At the 692 RS4;11 bins where an MLL1 N-terminus and menin co-occupy, VTP-50469 takes menin to 0.20 and DOT1L to 0.20, drops H3K79me2 to 0.66, and leaves MLL1-N-reactive material at 0.83 - four factors, one drug, one chromatin, and a clean ordering
also called the four-way ordering · DOT1L leaves with menin · the function arm
Measured side by side at the same places in the same cells, the drug removes menin and DOT1L almost completely, cuts the histone mark they produce by a third, and leaves the MLL1 protein itself essentially where it was. Whatever stays behind has lost its partners and most of its output.
The same thing, for a scientist
**Four antibodies, one drug, one set of bins, one within-sample statistic.** RS4;11 (KMT2A::AFF1), 330 nM VTP-50469 for 3 days, 2 kb genome-wide bins, universe 1,389,514 (input > 0, ENCODE hg19 blacklist v2 removed) [E1]: | at the 692 bins in BOTH the MLL1-N and MENIN top 0.1% | enrichment DMSO -> VTP | share ratio | floor-normalised | |---|---|---:|---:| | **MLL1-N** (Bethyl A300-086A) | 17.09x -> 11.75x | **0.8305** | **0.949** | | **MENIN** (Bethyl A300-105A) | 18.00x -> 2.36x | **0.2005** | **0.226** | | **DOT1L** (CST 77087S, clone 1W4Z) | 19.10x -> 2.46x | **0.2003** | **0.244** | | **H3K79me2** (CST 5427) | 16.14x -> 8.24x | **0.6578** | **0.668** | The floor is each track's own aggregate share ratio over ~555,700 mid-occupancy bins (DMSO percentile 50-90): MLL1-N 0.8753, MENIN 0.8874, DOT1L 0.8208, H3K79me2 0.9851. **And at the MLL1-N held set alone (1,390 bins), floor-normalised, MLL1-N-reactive material does not move at all: 0.9069/0.8753 = 1.036.** DOT1L there is 0.2455 (0.299 over floor), menin 0.2560 (0.288), H3K79me2 0.7008 (**0.711**). **This is the FUNCTION ARM falsifier of `hyp-q118-...`, run inside the beat, and it DOES NOT FIRE.** The falsifier as written was: *if H3K79me2 at the 1,390 MLL1-N held bins has a share ratio above 0.75, the retained complex is still functional and 'decoupling' understates what survives.* Observed **0.7008**, floor-normalised **0.7114**. **H3K79me2 falls where MLL1-N occupancy does not.** **It also refutes one of that hypothesis's own predictions.** Prediction 4 was *"DOT1L and SEC co-occupancy at the MLL1-N held set behaves like MLL1-N rather than like menin."* **DOT1L behaves like menin, to three decimal places at the co-occupied set (0.2003 against 0.2005).** So the retained MLL1-N-reactive material is not a functional complex minus menin - it is a remnant that has lost menin, lost DOT1L, and lost a third of its H3K79me2. **The ordering is the result.** `MENIN ~ DOT1L (0.20) < H3K79me2 (0.66) < MLL1-N (0.83-1.04)`. A histone mark should be the slowest thing to change, because it has to be diluted or actively removed rather than dissociating, and at 3 days it is intermediate - **which is what a mark decaying behind a departed writer looks like.** The thing that has not moved is the one the field measures.
Why the H3K79me2 arm existed only in the raw files
The depositors' own processed table is explicitly "Summary ROSE TSS signal output (for all samples excluding H3K79me2)". So the H3K79me2 arm of the founding VTP-50469 occupancy dataset cannot be reached from the deposited processed file at all, and Q117 listed the equivalent arm of GSE260742 as its own cheapest next test and did not run it.
It cost two downloads (GSM3635697 and GSM3635698, 299 and 314 MB), two minutes of binning and deletion. The falsifier was written into a hypothesis in this beat and scored in the same beat, which is the practice this project's rule 96 exists to encourage and which its own audit found happening in 17 of 624 cases.
Two properties of the H3K79me2 track that matter for reading it. Its zero fraction in the universe is 0.000 - it is a broad mark and every bin has signal, so no ratio here has a small-denominator problem. And its own top 0.1% has a Jaccard of only 0.0308 with the MLL1-N held set and is only 21.4% CpG islands against MLL1-N's 75.18%: H3K79me2's own domains are gene bodies and MLL1-N's summits are promoters, so these are two different compartments and the measurement above is deliberately at MLL1-N's compartment rather than at the mark's.
The 53 gained bins, measured with all four antibodies
At the 53 bins where MLL1-N-reactive material rises 1:
| share ratio at the 53 bins | enrichment DMSO -> VTP | |
|---|---|---|
| MLL1-N | 8.6712 (by construction - the set is selected on it) | 1.97x -> 15.28x |
| MENIN | 1.6872 | 2.58x -> 3.98x |
| DOT1L | 1.0452 | 1.61x -> 1.72x |
| H3K79me2 | 1.0068 | 0.64x -> 0.62x |
Menin rises with MLL1-N at these bins; DOT1L and H3K79me2 do not move. And H3K79me2's enrichment there is 0.64x the genome median - i.e. these are regions with below-average H3K79me2, which is not what a newly-active transcribed region looks like.
