Oncofusion

start / graph / ana-q143-enl-redistributes-to-the-gmp-like-peaks-rather-than-simply-persisting-and-both-instruments-agree-on-the-direction

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q143-menin-under-drug-in-infants

ENL redistributes rather than persists: it is the only channel that holds more of its signal at the myeloid-program peaks than at the B-lineage ones, by 2.06x in my instrument and 1.26x in the authors' own - and in absolute terms it falls further than menin does

The published headline is that ENL stayed on the fusion's target genes while menin and the fusion left, but ENL actually fell at least as much as menin did. What it did do, uniquely among the five proteins measured, is hold on at the genes of the new myeloid identity while letting go of the B-cell ones.

The same thing, for a scientist

The paper reports that *"ENL remained bound at appreciable levels over numerous canonical oncoprotein-target genes"* while the oncoprotein, menin and DOT1L became undetectable. Scored on the deposited fragments, the **absolute** half does not reproduce: ENL's 152985/148752 ratio over the authors' oncoprotein-target peaks is **0.449**, the *lowest* of the five channels and below menin's 0.586, and in the drug arm ENL sits at **0.80x** of an out-of-sample non-specific reference. The **relative** half reproduces robustly. Retention at AML-specific peaks divided by retention at ALL-specific peaks is **ENL 2.06**, against menin 0.72, DOT1L 0.88, KMT2A-C 0.90 and KMT2A-N 1.09 - ENL is the only channel above 1.2. The direction holds in **both** of ENL's library pairs taken separately (2.16 and 3.41) and in the **authors' own normalised table** (1.26), so it is robust to the drug arm's poor replicate agreement even though its magnitude is not.

Go deeper
The two halves of the published claim, scored separately
the paper saysscored hereverdict
ENL remained bound at appreciable levelsratio 0.449, the lowest of five channels; 0.80x of a non-specific reference in the drug armdoes not reproduce as an absolute statement
ENL was only significantly enriched over the GMP-like program in the AML sample, having been enriched over both programs in the B-ALL sampleAML-specific / ALL-specific retention = 2.06 (mine), 1.26 (authors'), 2.16 and 3.41 per replicatereproduces, as a redistribution

The distinction matters because the paper's therapeutic proposal - "warrant investigation of menin inhibitor and ENL inhibitor combination therapies" - rests on ENL being a persistent target in the drug-resistant state. A redistribution supports that proposal; an absolute persistence would support it more strongly, and the data support the weaker version.

The per-channel table

152985/148752 median density ratio, by the authors' own peak class 1:

channelALL-specific (n=52)AML-specific (n=45)AML/ALL
menin0.4410.3180.72
KMT2A-N0.2780.3021.09
KMT2A-C0.4810.4320.90
ENL0.2660.5482.06
DOT1L0.4820.4210.88

And in the authors' own normalised table over the same intervals: menin 0.905, ENL 1.259, DOT1L 1.024 2. Two instruments, two normalisations, same ordering and same single channel above 1.

The sample sizes, which are small and must be quoted with the ratios

n = 52 ALL-specific and n = 45 AML-specific peaks. These are not distributions from which a confident effect size can be read; the defensible statement is the ordering - ENL is the only one of five channels whose retention is higher at the myeloid peaks - and that ordering is reproduced independently in the authors' own table and in each of the two library pairs. The magnitude (2.06 versus 1.26) differs by 63% between instruments, which is the honest spread.

What holds this ENL up, if anything does

ENL binds acylated histones through its YEATS domain and is a component of the super elongation complex; neither route requires the KMT2A N-terminus or the menin pocket. So an ENL that survives a menin inhibitor better than menin does is mechanistically unsurprising - and it is the one channel in this pair whose persistence is not predicted to be lost by the drug's own mechanism. That is why it is the informative channel here and menin is not.

