Summit containment passes the positive control that the bin-overlap measure failed, in all three HA batches
An earlier check of this dataset used a method that claimed the destroyed cancer protein was still sitting on most of its sites - which cannot be true. Switching to a stricter method gives 2-9% instead of 40-93%, which is what a destroyed protein should look like.
The same thing, for a scientist
Direct replication and extension of the corpus's own measurement-trap finding. Using **summit containment** (a control peak is retained only if a treated-arm peak summit falls inside it), the HA self-control gives **1.9%** retention in batch 1909x2 (12,017 control peaks), **6.4%** in 190814 (5,049 peaks) and **9.1%** in 1912x2 at 4 d (3,204 peaks). The corresponding treated/control median peak width ratios are **2.79, 1.93 and 2.18** - all above the pre-registered 1.3 re-score trigger, and all reproducing the corpus's recorded 1.92x width inflation (the 190814 batch matches it to two decimal places). The prior bin-overlap measure reported 40-93% for the same files. The 4.5-point spread between the two 3 h batches (1.9% vs 6.4%) is the **only** replicate-derived error estimate anywhere in this deposit, and it is an estimate on a near-total collapse: it does not bound a 20% effect on a channel that barely moves.
Top-decile figures, for comparability with the existing corpus entries
| channel | prior corpus figure | recomputed here | delta |
|---|---|---|---|
| MEN1 top-decile summit retention | 80.0% | 83.4% | +3.4 |
| DOT1L top-decile summit retention | 68.7% | 68.8% | +0.1 |
| HA peak width ratio | 1.92x | 1.93x (batch 190814) | +0.01 |
The prior figures reproduce. The 3.4-point MEN1 difference is a top-decile definition difference (I rank by -log10 q). Nothing in the prior entries' arithmetic is wrong. What this beat changes is the inference drawn beside those numbers.
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.
| script | analyses/menin-retention-by-ha-occupancy/analyze_peaks.py |
|---|---|
| status | run |
| result summary | HA summit-containment self-control 1.9% / 6.4% / 9.1% across three batches, with width ratios 2.79 / 1.93 / 2.18. Prior corpus figures reproduce: MEN1 top-decile 83.4% vs 80.0%, DOT1L 68.8% vs 68.7%, width 1.93x vs 1.92x. |
| peak width ratio note | All three HA batches exceed the 1.3 trigger (2.79, 1.93, 2.18), which is why the HA peak figures are reported only as a positive control and never as a magnitude. |
inputs
[ "GSE173599 MACS peak tables" ]
- Any occupancy-based pharmacodynamic endpoint for a menin or DOT1L inhibitor needs its peak-width behaviour reported alongside it, or a 2x width inflation on a collapsing channel will be read as retained binding.
- Is summit containment itself biased when treated peaks are much wider - a wide treated peak has a summit that may fall outside a narrow control peak even when signal is retained? The direction of that bias is toward under-reporting retention, so it is conservative here, but it is not quantified.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Summit-containment retention of HA control peaks after fusion degradation is 1.9%, 6.4% and 9.1% in the three HA batches, against a bin-overlap measure that reported 40-93%.
GSE173599 - Chromatin state changes induced by targeted degradation of MLL-AF9 are phenocopied by combined DOT1L and MENIN inhibition [ChIP-Seq] · Olsen SN, Hatton C (Armstrong lab, Dana-Farber) · NCBI GEO (SubSeries of GSE173600) 2022What this evidence cannot show
Peak calling depends on library depth, which is not equalised between arms and has no spike-in control. The 1912x2 figure is a 4 d rather than 3 h arm.
Evidence that argues against this (1)
- A low summit-containment figure could in principle reflect summit position instability rather than protein loss; the paper's own western blots and the near-zero per-gene HA signal ratios make protein loss the parsimonious reading. Olsen SN et al. 2022
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Treated/control median peak width ratios for HA are 2.79, 1.93 and 2.18, reproducing and exceeding the 1.92x inflation already recorded in this corpus.
GSE173599 - Chromatin state changes induced by targeted degradation of MLL-AF9 are phenocopied by combined DOT1L and MENIN inhibition [ChIP-Seq] · Olsen SN, Hatton C (Armstrong lab, Dana-Farber) · NCBI GEO (SubSeries of GSE173600) 2022What this evidence cannot show
Width inflation on a collapsed channel is expected from MACS's lambda model; the ratio is reported here as a diagnostic, not as a biological quantity.
Evidence that argues against this (1)
- A width ratio computed on a channel with 997 surviving treated peaks is estimated from few observations, and the between-batch spread (1.93 to 2.79) is itself large enough to question any single value. Olsen SN et al. 2022
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
The two independent HA 3 h batches give 1.9% and 6.4%, a 4.5-point spread that is the only replicate-derived error estimate in the deposit.
GSE173599_TSS_signal_HA_ChIPs.xlsx - per-gene HA (MLL::AF9) TSS signal in four sequencing batches, 19,168 genes each · Olsen SN et al. · NCBI GEO supplementary file 2022What this evidence cannot show
This is a technical floor measured on a channel that lost ~98% of its signal. It does NOT bound the error on a 20% effect in a channel that barely moves, and must not be quoted as if it did.
Evidence that argues against this (1)
- The two HA 3 h batches differ in sequencing depth and in the input library used, so their 4.5-point spread mixes technical replication with batch effects and may overstate reproducibility rather than understate it. Olsen SN et al. 2022
replicated in all three HA batches: summit containment 1.9/6.4/9.1% with width ratios 2.79/1.93/2.18
- related to →Peak-call retention tracks peak strength, not fusion occupancy, and that inverts the answer
summit containment fixes the width trap but not the dynamic-range trap
- related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
the degraded fusion retains 1.9-9.1% of its own peak calls