1 binding. If it turns out SUV39H1 binding is also lost in that construct, the 86-residue window really is one functional unit and this adjudication is wrong. That is the falsifier and it is written to be scoreable. Lian A et al. 2026
Connected to