Oncofusion

start / graph / ana-q141-the-menin-gap-survives-signal-matching-but-only-two-deciles-of-the-two-strata-overlap

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q141-menin-at-fusion-targets

The menin gap survives matching on menin's own signal, but only two deciles of the two strata overlap

Genes the fusion sits on are also the busiest genes, so a difference between them and the rest can be an artefact of how busy they are rather than of the fusion. Correcting for that shrinks the menin effect by about two-fifths but does not remove it - and the honest caveat is that the two groups barely overlap, so the correction rests on a narrow slice of genes.

The same thing, for a scientist

Pre-registered control 2. Repeating the occupied-vs-non-occupied comparison **within deciles of each channel's own DMSO signal** shrinks the MEN1 gap from 0.467 to **0.285** (a 39% reduction). It rests on only **2 of 10 deciles**, because occupied and non-occupied genes barely co-occur in menin signal: in decile 9 (185 occupied, 41 non-occupied) the medians are 0.514 vs 0.793, gap 0.279; in decile 8 (6 occupied, 45 non-occupied) 0.350 vs 0.820, gap 0.470. DOT1L's gap is unchanged by matching (0.455 -> 0.470, 5 deciles), CDK9's falls 0.765 -> 0.425, AFF4's rises 0.173 -> 0.304, Pol II TSS 0.105 -> 0.209, H3K27ac 0.152 -> 0.110, H3K79me2 3 h 0.116 -> 0.054. **The two negative controls come out flat or negative** (H3K4me3 0.073 -> -0.026; wild-type MLL1 C-terminal -0.079 -> -0.055), which is what makes the surviving MENIN gap interpretable rather than generic.

Go deeper
Why this control was pre-registered

A self-stratified monotone gradient is free: split any population on a variable correlated with signal strength and ratios will drift across the split for floor and regression-to-the-mean reasons alone. The pre-registration declared that if MENIN's gap vanished under signal-matching while DOT1L's survived, the MENIN gap was a baseline artefact, and that if every channel's gap vanished the whole analysis was uninformative 1. Neither happened.

The part that weakens the result, stated plainly

Two deciles is a thin base. It is thin for a structural reason - fusion-occupied genes are nearly all in the top menin-signal decile - and no re-weighting fixes a region of no overlap. The 0.285 matched gap should be read as an estimate from 185 occupied and 41 non-occupied genes, not from 6,036. A second weakening: even non-occupied genes in the top menin decile lose 21% of signal (0.793), against 2.6% for the unmatched non-occupied stratum, so part of what the unmatched contrast attributes to fusion occupancy is attributable to signal strength.

Scale and limits

No spike-in exists anywhere in GSE173599, so every number is a within-sample share, and the treated arms are scaled by a 21-gene housekeeping panel whose own raw 3 h H3K27ac ratio averages 0.809 and is forced to 1. A housekeeping normalisation DEFINES the global change to be zero, so only the DIFFERENCE between strata is measurable; a uniform genome-wide loss of any of these proteins would be invisible here. n = 1 per arm for every channel except HA, which has two independent 3 h batches. The cells are HCB1, transformed by KMT2A::MLLT3 (MLL::AF9) - NOT the index fusion of this project, KMT2A::AFF1 - and a prior beat in this corpus found that per-gene ordering transfers between MLLT3 and AFF1 lines while magnitude does not.

Facts and figures
scriptanalyses/menin-retention-by-ha-occupancy/analyze_controls.py
statusrun
result summaryMEN1 occupied-vs-non-occupied gap 0.467 raw -> 0.285 matched on menin's own DMSO signal decile, on 2 of 10 deciles. DOT1L 0.455 -> 0.470 (5 deciles). Negative controls: H3K4me3 0.073 -> -0.026; MLL1 C-terminal -0.079 -> -0.055.
peak width ratio noteSignal instrument; MEN1 companion peak width ratio 0.93, DOT1L 1.13, both below the 1.3 trigger.
inputs
[
  "GSE173599_TSS_gene_body_signal_Figure3.xlsx",
  "GSE173599_ChIPseq_TSS_Figure3_revisions.xlsx",
  "GSE173599_TSS_signal_HA_ChIPs.xlsx"
]
How this could bite us back
  • If part of the menin loss at fusion targets is a property of highly transcribed genes rather than of the fusion, then a menin inhibitor's on-target chromatin effect is smaller than the raw contrast suggests, and the dose needed to produce a given occupancy change at target genes is correspondingly higher.
Still open
  1. Does the menin gap survive matching in a dataset where fusion-occupied and non-occupied genes overlap in menin signal - for example a line with lower fusion expression, or a titrated degron?
  2. Is the 21% menin loss at high-signal NON-occupied genes real, or a consequence of the housekeeping normalisation being anchored on genes that are themselves high-signal?
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Matching the occupied-vs-non-occupied comparison on each channel's own DMSO signal decile reduces the MEN1 gap from 0.467 to 0.285 and leaves the H3K4me3 and wild-type MLL1 negative controls at or below zero.

    What this evidence cannot show

    The MENIN matched estimate uses 2 of 10 deciles (185 occupied vs 41 non-occupied genes in decile 9) because the strata barely overlap in menin signal. n=1 per arm; no replicate-based error is available for any non-HA channel.

    Evidence that argues against this (1)
    • Menin is an adapter whose chromatin residence is partly determined by LEDGF/PSIP1 rather than by the MLL protein it is bound to, which supplies a mechanism by which menin could fall at busy genes independently of the fusion. Yokoyama A et al. 2008
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Wild-type MLL1 measured with a C-terminal antibody does not preferentially leave fusion-occupied promoters (matched gap -0.055), which is the internal negative control this comparison needs.

    What this evidence cannot show

    The MLL1 C-terminal sheet holds only 1,249 genes and only 34.5% of the authors' target genes survive into it, so the occupied arm of this control is 69 genes. The authors make the same point from the same data (Figure S4F).

    Evidence that argues against this (1)
    • If the C-terminal antibody has poor efficiency, a flat result is what a failed ChIP also looks like; the sheet's small size (1,249 genes) is consistent with either reading. Olsen SN et al. 2022
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    The stratification variable is the fusion's own HA signal, whose positive control separates the strata 61.7-fold.

    What this evidence cannot show

    See the companion positive-control entry; without that gate this control would be uninterpretable.

    Evidence that argues against this (1)
    • A stratification that works perfectly for the fusion need not be the right stratification for menin, whose chromatin distribution is broader than the fusion's and includes many non-MLL sites. Yokoyama A et al. 2008
Connected to