So whatever the 53 bins are, they are not newly transcribed and not newly DOT1L-marked. That narrows the readings: a genuine new MLL1-N/menin co-binding event that has not (yet) produced elongation, or a shared non-productive ChIP background at low-H3K79me2 regions. The second reading is more parsimonious than it was before this measurement, and it is the one an IgG control would test. GSE127507 deposits no IgG sample in any line.
| script | analyses/q118-menin-at-the-human-index-fusion/function-arm.py |
|---|---|
| status | run |
| result summary | At the 692 MLL1-N-AND-MENIN co-occupied bins: MENIN 0.2005, DOT1L 0.2003, H3K79me2 0.6578, MLL1-N 0.8305. The FUNCTION ARM falsifier of hyp-q118 does not fire (H3K79me2 0.7008 at the MLL1-N held set, threshold 0.75), and that hypothesis's prediction 4 is refuted: DOT1L behaves like menin, not like MLL1-N. |
| cell line | RS4;11 |
| fusion | KMT2A::AFF1 |
| drug | VTP-50469 330 nM, 3 days |
| genome build | hg19, verified from each of the nine files' own chromosome extents |
| bin bp | 2000 |
| n bins universe | 1389514 |
| epigenetic readout | H3K79me2 (CST 5427 lot 4) at the MLL1-N held set falls to a share ratio of 0.7008, floor-normalised 0.7114, while MLL1-N occupancy there is unchanged at 1.036 over floor |
| mark written | H3K79me2, by DOT1L, whose own occupancy at the same bins falls to 0.2003 |
| shown for fusion | KMT2A::AFF1 in human RS4;11 - and the MLL1-N and MENIN arms do not distinguish the fusion from wild-type MLL1, so the retained species is unidentified |
inputs
[ "GSM3635695/96 RS4;11 MLL1n", "GSM3635693/94 RS4;11 MENIN", "GSM3635691/92 RS4;11 DOT1L", "GSM3635697/98 RS4;11 H3K79me2-CST", "GSM3635699 RS4;11 input", "ENCODE hg19 blacklist v2", "UCSC hg19 cpgIslandExt" ]
- If what remains after the drug is an MLL1-N-containing remnant with no menin, no DOT1L and decaying H3K79me2, then the drug is doing nearly everything one would want at the level of function while leaving a scaffold in place. That scaffold is a re-assembly platform the moment drug exposure dips, which predicts steep trough-concentration dependence and fast transcriptional recovery on a missed dose.
- It also means an occupancy-based pharmacodynamic assay is measuring the one quantity that does not move. H3K79me2 at the fusion's own sites moves 0.70 and DOT1L moves 0.20 - either is a better biomarker than the MLL1 ChIP the field uses, and both are measurable in a patient sample.
- A DOT1L inhibitor added to a menin inhibitor has little left to remove at these sites: DOT1L is already at 0.20 under the menin inhibitor alone in RS4;11. The combination rationale in the index fusion is weaker than in MOLM13 or ML-2, where this beat's TSS-level DOT1L figures at HOXA9 are 0.99 and 0.76.
- H3K79me2 falls to 0.67-0.71 at 3 days while DOT1L falls to 0.20. Is the mark decaying passively behind a departed writer, or being actively removed? A time course would separate them and GSE127507 has one timepoint. If passive, the rate gives the mark's half-life in these cells, which is a number this field does not have.
- DOT1L leaves at 0.2003 and menin at 0.2005 - agreement to three decimal places at 692 bins. Is DOT1L's departure menin-dependent, or is it a consequence of the same complex disassembling? A DOT1L ChIP in a MEN1-knockout rather than drug-treated cell would separate a drug effect from a menin-loss effect.
- The 53 gained bins have BELOW-average H3K79me2 (0.64x the genome median) and gain none under drug. If they were real new fusion sites one would expect eventual elongation; if they are ChIP background one would expect exactly this. An IgG control would decide and GSE127507 has none in any line.
- AFF1, ENL and the rest of the SEC were not measured in this series at all. DOT1L leaving with menin is suggestive that the whole elongation module goes, but the SEC itself is unmeasured under a menin inhibitor in the index fusion.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
At the 692 RS4;11 2 kb bins in both the MLL1-N and MENIN top 0.1%, within-sample share ratios under 330 nM VTP-50469 are MLL1-N 0.8305, MENIN 0.2005, DOT1L 0.2003 and H3K79me2 0.6578 (floor-normalised 0.949, 0.226, 0.244, 0.668 against genome-wide mid-occupancy floors of 0.8753, 0.8874, 0.8208 and 0.9851 over ~555,700 bins each); at the MLL1-N held set alone MLL1-N is 0.9069 (1.036 over floor) and H3K79me2 is 0.7008 (0.711); and at the 53 gained bins MENIN is 1.6872, DOT1L 1.0452 and H3K79me2 1.0068 with H3K79me2 enrichment there 0.64-fold the genome median.
