Oncofusion

start / graph / ana-q118-menins-own-gained-set-is-cpg-island-enriched-and-the-mll1-n-one-is-not

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q118-menin-at-the-human-index-fusion

Under VTP-50469 in RS4;11, the 209 bins that rise to the top of the MENIN distribution are 3.7-fold CpG-island-enriched over a class-matched null at p = 5.2e-07, while the MLL1-N gained set is not enriched at all - so menin's residue is island-biased and an MLL1-N gain is not

also called menin's residual chromatin · the positive control for criterion C · menin redistribution

When the drug strips menin off its usual places, the places where menin is still most concentrated afterwards are unusually rich in CpG islands, unlike the equivalent set for the MLL1 antibody. The simplest explanation is not that menin moves, but that the menin that survives the drug was never the kind that needed MLL1.

The same thing, for a scientist

**Same cells, same universe, same 2 kb bins, same class-matched matcher, same test - and the opposite answer from the MLL1-N arm** [E1]: | | MENIN arm | MLL1-N arm | |---|---|---| | bins rising into the top 0.1% from below the 90th DMSO percentile | **209** | 53 | | on an hg19 CpG island | **21/209 = 10.05%** | 3/53 = 5.66% | | class-matched null rate | 2.71% | 4.23% | | **fold over matched null** | **3.71** | 1.34 | | **Fisher two-sided p** | **5.2e-07** | 0.49 | | amplitude in the VTP track, median | 5.06x the genome median | 15.28x | | the held set's own amplitude in VTP | **1.96x** | 12.18x | **Two things follow, and only the second is new.** **First, the instrument works.** A CpG-island enrichment test that returns p = 0.49 for MLL1-N is worth nothing unless the same test can return a positive; it returns p = 5.2e-07 for menin. This is the positive control that licenses the MLL1-N negative, and it was run in the same script. **Second, and this is the finding: under drug, menin's top bins are no longer menin's real sites.** The held set's amplitude in the VTP track is **1.96x** the genome median while the gained set's is **5.06x** - so the 209 gained bins outrank the 1,390 held bins in the treated sample. That is what near-total eviction looks like from the inside: the distribution's top becomes whatever did not leave. **The parsimonious reading is not relocation.** Menin has substantial MLL1-independent chromatin function - JUND, SMAD3, and in this beat's own data 698 menin-held bins that carry no MLL1-N signal and are only **31.4%** CpG islands against 80.8% for the MLL1n-only bins. A menin inhibitor blocks the MLL1-binding pocket, so menin's MLL1-independent occupancy should be the part that survives. **The 3.71-fold island enrichment of the survivors is then a statement about which menin sites are drug-resistant, not about menin moving anywhere.** Distinguishing the two needs a time course, and GSE127507 has one condition and one timepoint per line.

Go deeper
The TSS-window version, and the three-line pattern

On instrument A (the deposited ROSE table, top 1% = 215 windows), the MENIN arm produces gained sets of 75 windows in RS4;11, 110 in MOLM13 and 34 in ML-2, with gained amplitudes of 6.78x, 4.31x and 4.72x the track's non-zero median against held amplitudes of 3.20x, 1.21x and 2.88x 2 - the same inversion.

And their gene identity is an artefact signature, not a programme. The RS4;11 menin-gained windows are led by LOC401074, LOC284412, MIR4461, SNAR-C2, LOC100134317, MIR519B, FAM90A10P, FAM90A7P, MTRNR2L2, TRIM48, PRR20A/PRR20E, GOLGA6L10, USP17L2, FAM86B1 and FAM182B; the MOLM13 set by OR2T1, SNAR-C2, FRG1B, FAM86B1, SNX29P2, PPIAL4A, TBC1D3H, SUMO1P3, BTF3P11 and UPK3BL. Those are microRNAs, snaR genes, pseudogenes and segmental-duplication families - the classic multi-mapping compartment.

So the TSS-window menin-gained sets and the genome-wide one are telling different stories, and the honest summary is: at TSS windows menin's survivors look like mapping artefact; genome-wide they are island-biased. The 2 kb genome-wide analysis is blacklisted and input-filtered and the TSS table is neither, which is the likeliest reason.

