Menin does not outlast the fusion under revumenib in primary infant material - 0.586 against 0.485 at oncoprotein peaks, a ratio of 1.21 against a pre-registered threshold of 1.5 - so the fusion-independent menin plateau seen after a degron does not reproduce here
After the fusion protein is destroyed in a lab cell line, about half the menin stays put, which is why this project believes something other than the fusion holds it there. In a real patient on a menin-blocking drug, menin and the fusion fall away together. The lab result does not reproduce - but the two experiments remove menin by completely different routes, so this is a weaker blow than it looks.
The same thing, for a scientist
Pre-registered outcome **B**, scored on the declared threshold. Over the authors' 1,590 oncoprotein-target peak intervals, the 152985/148752 median density ratio is **menin 0.586** and **KMT2A-N 0.485** - a menin-over-fusion ratio of **1.21**, rising to **1.45** after each channel's own genome-wide movement is divided out, against a pre-registered firing threshold of **1.5**. The ENL leg of the falsifier also fails: ENL's ratio at the same peaks is **0.449**, below the declared 0.75. **The falsifier does not fire.** The previous cycle's KMT2A::MLLT3 degron result - menin retaining 50.3% at fusion-occupied promoters three hours after the fusion fell to 1.0-2.8% - has no counterpart here: menin and the fusion move together within 21%. Read strictly, that damages the generality of the loose-coupling reading. Read carefully, it barely touches its mechanism, for the reason set out in the companion concept entry: **revumenib removes menin by occupying menin's own pocket, so every model of what tethers menin at fusion sites predicts menin leaving.**
The scored falsifier, verbatim from the pre-registration
In the patient-matched pair 148752 -> 152985 (primary infant KMT2A::MLLT1, diagnosis -> relapse on revumenib), the median 152985/148752 CPM ratio for menin over the KMT2A-N-biased promoter set defined in 148752 is at least 1.5x the median ratio for KMT2A-N over the same promoters, while ENL's median ratio over the same promoters exceeds 0.75.
Scored: menin/N = 1.21 (1.45 after genome-wide correction), ENL = 0.449. DOES NOT FIRE.
Per the project's convention, the outcome is recorded inside the falsifier text wherever it appears: RUN IN CYCLE 122, DOES NOT FIRE.
The full channel table, published whatever the verdict
152985/148752 median density ratio on the authors' own peak intervals, and the same quantity genome-wide in 5 kb bins as the 'everything moves' control 1:
| channel | oncoprotein peaks (n=1590) | ALL-specific (n=52) | AML-specific (n=45) | genome-wide 5 kb | specificity index |
|---|---|---|---|---|---|
| menin | 0.586 | 0.441 | 0.318 | 1.074 | 0.55 |
| KMT2A-N | 0.485 | 0.278 | 0.302 | 1.289 | 0.38 |
| KMT2A-C | 0.610 | 0.481 | 0.432 | 1.153 | 0.53 |
| ENL | 0.449 | 0.266 | 0.548 | 1.020 | 0.44 |
| DOT1L | 0.538 | 0.482 | 0.421 | 0.620 | 0.87 |
The specificity index (peak ratio divided by genome-wide ratio) is the useful column: it asks whether a channel moved at fusion sites or everywhere. Menin, KMT2A-N, KMT2A-C and ENL all lost roughly half their signal specifically at oncoprotein peaks; DOT1L at 0.87 moved almost as much everywhere as it did at peaks, so DOT1L's loss is the least site-specific of the five.
The sharper cut, which gives a much larger number and a smaller claim
On the top 50 promoters defined in the drug-free aliquot rather than on the authors' 13 kb-median peak domains, the ratios are menin 0.037, KMT2A-N 0.028, KMT2A-C 0.113, ENL 0.071, DOT1L 0.040 - a 25- to 35-fold collapse rather than a halving. The two measures differ by an order of magnitude because the authors' peaks are broad domains that dilute a focal change, exactly as a broad interval always will. The menin-over-fusion ratio is 1.32 on this measure, still below the threshold, so the verdict is stable across both interval definitions even though the magnitudes are not.
Which direction this actually cuts, both ways, as required
Against loose coupling. The clean prediction of 'half the menin at fusion promoters is held by something that is not the fusion' is a menin surplus over the fusion when the fusion goes. There is no surplus here worth the name. In the only primary infant material anyone has, menin's residency at fusion promoters is not separable from the fusion's.
For loose coupling, or at least not against it. (i) A menin inhibitor is not a fusion degron; it strips menin from every KMT2A-family site including wild-type MLL1, so a fusion-independent menin pool that still needs the menin pocket is removed too, and the observed joint loss is the prediction of both models. (ii) The comparator is a different clone at relapse, with intrinsically reduced oncoprotein by the paper's own account. (iii) Nothing in the drug arm is above background, so this is a ratio of two small numbers whose replicate agreement is 0.25. (iv) The drug-free half of the same patient does show menin sitting with the fusion N-terminus and not with the wild-type C-terminus, which is the positive half of the loose-coupling picture and is the stronger result of the two.
