For AFF1 in primary infant material, which antibody you use moves the answer more than which patient, which lineage, or cell line versus patient - a pre-registered falsifier that fired
Two different antibodies aimed at the same protein in the same patient sample gave more different answers than two different patients gave with the same antibody. That means cross-sample comparisons of this protein are partly measuring the reagent.
The same thing, for a scientist
**Pre-registered falsifier F3:** *if the p279-to-p318 AFF1 Spearman is higher than the p279 AFF1-1-to-AFF1-2 antibody agreement, then between-lineage variation is smaller than between-reagent variation, and every cross-sample claim in this beat is a reagent claim.* **It fired, in every comparison available** [E1]. | comparison | Spearman rho over 19,404 promoter shares | |---|---:| | p279 AFF1-1 vs AFF1-2 (same sample, two antibodies) | **0.712** | | p318 AFF1-1 vs AFF1-2 | **0.765** | | p454 AFF1-1 vs AFF1-2 | **0.784** | | SEM AFF1-1 vs AFF1-2 | 0.885 | | p279 vs p318 (one patient, B-ALL vs its lineage-switched AML) | 0.831 | | p279 vs p454 (two patients, both infant B-ALL) | 0.893 | | p318 vs p454 (two patients, AML vs B-ALL) | 0.832 | | SEM vs p279 / p318 / p454 (cell line vs primary) | 0.851 / 0.820 / 0.868 | **Every between-sample AFF1 correlation (0.820-0.893) exceeds every within-sample between-antibody correlation in a patient (0.712-0.784).** A cell line and a patient agree better than two antibodies in the same patient. **The cell line is the exception, and that is itself informative.** In SEM the two antibodies agree at 0.885 - better than SEM agrees with any patient. Whatever degrades the agreement is a property of primary material, not of the antibody pair as such: lower input, more heterogeneous cells, or lower oncoprotein levels in lineage-switching samples. **This is not a depth artefact, which is the first explanation to check and it fails.** AFF1-2 has the *larger* dynamic range of the two (99th percentile over median non-zero share: 8.07 vs 3.88 in p279, 23.9 vs 5.74 in p318, 9.02 vs 4.14 in p454) and 7-20% of promoters at zero against 0.8-2.0% for AFF1-1 [E1]. AFF1-2 is sparser and more peaked, not flatter - it is enriched somewhere, and that somewhere is not the fusion sites: its own top 971 promoters land in the fusion-target set at 0.4-0.9x the random null. **Consequence for this beat and for the corpus.** Every AFF1 number in this project that compares two systems is partly a reagent comparison unless the reagent is stated. The published analysis pools the two antibodies [E2], which removes the ability to see this at all.
What a good control would have been
A third AFF1 antibody, or the same antibody at two dilutions, or an AFF1 knockdown arm. None exists in this deposit. With two reagents and no tiebreak, disagreement can be attributed to either one and this analysis does not attribute it.
Where the replicate structure IS clean
DOT1L, ENL and H3K4me3 have genuine same-antibody replicates. Their agreement is H3K4me3 0.972-0.979, ENL 0.629-0.785, DOT1L 0.357-0.629. So the assay can reach rho 0.98 when the reagent is good and the target is abundant; the AFF1 and DOT1L numbers are not a ceiling imposed by CUT&RUN in primary cells. p279 DOT1L fires the pre-registered U2 guard: its two replicates agree at 0.357, which is below the 0.5 floor, so p279 DOT1L numbers are reported but are not load-bearing.
| script | analyses/q142-aff1-in-primary-infants/analyze.py |
|---|---|
| status | run |
| result summary | F3 fired: between-antibody rho 0.712-0.784 < between-sample rho 0.820-0.893 in every patient. |
| falsifier | F3 |
| outcome | FIRED |
- If an occupancy assay is proposed as a clinical pharmacodynamic readout in infant marrow, this says the reagent lot and clone must be fixed and reported, because reagent swap moves the readout more than the disease state does.
- Does the same reagent-over-biology ordering hold for the KMT2A N- and C-terminal antibody pairs at peak resolution rather than promoter resolution?
- Is the SEM-versus-primary difference in antibody agreement explained by input amount, by cell purity, or by oncoprotein abundance? All three are confounded in this deposit.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In GSE252378 primary samples, two AFF1 antibodies in the same sample correlate at rho 0.712-0.784 over promoter shares while different samples correlate at 0.820-0.893 with a matched antibody; in the SEM cell line the antibodies agree at 0.885.
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
Spearman over 19,404 promoter shares from the deposited hg38 bigWigs, computed in this beat. Promoter-level rank correlation is a coarse instrument and a factor could agree at peaks while disagreeing genome-wide; no peak-level comparison was run. n = 2 patients.
Evidence that argues against this (1)
- The two antibodies were raised in different species against different epitopes, so imperfect agreement is expected and a threshold of 'more than the between-patient correlation' is a comparison this beat chose. A reader could argue the right benchmark is the between-patient correlation of a PERFECT reagent, which is unknown, and that 0.71-0.78 is acceptable. Janssens DH et al. 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The published analysis pooled reads from the two AF4 antibodies for downstream analysis.
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
Stated in the paper's methods. Pooling is a reasonable choice when two reagents are believed to report the same protein; this entry argues the belief should have been checked, not that the choice was unreasonable.
Evidence that argues against this (1)
- Pooling two antibodies is standard practice and can improve coverage; the authors may have checked concordance in analyses not shown. The paper reports no concordance statistic either way. Janssens DH et al. 2024
the reagent spread is why the 'neither' branch fired
two clones, one protein, rho 0.71-0.78 in primary cells
antibodies agree better in the line (0.885) than in any patient
the concordance statistic that makes pooling unsafe