Q118 re-analysis of GSE127507, DEVIATION 10: the H3K79me2 and DOT1L arms of RS4;11 binned genome-wide from the deposited .tdf files; analyses/q118-menin-at-the-human-index-fusion/function-arm.json, run 2026-09-23 · Iqbal Lab oncofusion knowledge graph 2026What this evidence cannot show
NEITHER ANTIBODY IN THIS SERIES IS FUSION-SPECIFIC AND THE ANTIBODY NAME SAYS SO. MLL1n is anti-MLL1 N-terminus (Bethyl A300-086A, lots 5 or 6); the N-terminus is retained in KMT2A::AFF1, so that track is the sum of the fusion and wild-type MLL1 in unknown proportion. MENIN is Bethyl A300-105A Lot 8; menin binds the fusion, wild-type MLL1 and non-KMT2A partners. Q117 could attribute its numbers because its fusion was FLAG-tagged; nothing here can. Every occupancy statement in this beat is about MLL1-N-reactive material or menin-occupied sites, never about the fusion. No scale factor exists for any sample. The depositors state that a Drosophila spike-in was run on one batch, that fly/human ratios were 'found to be small', and that NO spike-in normalisation was applied; no number for that measurement is reported and none is deposited. Every cross-condition figure here is therefore a WITHIN-SAMPLE share or a within-sample rank, which cancel an unknown per-sample scale factor exactly, and never a fold change. n = 1 per antibody per condition per line - there is no replicate anywhere in GSE127507 - so no number here carries a spread and none is precise. The four ChIPs are separate experiments with different efficiencies, so the ORDERING is the result and the individual ratios are not comparable as amounts between antibodies. All four tracks' genome medians were asserted non-zero before any ratio, and the H3K79me2 zero fraction in the universe is 0.000. RS4;11 has no VTP input, so none of these figures is input-corrected; the one input-corrected measurement in the series is MOLM13's and the correction there is 4.6%. The H3K79me2 samples are titled 'H3K79me2-CST' and dated 181213, a different sequencing batch from the MLL1n pair (180402) and the MENIN and DOT1L pairs (180110).
Evidence that argues against this (2)
- A share ratio is blind to a uniform global loss, so 'MLL1-N does not move' is a statement about distribution and not amount. If the drug removed MLL1-N-reactive protein uniformly, this measurement would read 1.036 over floor and the ordering would be an artefact of four different uniform losses. Nothing in GSE127507 distinguishes that, and it is the largest single vulnerability of this entry. Iqbal Lab oncofusion knowledge graph 2026
- The four antibodies come from three different sequencing batches (180110, 180402, 181213). A batch effect that differed systematically between the DMSO and VTP libraries of one batch but not another would produce an apparent ordering. The blacklist artefact load across these files is 5.5-10.1% with no catastrophic single bin, which argues against a large batch artefact, but it is not excluded. Iqbal Lab oncofusion knowledge graph 2026
the FUNCTION ARM falsifier, run inside the beat: H3K79me2 at the MLL1-N held set is 0.7008 against a threshold of 0.75, so it DOES NOT FIRE and the mark falls where occupancy is retained
and it refutes that hypothesis's own prediction 4: DOT1L behaves like MENIN (0.2003 against 0.2005) and not like MLL1-N, so the retained material is a remnant and not a functional complex minus menin
a third and fourth antibody at the same 692 bins: DOT1L leaves with menin and H3K79me2 falls, while MLL1-N does not move (1.036 over its genome-wide floor)
DOT1L occupancy at the MLL1-N held set falls to 0.2455 (0.299 over floor) and to 0.2003 at the co-occupied set - indistinguishable from menin's 0.2005
menin is 0.2005 at the co-occupied set, the reference point the other three are read against
- related to →KMT2A / MLL1 — the gene that breaks
the MLL1 N-terminus track is the one that does not move, and it cannot be attributed to the fusion
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
measured in RS4;11, the index-fusion line, at 330 nM VTP-50469 for 3 days
the quantitative form of exactly this hypothesis: what stays has lost menin, lost DOT1L and lost a third of its H3K79me2
a share statistic is blind to a uniform global loss, so 'MLL1-N stays' is a claim about distribution and not amount
at the 53 gained bins menin rises 1.6872 while DOT1L (1.0452) and H3K79me2 (1.0068) do not move, and H3K79me2 there is 0.64x the genome median - so the gained bins are not newly transcribed
the H3K79me2 arm is excluded from the depositors' own processed table by their own processing note, so this measurement exists only because the raw .tdf files were re-binned
DOT1L is already at 0.20 under the menin inhibitor alone in RS4;11, which weakens the menin-plus-DOT1L combination rationale in the index fusion specifically
the four additional files verify hg19 from their own extents on the same test
H3K79me2 at the fusion's own sites (0.70) and DOT1L (0.20) both move far more than the MLL1 ChIP the field uses as a readout (1.04 over floor), so either is a better pharmacodynamic biomarker
the four-way ordering does not replicate in a second line: MENIN, DOT1L and MLL1-N reproduce within 0.04 but H3K79me2 goes from 0.6578 to 0.8667 and overtakes MLL1-N, so the ordering has three levels in MOLM13 and not four