IDH2 - the hypermethylating co-mutation that IS enriched in KMT2A-rearranged AML (16.0% vs 10.4%), at p = 0.0004 and not the p < 0.0001 the abstract claims

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• If the IDH2 enrichment is real in fusions, then a subset of KMT2A-rearranged patients is receiving a menin inhibitor while carrying a co-mutation that raises CpG methylation - and no menin-inhibitor trial reports IDH status as a covariate, so the interaction is unmeasurable retrospectively.
• Enasidenib and revumenib both cause differentiation syndrome. Combining an IDH2 inhibitor with a menin inhibitor in an IDH2-mutant KMT2A-rearranged patient stacks two differentiation-syndrome mechanisms, and no trial has done it.
• The abstract-versus-table p-value discrepancy in the source paper is a reminder that this entire threat rests on one retrospective commercial-panel series with no outcome data. It should not be used to stratify anybody until it is replicated in a fusion-only cohort.

1. Does the IDH2 enrichment survive restriction to KMT2A FUSIONS? 43.1% of the KMT2A-altered arm is KMT2A-PTD and the table is not stratified. One reanalysis of an annotated cohort settles the single live threat this beat identified.
2. Has any infant or paediatric KMT2A-rearranged ALL cohort ever reported IDH1/IDH2 status? Not found in this beat's searches, so the index disease has no estimate at all - not a zero, an absence of measurement.
3. In an IDH2-R140Q cell, does 5mC gain fall on fusion-bound CpG islands or elsewhere? The fusion uses 20% of uCpG sites, so the answer decides whether hypermethylation can reach it.
4. 2-HG inhibits KDM2A and KDM2B, two competing CXXC readers, as well as TET. Does mutant IDH2 therefore help the fusion at the reader-competition level while hindering it at the DNA level? Nothing separates the two.
5. Why IDH2 and not IDH1, when both make the same oncometabolite? Lesion type, compartment, power and chance are all live and none is excluded.

1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In 3,863 adult AML cases profiled by one hybrid-capture DNA and RNA assay (median age 62), IDH2 genomic alteration was significantly more frequent in the 521 KMT2A-altered cases than in the 3,342 KMT2A-wild-type cases - 16.0% versus 10.4% - while IDH1 (6.7% vs 7.6%) and DNMT3A (16.0% vs 17.2%) showed no significant difference and TET2 was depleted (10.1% vs 16.4%).

   Characterizing KMT2A Rearrangement in Acute Myeloid Leukemia: A Comprehensive Genomic Study. · Batayneh O, Moein M, Naji NS, et al. · Cancers (Basel) 2026
   PMID 41514668DOI
   ▸What this evidence cannot show

   Read in full text from PMC12784994 on 2026-09-21, including Table 2, not from the abstract alone. THE PAPER DISAGREES WITH ITSELF ON THIS ROW: the abstract gives p < 0.0001 for IDH2 and Table 2 gives p = 0.0004; the recomputation in E2 reproduces the table. Adult AML, median age 62 - this says nothing directly about infant KMT2A::AFF1 ALL. A commercial targeted panel, not whole genome; germline filtering inferred. 43.1% of the KMT2A-altered arm are KMT2A duplications rather than fusions and the co-mutation table is NOT broken down by lesion type, so the IDH2 row may be a property of KMT2A-PTD. No multiplicity adjustment is reported across the ~17 genes compared. The paper reports no clinical outcome data, which its authors name as the study's main limitation.

   ▸Evidence that argues against this (1)

   • IDH1/IDH2 mutations are mutually exclusive with TET2 mutations in AML, so an IDH2 enrichment and a TET2 depletion measured in the same cohort are not two independent observations; part of each is the arithmetic shadow of the other, and the paper performs no joint analysis that would separate them. Figueroa ME et al. 2010
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

   Recomputing the FoundationOne Heme co-mutation comparison by two-sided Fisher exact test from counts back-calculated from the published percentages gives IDH2 odds ratio 1.630, p = 4.1e-4; IDH1 odds ratio 0.876, p = 0.53; TET2 odds ratio 0.577, p = 1.6e-4; DNMT3A odds ratio 0.912, p = 0.49 - reproducing the paper's Table 2 p-values in every case and contradicting the abstract's p < 0.0001 for IDH2.

