H3K4me3 is by far the most reproducible track in primary infant leukaemia CUT&Tag - replicate rho 0.97-0.98 - which makes it both the assay's proof of competence and the benchmark that every fusion-site enrichment should be quoted against
A general marker of switched-on genes is measured almost identically twice in the same patient sample, which shows the technique works on this difficult material. It also sets the bar that any claim about the fusion has to beat.
The same thing, for a scientist
**Same-antibody replicate agreement in the three primary KMT2A::AFF1 samples, Spearman over 19,404 promoter shares** [E1]: | target | p279 | p318 | p454 | |---|---:|---:|---:| | **H3K4me3** | **0.977** | **0.972** | **0.979** | | KMT2A-N (two antibodies) | 0.905 | 0.846 | 0.903 | | ENL | 0.629 | 0.735 | 0.785 | | AFF1 (two antibodies) | 0.712 | 0.765 | 0.784 | | KMT2A-C (two antibodies) | 0.730 | 0.620 | 0.692 | | DOT1L | **0.357** | 0.540 | 0.629 | **Two things follow.** **One: the assay is not the limiting factor.** AutoCUT&Tag on uncultured infant marrow reaches rho 0.98 between replicates. So a factor track at rho 0.36-0.79 is reporting something about that factor's abundance or its reagent, not a ceiling imposed by primary material. This removes the most common excuse for weak primary-sample chromatin data. **Two: H3K4me3 is the correct noise benchmark for a fusion-site enrichment claim, and it is a demanding one.** Its own enrichment at N/C-defined fusion promoters over wild-type KMT2A promoters is 0.97-1.06 - flat, exactly as the source paper's use of it as a negative control requires [E2]. But its **top 971 promoters** land in the fusion-target set at **2.1x, 2.5x and 2.6x the random null**, because fusion targets are active promoters and H3K4me3 marks active promoters. **A factor that hits fusion targets at 2x the null has achieved nothing that H3K4me3 does not achieve without any fusion specificity at all.** By that benchmark ENL (3.5-5.5x) passes, DOT1L (2.2-3.5x) passes in one sample of three, and AFF1 (1.8-4.6x with its better antibody) passes in one of three. **The general form.** A ratio-style negative control and an overlap-style negative control give different answers for the same track, and both are needed. H3K4me3's ratio says 'flat, no fusion preference'; H3K4me3's overlap says 'half the apparent fusion-site enrichment of any factor is promoter activity'. Quoting only the first makes every factor look better than it is.
Why the DOT1L number is the one to worry about
p279 DOT1L replicates agree at rho 0.357 - below this beat's pre-registered 0.5 guard and below the same sample's AFF1 two-ANTIBODY agreement of 0.712. Two aliquots of one antibody in one sample disagreeing more than two different antibodies do is a signature of a failed library rather than a biological property, and it is why p279 DOT1L is reported but not relied on anywhere in this beat.
| script | analyses/q142-aff1-in-primary-infants/extra.py |
|---|---|
| status | run |
| result summary | H3K4me3 replicate rho 0.972-0.979; its top promoters hit fusion targets at 2.1-2.6x null despite a flat 0.97-1.06 ratio. |
- Should this corpus adopt a standing rule that a fusion-site enrichment is quoted against an active-promoter-mark benchmark measured in the same samples?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In GSE252378 primary infant samples, H3K4me3 replicate Spearman is 0.972-0.979 while DOT1L replicates reach only 0.357-0.629, and H3K4me3's top 971 promoters overlap the N/C-defined fusion set at 2.1-2.6x the random null despite a flat 0.97-1.06 enrichment ratio.
GSE252378 - AutoCUT&RUN and AutoCUT&Tag profiles of 36 MLL-rearranged leukemia samples (171 GSMs, hg38) · Janssens DH, Henikoff S (Fred Hutchinson Cancer Center / Van Andel Institute) · NCBI Gene Expression Omnibus 2024What this evidence cannot show
Promoter windows, within-sample shares, no spike-in. H3K4me3 is a far more abundant target than any factor here, so its reproducibility advantage is partly a signal-to-noise effect and the comparison is not like-for-like; that is precisely why it is used as a ceiling and a benchmark rather than as a peer.
Evidence that argues against this (1)
- Using an abundant histone mark to set the bar for a low-abundance transcription factor can be argued to be unfair in both directions - too easy on the ratio test, too hard on the overlap test - and no third benchmark exists in this deposit to adjudicate. A synthetic null built by shuffling promoter labels would be a cleaner benchmark and was not built. Janssens DH et al. 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The source paper uses H3K4me3 as a negative control that does not distinguish wild-type from oncoprotein-target sites.
MLL oncoprotein levels influence leukemia lineage identities · Janssens DH, Duran M, Otto DJ, Wu W, Xu Y, Kirkey D, Mullighan CG, Yi JS, Meshinchi S, Sarthy JF, Ahmad K, Henikoff S · Nature Communications 15:9341 2024What this evidence cannot show
Read from PMC11522475.
Evidence that argues against this (1)
- H3K4me3 is catalysed by wild-type KMT2A and is a downstream consequence of fusion activity, so calling it a negative control is a modelling choice; if the fusion raises H3K4me3 at its targets the control is partially positive and the bar it sets is too high. Janssens DH et al. 2024
- related to →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here
H3K4me3 as both competence proof and benchmark
supplies the benchmark that entry scores against