Oncofusion

start / graph / hyp-q136-catalytically-inhibited-dot1l-is-retained-on-chromatin-at-fusion-target-genes

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q136-menin-vs-dot1l-inhibitors

Under all three DOT1L inhibitors, DOT1L's fold-enrichment at canonical target loci is HIGHER than DMSO (PBX3 4.85 to 13.92, SENP6 3.37 to 9.68) - either catalytically dead DOT1L is retained or trapped on chromatin, or the reading is an artefact of a 1.35-fold spread in MACS fragment-size estimates, and this deposit cannot tell which

also called DOT1L trapping hypothesis · the confounded gain

When the DOT1L enzyme is switched off by a drug, it seems to stay stuck on the leukaemia genes even harder than before. That would be interesting, but the experiment has a technical quirk that could produce the same appearance, so it is written down as a question rather than a result.

The same thing, for a scientist

**The observation.** In GSE149183's DOT1L ChIP arm, every DOT1L inhibitor raises DOT1L's fold-enrichment at canonical target loci above DMSO [E1]: | locus | DMSO | EPZ5676 | compound 10 | compound 11 | |---|---:|---:|---:|---:| | HOXA9/HOXA10 | 2.56 | 4.39 | 5.08 | **7.47** | | MEIS1 | 3.67 | 5.65 | 4.93 | 6.82 | | PBX3 | 4.85 | 8.32 | 9.41 | **13.92** | | SENP6 | 3.37 | 6.17 | 6.03 | **9.68** | | JMJD1C | no peak | 6.91 | 7.04 | 12.03 | And genome-wide the DOT1L track sharpens: peaks above 5-fold go from **80 (DMSO) to 192 (EPZ5676), 197 (compound 10) and 364 (compound 11)** [E1]. **The hypothesis.** DOT1L bound to a SAM-competitive or SAM-pocket inhibitor is retained on chromatin - a trapped, catalytically dead complex - rather than cycling off. This would be mechanistically important: it predicts that a DOT1L **degrader** and a DOT1L **inhibitor** are not interchangeable, and this corpus already holds `exp-dot1l-degrader-vs-inhibitor-matched-exposure` and `mechanism-dot1l-catalysis-independent-function` as the places that question lives. **The confound that prevents calling it a result, stated first because it probably explains the data.** MACS's per-sample fragment-size estimate in this arm runs **DMSO 822 bp, compound 10 859, VTP-50469 969, EPZ5676 1064, compound 11 1112** - a 1.35-fold spread [E2]. Pileup scales with fragment length, and the two arms with the largest apparent gains (compound 11 at 1112 bp, EPZ5676 at 1064 bp) are exactly the two with the longest estimated fragments, while DMSO has the shortest. Across the five DOT1L samples, peak count correlates **+0.7 with fragment size and -0.8 with depth** [E2] - the opposite of what a depth-driven artefact looks like and consistent with fragment size driving the whole arm. **So the ordering compound 11 > EPZ5676 ~ compound 10 > DMSO is also the ordering of fragment-size estimates**, and no statistic computable from the deposited peak tables can separate the two explanations.

Go deeper
Why this is filed as a hypothesis and not dropped

A confounded observation that is discarded silently is indistinguishable from an observation that was never made. The effect is large (PBX3 nearly tripling), it is consistent across three chemically distinct inhibitors and two factors' worth of loci, and the experiment that separates it from the artefact is cheap and already partly specified in this corpus. It is recorded with its confound named so that the next person can either cash it or kill it.

