A housekeeping normalisation defines the global change to be zero, so only redistribution is measurable
To compare two sequencing experiments you have to put them on the same scale, and the usual trick is to assume a set of ordinary genes did not change. That assumption makes it impossible to detect a change that affected everything, so 'nothing changed overall' is a property of the method, not a finding.
The same thing, for a scientist
Stated because it bounds every quantitative claim in this beat and in the corpus entries built on GSE173599. The deposit contains no spike-in. The authors scale each treated arm by a 21-gene housekeeping panel (ACTB, ACTG1, B2M, GAPDH, GUSB, HMBS, HPRT1, MRPL18, POM121C, PPIA, RPL13A, RPL32, RPS18, RPS27, SDHA, TBP, TFRC, TOMM5, TPT1, YTHDF1, YWHAZ); the panel's own **raw** H3K27ac ratio at 3 h averages **0.809 (SD 0.113)** and at 24 h **0.654 (SD 0.131)**, and both are forced to 1. Consequences: (i) a uniform genome-wide loss is invisible by construction; (ii) a large loss concentrated at a few strong sites must push everything else up, which is the likeliest reading of the 1.271 non-target menin ratio at 4 days; (iii) only **differences between strata** are interpretable, and every claim in this beat is written as one. **Trust the shape of a share; distrust its level.**
What the panel's own numbers show
The raw housekeeping H3K27ac ratio of 0.809 at 3 h is not nothing - if it is real, global H3K27ac fell by a fifth in three hours. Normalisation converts that into 'no change at housekeeping genes' by definition. The 24 h panel value of 0.654 is larger still. Neither can be resolved without an external scale.
The operational rule
When quoting a ratio from a housekeeping- or read-depth-normalised ChIP experiment, say so in the same sentence, and never convert it into a statement about the amount of protein on chromatin. Reserve absolute-occupancy language for experiments with a spike-in or an orthogonal quantitative measure.
This corpus has already demonstrated the general form, with a ground truth - read those entries first
This entry is the deposit-specific instance and is not the first statement of the principle here. Two king entries got there first and got further:
ana-the-spike-in-ruler-is-deposited-in-plain-text-and-the-global-change-is-5-to-23-foldfound a Drosophila spike-in deposited for GSE149183, in RS4;11 - a KMT2A::AFF1 line - showing a 5- to 23-fold genome-wide collapse that a share-based statistic would barely register.ana-a-share-renders-a-global-collapse-as-a-rising-gradient-into-the-bulkthen ran this project's own share statistic against that known truth and showed it does not merely miss the collapse: it reports a large change of the wrong kind, a rising gradient into the bulk of up to ten-fold, when nothing moved.
What this entry adds is narrow and specific: the normalisation in GSE173599 is a 21-gene housekeeping panel rather than a total-signal share, the panel's own raw ratios are 0.809 (3 h) and 0.654 (24 h), and those are the numbers the next reader of this deposit needs.
What this does and does not do to the menin result in this beat
It bites, and the bite is bounded. Under a housekeeping normalisation anchored on high-signal genes, a genuine target-specific loss and a global collapse with a larger loss at targets produce the same table: top strata below 1, bulk near 1. The observed MENIN pattern (0.473 at the top 1%, 0.974 in the bulk) is consistent with both. What survives either reading is that menin falls roughly twice as far at fusion-occupied promoters as it does in the bulk, and that is the quantity every claim in this beat is written as.
One diagnostic distinguishes this deposit from the GSE149183 case: there, the share statistic pushed the bulk up to 3.859, a loud artefact signature. Here the bulk sits at 0.974, which is not that signature. That is reassurance about the magnitude of any artefact, not proof of its absence.
- A pharmacodynamic endpoint defined as a normalised occupancy ratio cannot detect a drug that reduces a protein's chromatin binding everywhere, which is precisely what a competitive protein-protein interaction inhibitor might do.
- How many entries in this corpus quote a ChIP-seq ratio from a normalised dataset in absolute-occupancy language?
- Would a spike-in arm in this system change the sign of any conclusion, or only the levels?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
The deposit's normalisation panel is 21 housekeeping genes whose own raw treated/control ratios average 0.809 at 3 h and 0.654 at 24 h and are forced to 1.
GSE173599_TSS_gene_body_signal_Figure3.xlsx - per-gene housekeeping-normalised TSS and gene-body ChIP signal, DMSO vs dTAG-VHL 3 h and 24 h, 16 channels · Olsen SN et al. · NCBI GEO supplementary file 2022What this evidence cannot show
Read directly from the deposited normalisation sheet; the panel composition and the raw ratios are the authors' own.
Evidence that argues against this (1)
- Housekeeping-gene normalisation is standard where no spike-in exists, and the alternative - total-read normalisation - makes the same assumption less explicitly, so this is a limitation of the field's practice rather than an error by these authors. Olsen SN et al. 2022
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
No spike-in chromatin or exogenous reference is described anywhere in GSE173599's sample metadata or the paper's methods.
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
An absence established by reading the deposited MACS command lines and the sample descriptions; a spike-in used but not described would not be visible this way.
Evidence that argues against this (1)
- An absence established from deposited metadata is bounded by what the submitters chose to describe; a spike-in used in the wet lab and not reported would be invisible to this check. Olsen SN et al. 2022
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
At 4 days, menin signal at non-fusion promoters reads 1.271, a 27% rise that is exactly what a renormalised large loss elsewhere would produce.
GSE173599_TSS_signal_Figure4.xlsx - per-gene HA, DOT1L and MENIN TSS signal after 4 days of EPZ-5676, VTP-50469 or dTAG-VHL (batch 1912x2) · Olsen SN et al. · NCBI GEO supplementary file 2022What this evidence cannot show
Consistent with, but not proof of, a normalisation artefact; genuine redistribution would look identical here.
Evidence that argues against this (1)
- Menin is an abundant nuclear protein with many non-MLL chromatin sites, so a genuine relocation to those sites after the fusion is removed is biologically plausible and would produce the same number. Yokoyama A et al. 2008
bounds every ratio in the beat: only between-stratum differences are interpretable, never levels
- related to →Four days after the fusion is destroyed, menin signal at non-fusion promoters is 27% higher than control
the 27% rise at non-fusion promoters is exactly what renormalising a large concentrated loss produces
second reason a level from these tables is not an absolute quantity
deposit-specific instance: GSE173599 uses a 21-gene housekeeping panel whose own raw ratios are 0.809 (3 h) and 0.654 (24 h)
that entry demonstrated the general form against a spike-in ground truth; this one records the normalisation actually used in GSE173599