Oncofusion

start / graph / ana-q118-in-ml-2-the-dot1l-track-moves-most-and-the-menin-eviction-is-weakest

Analysisweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q118-menin-at-the-human-index-fusion

ML-2 (KMT2A::AFDN, a non-elongation partner) is the line where menin eviction is weakest (retains 37.2% of its own top windows against 20.5% and 13.0%) and where the DOT1L track moves most of any track in the beat (retains 31.6%, 13.69 percentile points below a rank-matched null at p = 3e-10) - the opposite pattern to what an AF9/ENL-free fusion would predict

also called ML-2, the third answer · the MLL-AF6 line · DOT1L moves most where the fusion cannot recruit it

The third cell line carries a fusion partner that is not part of the transcription-elongation machinery, so DOT1L should have been the least affected thing in it. It is the most affected, and menin is the least - which is the reverse of the expected pattern and worth checking before it is believed.

The same thing, for a scientist

**Three lines, three answers - the reason they are three entries and not one.** ML-2 carries **KMT2A::AFDN (MLL-AF6)** per `!Series_overall_design`, a partner with no AF9/ENL-family DOT1L-recruiting moiety, and it behaves unlike the other two in both directions [E1]. At TSS windows, the fraction of each track's own top 1% still in its own top 1% under 330 nM VTP-50469: | line | fusion | MENIN | MLL1-N | DOT1L | |---|---|---:|---:|---:| | RS4;11 | KMT2A::AFF1 | 0.205 | 0.549 | 0.712 | | MOLM13 | KMT2A::MLLT3 | **0.130** | 0.665 | 0.721 | | **ML-2** | **KMT2A::AFDN** | **0.372** | 0.637 | **0.316** | And the within-sample share ratios at each track's own top 1%: ML-2 MENIN **0.3517**, MLL1-N **0.6463**, DOT1L **0.2859**. **Two things are anomalous and they point in opposite directions.** **ML-2's menin eviction is the weakest of the three** (0.372 retained against 0.205 and 0.130), and its unnormalised MENIN total falls least (0.658 against 0.296 and 0.128). If menin engagement is what the drug does, ML-2 is the line where it does it least - and ML-2 is also the line most often described as the menin-inhibitor-insensitive comparator in this literature. **ML-2's DOT1L track moves most of any track in the whole beat.** Its DOT1L held set falls 4.91 percentile points against a rank-neighbourhood null's 18.60 - **13.69 points, Mann-Whitney p = 3.2e-10**, the largest and most significant rank effect measured anywhere in this analysis - and it produces **91 gained TSS windows**, more than any MLL1-N arm in any line. **That is the reverse of the naive prediction**, which is that a fusion without an AF9/ENL moiety recruits DOT1L least and so has least DOT1L to lose. **This entry reports the anomaly and does not explain it.** Three candidate readings, none tested here: (i) DOT1L in ML-2 is recruited by wild-type MLL1 or by another complex whose menin dependence is greater than the fusion's; (ii) the ML-2 DOT1L libraries are of lower quality - ML-2 DOT1L DMSO has the smallest total in the deposited table (3,140,979) and the ML-2 DOT1L VTP the smallest of all (1,477,570), and a 91-window gained set from a weak library is what tail noise looks like; (iii) ML-2's fusion is not what the series says it is. **Reading (ii) is the one to exclude first** and it is cheap: the ML-2 DOT1L pair was not downloaded, so no genome-wide, blacklisted, input-filtered version of this measurement exists. **Until it does, the ML-2 DOT1L result should be treated as a flag and not a finding.**

Facts and figures
scriptanalyses/q118-menin-at-the-human-index-fusion/roseA.py and roseB.py
statusrun
result summaryML-2 retains 37.2% of its menin top windows (weakest eviction of the three lines) and 31.6% of its DOT1L top windows (strongest effect on any track in the beat), with 91 gained DOT1L windows from the weakest library in the table.
cell lineML-2
fusionKMT2A::AFDN (MLL-AF6), per !Series_overall_design
drugVTP-50469 330 nM, 3 days
genome buildhg19
shown for fusionKMT2A::AFDN in human ML-2, via MENIN, MLL1-N and DOT1L antibodies, none of which is fusion-specific
absence classmeasured-not-analysed
absence class noteThe gap this token describes is that the ML-2 .tdf files ARE deposited and were not downloaded, so no genome-wide blacklisted input-filtered version of the ML-2 measurements exists. It does not describe the missing ML-2 VTP input, which is never-measured for this deposit.
inputs
[
  "GSE127507 ROSE TSS signal table, ML2 columns"
]
How this could bite us back
  • If a menin inhibitor has its largest chromatin effect on DOT1L in the line whose fusion partner cannot recruit DOT1L, then the drug's DOT1L effect is not fusion-mediated, and the mechanistic argument for combining a menin inhibitor with a DOT1L inhibitor in that fusion class weakens.
  • ML-2 is used as the insensitive comparator in this field. If its menin eviction is genuinely weaker rather than its ChIP being worse, then 'insensitive' may mean 'the drug engages less' rather than 'the cells tolerate engagement' - which are different resistance mechanisms with different countermeasures.
Still open
  1. Does the ML-2 DOT1L result survive a genome-wide, blacklisted, input-filtered re-analysis? The four ML-2 .tdf files needed (GSM3635674/75 DOT1L, GSM3635680 input) are deposited and were not downloaded. Until that is run, this is a flag rather than a finding.
  2. Why is menin eviction weakest in ML-2? A less efficient menin ChIP and a genuinely less menin-dependent fusion give the same within-sample signature, and the deposited data cannot separate them.
  3. Is ML-2's fusion really KMT2A::AFDN? The identity is from the series design text only. Cellosaurus and an RNA-seq junction check would confirm it and neither was run.
Every source behind this entry (1)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    At TSS windows under 330 nM VTP-50469, ML-2 (KMT2A::AFDN) retains 37.2% of its own top 1% menin windows against 20.5% in RS4;11 and 13.0% in MOLM13, and retains only 31.6% of its own top 1% DOT1L windows with the held set falling 13.69 percentile points below a rank-neighbourhood null (Mann-Whitney p = 3.2e-10) and 91 gained windows; ML-2's MENIN, MLL1-N and DOT1L share ratios are 0.3517, 0.6463 and 0.2859.

    What 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. ML-2's fusion identity (MLL-AF6 / KMT2A::AFDN) is taken from !Series_overall_design and was not independently verified from a karyotype or RNA-seq. The ML-2 DOT1L libraries have the smallest totals in the deposited table (3,140,979 DMSO and 1,477,570 VTP against 34,279,592 for MOLM13 DOT1L DMSO), so a low-library-depth explanation for the 91-window gained set is live and has NOT been excluded. ML-2 has no VTP input, so no differential-input correction is possible. No ML-2 sample was downloaded for the genome-wide instrument, so this entry rests entirely on the deposited TSS table with no blacklist or input filtering.

    Evidence that argues against this (2)
    • The ML-2 DOT1L VTP column has the smallest total signal of all eighteen columns. A gained set derived from the tail of the weakest library in a table is the classic shape of a depth artefact, and this entry's own reading (ii) is that explanation. Nothing here distinguishes it from a biological result and the genome-wide re-analysis that would is not done. Iqbal Lab oncofusion knowledge graph 2026
    • ML-2 is widely used as the relatively menin-inhibitor-INSENSITIVE comparator in this field, and a weaker menin eviction in it is consistent with that. But it is also consistent with a less efficient menin ChIP in that line, and the two cannot be separated by a within-sample statistic. Krivtsov AV et al. 2019
Connected to