   Independent Fisher exact and Wilson-interval recomputation of the KMT2A-altered versus KMT2A-wild-type co-mutation table of PMID 41514668 · Oncofusion knowledge-graph project, Q81-methylome-axis · Recomputed 2026-09-21 from percentages read out of PMC12784994 Table 2; counts back-calculated as round(pct x n) and are accurate to about one case 2026
   ▸What this evidence cannot show

   This is a recomputation from published marginals, not from patient-level data, so it inherits every ascertainment property of the original assay and adds rounding error of about one case per cell. It can detect an internal inconsistency in the paper - which it did - but it cannot detect an error in the paper's own counting. No multiplicity adjustment is applied here either; at 17 genes a Bonferroni threshold is 0.0029, which IDH2 and TET2 both clear and nothing else does.

   ▸Evidence that argues against this (1)

   • Back-calculating counts from rounded percentages can shift a borderline p-value, and the agreement with Table 2 could in principle be coincidental rather than a validation of the method. The correct check is the authors' own count table, which is not published. Batayneh O et al. 2026
3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   In a large adult AML cohort, IDH1/2-mutant cases display global DNA hypermethylation and a specific hypermethylation signature; IDH1/2 mutations are mutually exclusive with TET2 mutations; TET2 loss-of-function produces similar epigenetic defects; and expression of mutant IDH impairs TET2 catalytic function in cells.

   Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation. · Figueroa ME, Abdel-Wahab O, Lu C, et al. · Cancer Cell 2010
   PMID 21130701DOI
   ▸What this evidence cannot show

   Adult AML, 2010, pre-dating both the KMT2A-rearranged cohorts used here; no KMT2A-rearranged subgroup is analysed and the word KMT2A does not structure the analysis. Methylation was assayed by HELP, an enrichment method with a different dynamic range from bisulfite sequencing. Mutual exclusivity is a cohort-level statistical statement whose strength depends on the mutation frequencies and the cohort size, and the paper reports it without an exclusivity test statistic in the abstract.

   ▸Evidence that argues against this (1)

   • Mutual exclusivity of IDH and TET2 is not universal - co-mutated cases are reported in later series - and treating the two as interchangeable routes to one phenotype conceals that IDH mutation also inhibits histone demethylases, which TET2 loss does not. Xu W et al. 2011
4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   2-hydroxyglutarate, produced by tumour-derived IDH1 and IDH2 mutants, is a competitive inhibitor of multiple alpha-ketoglutarate-dependent dioxygenases including histone demethylases and the TET family of 5-methylcytosine hydroxylases, occupying the same site as alpha-KG in the demethylase active site.

   Oncometabolite 2-hydroxyglutarate is a competitive inhibitor of alpha-ketoglutarate-dependent dioxygenases. · Xu W, Yang H, Liu Y, et al. · Cancer Cell 2011
   PMID 21251613DOI
   ▸What this evidence cannot show

   Biochemistry plus ectopic expression; the in vivo glioma correlation is in glioma, not leukaemia, and not in any KMT2A-rearranged system. Competitive inhibition constants differ greatly between dioxygenases, so 'inhibits TET' and 'inhibits KDM' are not equally potent statements and the paper does not rank them for the enzymes this graph cares about (KDM2A, KDM2B).

   ▸Evidence that argues against this (1)

   • Because 2-HG inhibits KDM histone demethylases as well as TET enzymes, a 2-HG-high cell has BOTH more 5mC (which should hinder a CXXC reader) and impaired KDM2A/KDM2B (which are two of the fusion's competing CXXC readers). The net effect on a KMT2A fusion is not predictable from this paper and has never been measured. Kerry J et al. 2017
5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

   IDH2 is not a dependency in any leukaemia line and is not selective for KMT2A rearrangement: DepMap Chronos mean gene effect is -0.0012 in the 11 KMT2A-rearranged lines, -0.0079 in the 43 other leukaemia lines, -0.0196 across all 1,208 lines screened, rank 10,005 of 17,728.

   Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap CRISPR Chronos via the Open Targets Platform GraphQL API) · Oncofusion knowledge-graph project · analyses/depmap-kmt2a-selective-scan-full/results.json, generated 2026-09-19T10:12:06Z; Open Targets dataVersion year 26 month 06 2026
   ▸What this evidence cannot show

   Cell lines carry wild-type IDH2 unless annotated otherwise, and mutant IDH2 is a neomorphic gain of function - a CRISPR knockout scan of wild-type IDH2 is close to uninformative about a mutant-IDH2 cell's biology, and is recorded here only to close the dependency question rather than to say anything about the co-mutation. None of the 11 KMT2A-rearranged lines is infant-derived and the arm is myeloid-enriched. Open Targets dataVersion 26.06; DepMap's own release identifier is not returned by that endpoint.

   ▸Evidence that argues against this (1)

   • A knockout scan cannot detect a neomorphic dependency: an IDH2-R140Q cell may depend on the mutant allele's product while showing no effect from losing the gene, which is precisely why enasidenib is an inhibitor of the mutant enzyme rather than a degrader. Stein EM et al. 2017
6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   The KMT2A CXXC (MT) domain binds DNA containing unmethylated CpG and 5-methyl-CpG on one or both strands precludes efficient interaction, so any rise in CpG-island methylation is mechanistically opposed to the fusion's DNA contact.

   The MT domain of the proto-oncoprotein MLL binds to CpG-containing DNA and discriminates against methylation. · Birke M, Schreiner S, Garcia-Cuellar MP, et al. · Nucleic Acids Res 2002
   PMID 11842107DOI
   ▸What this evidence cannot show

   Recombinant GST-MT in vitro with short defined oligonucleotides; no cellular methylation manipulation, and no fusion protein. The quantitative affinity reported (KD ~3.3e-8 M by SPR) is for a dimerising GST fusion and is not comparable with the ~44 uM NMR figure this graph quotes elsewhere for the isolated domain. RESOLVED 2026-09-21 by Q82. This caveat was right to flag the discrepancy and its GST-dimerisation attribution is now bounded rather than confirmed. Ten published measurements of this contact form a continuous series from 44 uM to 33 nM. A same-laboratory, same-construct (1147-1203), same-oligonucleotide pair puts GST plus a 300-to-150 mM salt drop at 5.5-fold, and a tagless-against-tagless pair (44 uM at 300 mM NaCl against 5.7 uM at 150 mM NaCl) puts about 8-fold on the salt alone - so GST dimerisation explains very little of the gap. The remaining 19-fold, between GST-1147-1203 at 0.64 uM and GST-1147-1244 at 33 nM, confounds the 41-residue post-CXXC basic patch, six CpGs against two, a 52-bp duplex against a 12-mer, and an immobilised kinetic geometry, and is unresolved. See ana-cxxc-dna-affinity-ladder. This entry's own use of the paper - qualitative methylation discrimination - is unaffected.

   ▸Evidence that argues against this (2)

   • Methylation discrimination in vitro does not establish that a physiological rise in 5mC displaces the fusion in a cell, where the fusion is additionally held by AT-hooks, menin and LEDGF - and in SEM cells the fusion's site selection is dominated by menin occupancy, not by uCpG availability. Kerry J et al. 2017
   • The reconciliation this caveat called for was published in 2008, by the same laboratory that produced the 44 uM figure, and this graph did not cite it. Erfurth et al. state the SPR value is about two orders of magnitude tighter than their ITC and attribute it to GST dimerisation and bivalent binding, and the same paper gives the MT domain boundaries as MLL 1147-1244 and an ITC ladder of 3.5 to 32 uM across five Hoxa9-derived oligonucleotides at fixed construct, salt and method. Erfurth FE et al. 2008
7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   In SEM cells (human t(4;11) KMT2A::AFF1 B-ALL), all MLL-AF4 binding sites lie on open unmethylated-CpG regions but the fusion occupies only 20% of available uCpG sites, so unmethylated CpG is abundant relative to the fusion's use of it.

   MLL-AF4 Spreading Identifies Binding Sites that Are Distinct from Super-Enhancers and that Govern Sensitivity to DOT1L Inhibition in Leukemia. · Kerry J, Godfrey L, Repapi E, et al. · Cell Rep 2017
   PMID 28076791DOI
   ▸What this evidence cannot show

   One cell line; the uCpG denominator is defined by Bio-CAP, a recombinant CXXC affinity reagent, so it is a CXXC-accessible denominator rather than a bisulfite one. Read from full text (PMC5263239) on 2026-09-21; the percentages appear in the running text and Figure S1F and not in the abstract.