Facts and figures
published claimENL remained bound at appreciable levels over numerous canonical oncoprotein-target genes while oncoprotein, menin and DOT1L became undetectable
peak classes defined inthe p279/p318 KMT2A::AFF1 pair, not in this patient
why enl is the informative channelENL's chromatin engagement runs through its YEATS domain and the super elongation complex, neither of which requires the menin pocket the drug occupies
therapeutic consequencesupports the authors' menin-plus-ENL combination proposal in its weaker, redistribution form
statusrun
scriptanalyses/menin-under-revumenib-in-an-infant-patient-pair/run2_controls.py
absolute half
{
  "verdict": "does not reproduce",
  "enl_ratio_at_oncoprotein_peaks": 0.449,
  "menin_ratio_same_peaks": 0.586,
  "enl_vs_nonspecific_reference_in_drug_arm": 0.8
}
relative half
{
  "verdict": "reproduces",
  "aml_over_all_specific_retention": {
    "MENIN": 0.72,
    "KMT2A_N": 1.09,
    "KMT2A_C": 0.9,
    "ENL": 2.06,
    "DOT1L": 0.88
  },
  "enl_per_library_pair": [
    2.164,
    3.413
  ],
  "authors_own_table": {
    "MENIN": 0.905,
    "ENL": 1.259,
    "DOT1L": 1.024
  },
  "n_peaks": {
    "ALL_specific": 52,
    "AML_specific": 45
  }
}
How this could bite us back
  • The authors propose menin-plus-ENL combination therapy on the strength of ENL persistence. The persistence is a redistribution, not a retention, and it is measured in one patient with drug-arm replicates agreeing at 0.247.
  • An ENL inhibitor added to a menin inhibitor would be given to a patient whose leukaemia has already switched lineage. If ENL's role in the switched state is to maintain a myeloid program rather than the fusion program, blocking it may select for a third state rather than restore sensitivity.
  • ENL YEATS inhibition also hits ENL's normal role in haematopoiesis and in MLL-independent elongation, so the therapeutic window in an infant is not established by this result.
Still open
  1. Does ENL redistribution to the myeloid program happen in lineage switches that occur WITHOUT a menin inhibitor - for instance the p279/p318 pair in this same deposit? That would separate drug from lineage switch and is computable from files already public.
  2. Is the ENL that persists at GMP-like peaks YEATS-domain-dependent? An ENL YEATS inhibitor in a lineage-switched sample would answer it.
  3. The peak classes come from a KMT2A::AFF1 pair. Do this patient's own ALL- and AML-specific peaks give the same answer? They were not derived here.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    ENL is the only one of five channels whose retention is higher at AML-specific than at ALL-specific peaks (2.06), and the direction holds in each of its two library pairs separately (2.16, 3.41).

    analyses/menin-under-revumenib-in-an-infant-patient-pair - pre-registered re-analysis of GSE252378 samples 148752 and 152985 · Q143-menin-under-drug-in-infants (Oncofusion Knowledge Graph) · this repository 2026
    What this evidence cannot show

    n = 52 and n = 45 peaks. The drug arm's ENL replicates agree with each other at Spearman 0.247 over promoters, so the magnitude is unreliable; the direction is reported because it is reproduced per-replicate and in a second instrument. Peak classes are the authors' construction, derived from the p279/p318 pair, not from this patient.

    Evidence that argues against this (1)
    • The ALL-specific and AML-specific peak classes were defined in a DIFFERENT patient pair (p279/p318, KMT2A::AFF1), so applying them to this KMT2A::MLLT1 patient imports a definition across fusion backgrounds - which this project has measured to transfer in ordering and not in magnitude. Janssens DH et al. 2024
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    The authors' own normalised signal table over the same peak intervals gives the same ordering: ENL 1.26 AML/ALL, menin 0.905, DOT1L 1.024.

    MLL_oncoprotein_levels_NatComms - processed data tables and the figure-generation notebook for Janssens et al. 2024 · Janssens DH · GitHub / Zenodo (the paper's Code Availability statement) 2024
    DOI
    What this evidence cannot show

    Their normalisation divides by total base-pair coverage of the whole fragment file and adds a pseudocount of 1, so it is a within-library share like mine and is subject to the same inability to see a global change. Two instruments sharing that limitation agreeing is weaker evidence than it appears.

    Evidence that argues against this (1)
    • Two instruments that share a normalisation family can agree while both being wrong in the same direction: the authors' measure and mine are both within-library shares, so neither can see a global change in ENL, and their agreement is not independent confirmation that ENL's absolute level is preserved at myeloid peaks. Janssens DH 2024
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    ENL's absolute retention over the authors' oncoprotein-target peaks is 0.449, the lowest of the five channels and below menin's 0.586.

    analyses/menin-under-revumenib-in-an-infant-patient-pair - pre-registered re-analysis of GSE252378 samples 148752 and 152985 · Q143-menin-under-drug-in-infants (Oncofusion Knowledge Graph) · this repository 2026
    What this evidence cannot show

    Within-library shares, different clones, n = 1 patient. The authors' own table over the same intervals gives ENL 0.827 and menin 0.871 - the same ordering but a much milder effect, so the two instruments disagree on magnitude while agreeing that ENL is not retained MORE than menin.

    Evidence that argues against this (1)
    • The paper's claim is anchored on named loci - MBNL1, RUNX2, KLF6 - shown in genome-browser tracks, not on a genome-wide median, and a claim about specific loci is not refuted by a median over 1,590 intervals. Janssens DH et al. 2024
Connected to