Facts and figures
statementDOT1L bound to a catalytic-site inhibitor is retained or trapped on chromatin at KMT2A-fusion target genes rather than released, so DOT1L inhibitors and DOT1L degraders are not mechanistically interchangeable.
origindata-pattern
prior probability0.3
confound namedMACS per-sample fragment-size estimate spans 822-1112 bp across the five DOT1L samples and its rank order matches the rank order of apparent gain
geoGSE149183
shown for fusionKMT2A::MLLT3 in MOLM13; nothing in KMT2A::AFF1
predictions
[
  "A DOT1L degrader at matched H3K79me2 suppression removes DOT1L from target loci; a catalytic inhibitor does not.",
  "The retained DOT1L under inhibitor still co-immunoprecipitates AF9/ENL and menin, i.e. the complex is intact and only the catalysis is dead.",
  "H3K79me2 loss and DOT1L occupancy loss are dissociable in time under a catalytic inhibitor, with the mark falling while occupancy holds."
]
analyses
[
  "ana-q136-the-menin-inhibitor-evicts-dot1l-from-fusion-target-genes-while-dot1l-inhibitors-leave-it-there"
]
experiments
[
  "exp-q136-spike-in-normalised-men1-and-dot1l-chip-with-replicates-and-both-drug-classes-in-rs4-11",
  "exp-dot1l-degrader-vs-inhibitor-matched-exposure"
]
falsifiers
[
  "A spike-in-normalised DOT1L ChIP with at least two replicates per arm in MOLM13 at 4 d, with fragment-size distributions reported per sample: if DOT1L occupancy at HOXA9, MEIS1, PBX3, SENP6 and JMJD1C is within 1.3-fold of DMSO under compound 10 at 100 nM once fragment size is matched, the apparent gain was an artefact and this hypothesis is dead.",
  "If a DOT1L degrader at matched H3K79me2 suppression leaves DOT1L occupancy at those five loci unchanged relative to a catalytic inhibitor, the inhibitor-versus-degrader prediction fails and the trapping model loses its main consequence.",
  "If the fragment-size spread is shown to be an artefact of sonication batch rather than of treatment - e.g. by re-estimating fragment size from the raw paired-end data in SRP258055 and finding it uniform - then the confound named here is not the explanation and the observation returns as a candidate result."
]
How this could bite us back
  • If catalytic DOT1L inhibitors leave a dead enzyme parked on target genes, resistance could arise through a catalysis-independent scaffolding function that the drug does not touch - and pinometostat's clinical failure at achievable exposures is at least consistent with that.
  • A degrader would avoid the trapping liability but removes DOT1L from normal haematopoiesis too, where this corpus already records a requirement. The trapping hypothesis, if true, makes the degrader more attractive and more toxic at the same time.
Still open
  1. Can the fragment-size confound be removed without new wet work? Re-estimating fragment size directly from the paired-end runs in SRP258055 and re-running MACS with a fixed extension size would test it computationally, at the cost of downloading raw data this beat avoided.
  2. Is DOT1L trapping, if real, a liability or an opportunity? A catalytically dead DOT1L parked on target genes could either block the locus usefully or scaffold something else - and the corpus already holds mechanism-dot1l-catalysis-independent-function, which predicts the second.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In GSE149183's DOT1L ChIP arm, DOT1L fold-enrichment at canonical menin-MLL target loci is higher under all three DOT1L inhibitors than under DMSO (PBX3 4.85 to 8.32/9.41/13.92; SENP6 3.37 to 6.17/6.03/9.68; HOXA9 2.56 to 4.39/5.08/7.47), and peaks above 5-fold rise from 80 to 192/197/364.

    What this evidence cannot show

    n = 1 per condition, within-sample contrasts only, and see E2 - the confound is severe enough that this evidence item does not by itself support the hypothesis, only the existence of the observation.

    Evidence that argues against this (1)
    • The menin inhibitor arm, which has an intermediate fragment size (969 bp), shows the OPPOSITE - no DOT1L peak at any target locus. If fragment size drove everything, VTP-50469 at 969 bp should sit between DMSO and EPZ5676, and it does not. That argues the biology is not wholly artefact - but it does not rescue the specific gain being real. Oncofusion knowledge graph 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    MACS fragment-size estimates across the five DOT1L samples span 822-1112 bp, peak count correlates +0.7 with fragment size and -0.8 with library depth, and the rank order of apparent gain matches the rank order of fragment size.

    What this evidence cannot show

    Spearman over n = 5 samples; the correlation coefficients are descriptive and carry no useful p-value at that n. They are reported because the rank agreement is complete, not because the correlation is significant.

    Evidence that argues against this (1)
    • A drug that changes chromatin compaction genuinely changes the sonication fragment-size distribution, so the fragment-size spread may itself be a biological readout rather than a technical nuisance. That reading would make the confound inseparable in principle from the effect, not just in this deposit. Oncofusion knowledge graph 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    Compound 10 at 100 nM is target-engaging in this series - the dm6 spike-in H3K79me2 ratio in MOLM13 reaches 5.57 by day 3 - so the retained DOT1L, if real, is retained while catalytically inhibited.

    What this evidence cannot show

    Engagement is demonstrated for compound 10 at 100 nM only; EPZ5676's H3K79me2 arm used 1 microM and compound 11 has no H3K79me2 arm at all.

    Evidence that argues against this (1)
Connected to