   ▸Evidence that argues against this (1)

   • Abundance of uCpG at steady state says nothing about whether a hypermethylating insult removes sites at random. If IDH2-driven hypermethylation preferentially hits menin-bound promoters, the 80% headroom is irrelevant - and no experiment has distributed 5mC gain over fusion-bound versus unbound uCpG in any cell. Figueroa ME et al. 2010
8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   Human IDH2 resolves to a single reviewed UniProt entry P48735 (IDHP_HUMAN), primary gene name IDH2, 452 aa, recommended name Isocitrate dehydrogenase [NADP] mitochondrial, whose only subcellular-location comment - Mitochondrion - carries evidence code ECO:0000250, i.e. inference by similarity to rat P33198 and not an experimental observation.

   UniProtKB P48735 (IDHP_HUMAN), Isocitrate dehydrogenase [NADP], mitochondrial · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Fetched 2026-09-21 via gene_exact:IDH2 AND organism_id:9606 AND reviewed:true; one row returned; primary gene name and length read back from the same fetched record. The fields queried (ft_domain, ft_zn_fing) returned no features, so no residue coordinates are asserted here; that is an absence in the queried fields, not a claim that the protein has no domain structure. Rule 91: the absence of an ECO:0000269 for mitochondrial localisation is a statement about UniProt's curation, not evidence that IDH2 is not mitochondrial - the biochemistry is not in doubt.

   ▸Evidence that argues against this (1)

   • IDH1, the paralogue, carries experimental ECO:0000269 localisation to cytosol and peroxisome from PubMed 10521434, so the difference in evidence code between the two paralogues is a curation asymmetry in the very database a beat would use to argue that compartment explains the IDH1/IDH2 difference. UniProt Consortium 2026

• ← related toNo infant or childhood KMT2A-rearranged cohort has REPORTED IDH status — and at least one has certainly MEASURED it

  THE GAP UNDER THE ENRICHMENT. gene-idh2 records IDH2 mutation at 16.0% in KMT2A-rearranged AML against 10.4% elsewhere, recomputed at OR 1.630, p = 4.1e-4. That figure is adult, and 43.1% of its KMT2A-rearranged arm is KMT2A-PTD rather than a fusion. No infant or paediatric KMT2A-rearranged cohort has ever REPORTED an IDH frequency, so the number this project reasons from has never been measured in the disease this project is about.

• co-occurs with →KMT2A-rearranged acute myeloid leukaemia in children and infants

  IDH2 alteration enriched in KMT2A-altered AML: 16.0% of 521 vs 10.4% of 3,342, OR 1.63, Fisher p = 4.1e-4 (Table 2 p = 0.0004, NOT the abstract's <0.0001).

• related to →IDH1 - the same oncometabolite, the same assay, and NO enrichment in KMT2A-rearranged AML (6.7% vs 7.6%, p = 0.53): the control that makes the IDH2 result interpretable

  Paralogues producing the same oncometabolite; one is enriched in KMT2A-rearranged AML and the other is not, which is the asymmetry this beat could not explain.

• inhibits →TET2 is NOT a CXXC protein - its CXXC domain is a separate gene, and it is DEPLETED in KMT2A-rearranged AML rather than enriched

  Mutant IDH2 makes 2-hydroxyglutarate, a competitive inhibitor of the TET 5mC hydroxylases; IDH and TET2 mutations are mutually exclusive in AML.

• inhibits →TET1 (LCX / CXXC6) - a CXXC reader that is also a KMT2A fusion partner, and whose leukaemic fusion throws its own CXXC domain away

  2-HG is a competitive inhibitor of the TET family as a class, not of TET2 alone.

• inhibits →TET3 - the CXXC reader whose domain prefers 5-carboxylcytosine, and the only CXXC-family gene ever found mutated in a KMT2A-rearranged cohort

  2-HG is a competitive inhibitor of the TET family as a class, not of TET2 alone.

• related to →The CXXC domain (aa 1147-1195, PDB 2KKF) is a chromatin address, not a compartment address — CpG mutants lose transformation but keep their subnuclear position

  The only route by which a co-mutation could contest the fusion's address: raise 5mC at CpG islands and the CXXC contact is blocked.