Facts and figures
scriptanalyses/q118-menin-at-the-human-index-fusion/binsB.py and roseB.py
statusrun
result summaryMENIN gained set 209 bins, CpG island 10.05% vs matched null 2.71%, fold 3.71, Fisher p = 5.2e-07; gained amplitude 5.06x vs held 1.96x in the treated track. The MLL1-N arm in the same cells gives fold 1.34 at p = 0.49.
n gained menin genomewide209
genome buildhg19
bin bp2000
epigenetic readouthg19 CpG-island overlap of the menin survivors: 10.05% vs a class-matched null of 2.71%, against menin's own held set at 50.36%
cell lineRS4;11
shown for fusionKMT2A::AFF1 in human RS4;11 for the menin arm; this entry is about MENIN occupancy and makes no claim about the fusion
inputs
[
  "GSM3635693/94 RS4;11 MENIN",
  "GSM3635699 RS4;11 input",
  "GSE127507 ROSE TSS table",
  "UCSC hg19 cpgIslandExt"
]
How this could bite us back
  • If about a tenth of menin's occupancy is drug-resistant and island-biased, a menin inhibitor is not a menin-function-off switch, and any toxicity or efficacy attributed to complete menin loss is being attributed to something that does not happen at 330 nM for 3 days.
  • The menin sites that survive are the ones least likely to be MLL1-dependent, which means they are the ones a more potent or longer-dosed inhibitor is least likely to reach. Escalation would not close this residue.
  • An occupancy-based pharmacodynamic assay reading menin ChIP amplitude would see the drug work and would also see the top of the distribution move to a different, island-biased set of sites. Reported as 'menin redistributes', that would be a real observation with a misleading name.
Still open
  1. Are the 209 menin survivor bins NEW sites or PRE-EXISTING weak ones? They are below the 90th DMSO percentile by construction, so they were weakly occupied already; a time course, or a comparison against a MEN1-knockout menin ChIP, would separate new binding from surviving binding. GSE127507 has one timepoint per line.
  2. Do the survivors overlap the published MLL1-independent menin cistrome - JUND, SMAD3, the menin sites in endocrine tissue? No such overlap was computed here and the datasets exist.
  3. MOLM13 has both inputs and would allow the only differential-input-controlled version of this analysis. Its MENIN pair (GSM3635683/84) was not downloaded, and it is a 4-day exposure rather than 3, so the comparison would not be exposure-matched to RS4;11.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    In RS4;11 2 kb bins, 21 of the 209 bins that rise into the MENIN top 0.1% from below the 90th DMSO percentile overlap an hg19 CpG island (10.05%) against a class-matched null of 2.71% - fold 3.71, Fisher two-sided p = 5.2e-07 - and their median VTP amplitude is 5.06-fold the genome median against 1.96-fold for the 1,390 menin-held bins.

    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. The set is defined by selecting the top of the treated distribution, so its amplitude in the treated sample is large by construction; what is not by construction is the CpG-island contrast against a null matched on DMSO-MENIN decile and input decile. There is no VTP input for RS4;11, so this comparison has no differential-input control; the equivalent MENIN samples for MOLM13, which does have both inputs, were NOT downloaded. A 2 kb bin is wider than most islands.

    Evidence that argues against this (1)
    • The gained set's island rate of 10.05% is far BELOW menin's own held set at 50.36%, so calling this 'menin moving onto islands' overstates it. It is equally consistent with the surviving menin being a pre-existing weak, partly-island subpopulation that was always there and simply became the top of the distribution once the strong sites were stripped. Nothing here distinguishes a new site from a surviving one. Iqbal Lab oncofusion knowledge graph 2026
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    At TSS windows the MENIN gained sets are 75 (RS4;11), 110 (MOLM13) and 34 (ML-2) windows with amplitudes above those of their own held sets, and their gene identity is dominated by microRNAs, snaR genes, pseudogenes and segmental-duplication families (FAM90A10P, FAM90A7P, FAM86B1, FRG1B, PPIAL4A, TBC1D3H, SUMO1P3, BTF3P11, GOLGA6L10, USP17L2, MTRNR2L2, SNAR-C2).

    What this evidence cannot show

    The ROSE TSS table is neither blacklist-filtered nor input-thresholded by this beat, which is the likeliest reason its gained sets look like mapping artefact while the blacklisted genome-wide analysis does not. Gene symbols are the deposited table's own RefGene annotation and were not resolved through HGNC. 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.

    Evidence that argues against this (1)
    • The same families appear in the MOLM13 and RS4;11 lists, which could be read as a reproducible biological result rather than a shared artefact. The reading here is artefact because these are the known multi-mapping families and because the genome-wide, blacklisted analysis of the same line does not produce them - but a repeat-aware re-analysis would settle it and was not run. Boyle-Lab/Blacklist (GitHub) 2019
Connected to