| preregistered outcome | B |
|---|---|
| falsifier | RUN IN CYCLE 122, DOES NOT FIRE: menin/KMT2A-N = 1.21 (1.45 after genome-wide correction) against a threshold of 1.5; ENL = 0.449 against a threshold of >0.75 |
| oncoprotein peaks n | 1590 |
| fusion | KMT2A::MLLT1, not the index fusion - carry the ordering, not the magnitude |
| status | run |
| script | analyses/menin-under-revumenib-in-an-infant-patient-pair/run2_controls.py |
| preregistration | analyses/menin-under-revumenib-in-an-infant-patient-pair/PREREGISTRATION.md |
ratios 152985 over 148752 — the numbers behind the chart
{
"MENIN": 0.586,
"KMT2A_N": 0.485,
"KMT2A_C": 0.61,
"ENL": 0.449,
"DOT1L": 0.538
}genomewide 5kb ratios — the numbers behind the chart
{
"MENIN": 1.074,
"KMT2A_N": 1.289,
"KMT2A_C": 1.153,
"ENL": 1.02,
"DOT1L": 0.62
}specificity index — the numbers behind the chart
{
"MENIN": 0.546,
"KMT2A_N": 0.376,
"KMT2A_C": 0.529,
"ENL": 0.44,
"DOT1L": 0.867
}sharper interval definition top50 drugfree promoters — the numbers behind the chart
{
"MENIN": 0.037,
"KMT2A_N": 0.028,
"KMT2A_C": 0.113,
"ENL": 0.071,
"DOT1L": 0.04,
"menin_over_kmt2a_n": 1.32
}comparison point
{
"system": "KMT2A::MLLT3 degron, HCB1",
"menin_3h": 0.503,
"menin_4d": 0.437,
"fusion_3h": "1.0-2.8%",
"why_not_comparable": "a degron removes the fusion and leaves menin; revumenib removes menin"
}- The loose-coupling reading is the basis for arguing that menin monotherapy leaves a reservoir. If it does not generalise, that argument weakens - and this beat is the first evidence pointing that way, even though it is weak evidence.
- The opposite risk is worse: treating this non-reproduction as a refutation would licence menin monotherapy on the grounds that menin and the fusion leave together, in a patient who RELAPSED on that monotherapy with the fusion transcript still expressed.
- What would a fusion DEGRON in primary infant KMT2A::MLLT1 cells show? That is the experiment that would make this comparable to the KMT2A::MLLT3 result, and it does not exist in any primary material.
- Was MEN1 sequenced in this patient? A pocket mutation would make the occupancy result a statement about resistance rather than about pharmacology, and the raw sequencing for these two samples is public.
- Does the menin-over-fusion ratio exceed 1.5 in any menin-inhibitor occupancy dataset in any system? This project has not scored the existing cell-line datasets on this quantity.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Over 1,590 oncoprotein-target peak intervals the 152985/148752 median density ratio is 0.586 for menin and 0.485 for KMT2A-N, a menin-over-fusion ratio of 1.21 and 1.45 after genome-wide correction - below the pre-registered 1.5.
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 2026What this evidence cannot show
n = 1 patient; the two arms are different clones in different lineages; no spike-in, so both numbers are within-library shares; the drug arm's CUT&Tag replicates agree at 0.246. Peak-interval signal was derived from 5 kb bins whose centres fall inside each peak, with density computed over the actual binned length - unbiased for a 13 kb median peak but coarse.
Evidence that argues against this (2)
- The degron experiment this is being compared against removes the fusion and leaves menin untouched; revumenib removes menin directly. The two experiments cannot be placed on the same axis, so a non-reproduction here is not a refutation of the degron result. Olsen SN et al. 2022
- Menin-inhibitor resistance in patients arises through MEN1 pocket mutations that prevent the drug binding, which means occupancy under drug in a resistant relapse sample may report resistance genotype rather than drug pharmacology - and MEN1 was not sequenced in this patient. Perner F et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The previous cycle measured menin retaining 50.3% at fusion-occupied promoters 3 h after a KMT2A::MLLT3 degron reduced the fusion to 1.0-2.8%, and 0.437 at 4 days - a plateau rather than slow kinetics.
MLL::AF9 degradation induces rapid changes in transcriptional elongation and subsequent loss of an active chromatin landscape · Olsen SN, Godfrey L, Healy JP, Choi YA, Kai Y, Hatton C, Perner F, Haarer EL, Nabet B, Yuan GC, Armstrong SA · Molecular Cell 82:1140-1155.e11 2022What this evidence cannot show
KMT2A::MLLT3 in HCB1 cord-blood cells, n = 1 per arm, housekeeping-normalised with no spike-in, so that number is also a within-sample share and a relative one.
Evidence that argues against this (1)
- A DOT1L inhibitor strips menin from those same promoters about as effectively as destroying the fusion does (0.467 vs 0.437), so menin displacement is not diagnostic of the menin-KMT2A interface being broken - which independently limits how much any menin-displacement measurement can say about the interface. Krivtsov AV et al. 2019
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
DOT1L's loss between the two aliquots is the least site-specific of the five channels: specificity index 0.87 against 0.38-0.55 for the others.
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 2026What this evidence cannot show
DOT1L is the weakest channel in this deposit, with replicate agreement of 0.311 in the drug-free sample and 0.110 in the drug arm, so this index is the least reliable number in the table.
Evidence that argues against this (1)
- The paper reports DOT1L as reduced to undetectable levels alongside menin, which is a stronger statement than a specificity index of 0.87 supports; the disagreement is about the denominator, not the data. Janssens DH et al. 2024
the confound bounds this result
the drug arm's replicate agreement bounds this result
the drug arm is at background in every channel
menin's retention at fusion peaks relative to the fusion's
- related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
the degron result this fails to reproduce, in a different fusion and a different perturbation
the perturbation
the falsifier is scored on one of these three denominators
the reason the failed falsifier is weaker evidence than it looks
the scored falsifier
the design that would make the falsifier's quantities measurable