• related to →The KMT2A-r DNA methylation signature — focal hypermethylation, no global hypomethylation, and a hypomethylating agent that did not help

  A candidate genetic cause of hypermethylation in KMT2A-rearranged disease that the existing signature entry does not consider.

• depends on →The measurement everybody said was missing exists: in SEM cells KMT2A::AFF1 occupies 20% of unmethylated CpG sites, CXXC1 50% and KDM2B 89%, and only the fusion tracks menin (r2 = 0.96)

  Whether hypermethylation can reach the fusion depends on the 20% figure: the fusion leaves most unmethylated CpG unused.

• inhibits →KDM2B — 513 potent ChEMBL rows that are mostly patent bins, and no structure of the domain they claim to inhibit

  2-HG inhibits KDM histone demethylases as well as TET, so a 2-HG-high cell may impair two of the fusion's competing CXXC readers at the same time as it raises 5mC.

• inhibits →KDM2A - the CXXC reader that antagonises KMT2A leukaemogenesis, and displaces wild-type MLL rather than the fusion

  Same class effect: KDM2A is an alpha-KG-dependent demethylase and a CXXC reader.

• ← related toIDH1 - the same oncometabolite, the same assay, and NO enrichment in KMT2A-rearranged AML (6.7% vs 7.6%, p = 0.53): the control that makes the IDH2 result interpretable

  Negative control relationship: identical mechanism, opposite co-occurrence result in one instrument.

• ← inhibitsEnasidenib - an approved mutant-IDH2 inhibitor sitting next to the one methylome co-mutation that IS enriched in KMT2A-rearranged AML, and never tested there

  First-in-class oral selective inhibitor of MUTANT IDH2; lowers 2-HG and relieves competitive inhibition of the alpha-KG dioxygenases.

• ← testsNo IDH inhibitor has ever been reported in KMT2A-rearranged leukaemia: 4 literature records, all reviews, and 1 registry record, an adult master protocol - and the first version of this query returned a false zero

  Tests the therapeutic consequence of the enrichment and finds none reported at any age.

• ← depends on2-hydroxyglutarate competitively inhibits the alpha-KG dioxygenases - and it hits the TET demethylases AND the KDM histone demethylases, which is why its net effect on a CXXC reader is not predictable

  The neomorphic activity that produces the oncometabolite.

• ← related toIDH and TET2 mutations are mutually exclusive in AML - so 'IDH2 enriched, TET2 depleted' in one cohort is ONE observation, not two

  Applies the same brake to the enrichment as to the depletion; neither row should be cited as independent of the other.

• ← supportsThe methylome co-mutation table, recomputed: IDH2 up at p = 0.0004 and not p < 0.0001, IDH1 and DNMT3A flat, and the abstract disagrees with its own Table 2

  Supplies the verified figure: OR 1.630, Fisher p = 4.1e-4, Wilson 13.0-19.3 vs 9.4-11.5.

• ← depends onIf the fusion's address requires unmethylated CpG, a hypermethylating co-mutation should be selected AGAINST - and IDH2 is selected FOR, which means the address is not the limiting resource

  The enrichment that falsifies the naive antagonism prediction with the sign reversed.

• causes →2-hydroxyglutarate competitively inhibits the alpha-KG dioxygenases - and it hits the TET demethylases AND the KDM histone demethylases, which is why its net effect on a CXXC reader is not predictable

  Mutant IDH2 is the source of the oncometabolite.

• part of →IDH and TET2 mutations are mutually exclusive in AML - so 'IDH2 enriched, TET2 depleted' in one cohort is ONE observation, not two

  One half of the exclusivity relationship that makes the IDH2 and TET2 rows non-independent.

• tests →If the fusion's address requires unmethylated CpG, a hypermethylating co-mutation should be selected AGAINST - and IDH2 is selected FOR, which means the address is not the limiting resource

  The observation the hypothesis has to account for, and the sign is the wrong way round for the naive prediction.

• part of →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here

  A metabolic lesion that reaches chromatin: 2-HG inhibits both the DNA and the histone demethylases.

• ← supportsThe CXXC-DNA affinity is a ten-point ladder from 44 uM to 33 nM, not two numbers 1,300-fold apart, and every step but the last is under 5.5-fold

  Same resolution for the same caveat carried by this gene entry.