The deposit's two lineage-switching patients behave in opposite directions: the untreated pair keeps 1126 of its 2045 oncoprotein-target promoters across the switch, while the menin-inhibited pair falls from 345 to zero - so lineage switching does not by itself clear the oncoprotein
Two children in this dataset switched their leukaemia from one type to another. The one who was not on the drug kept most of its fusion-bound genes; the one on the drug lost all of them. The switch alone does not remove the fusion protein.
The same thing, for a scientist
**Two paired B-ALL/AML patients, one confound each, and they separate the confounds between them.** | pair | drug | B-ALL calls | AML calls | Jaccard | Spearman of the score | |---|---|---:|---:|---:|---:| | p279 / p318 | none | 2045 | 1894 | 0.400 | 0.358 | | 148752 / 152985 | menin inhibitor | 345 | 0 | 0.000 | 0.031 | **The untreated switch keeps its oncoprotein.** p318, the lineage-switched AML of p279, still calls **1894** oncoprotein-target promoters, close to its own B-ALL's **2045**, and the two aliquots' scores correlate at Spearman **0.358** - above the cohort median of **0.231** for unrelated pairs [E1]. **The menin-inhibited switch loses it entirely.** 152985 is the only sample of 27 calling **zero**, from **345** in its own pre-treatment aliquot, and the two correlate at Spearman **0.031** - at the bottom of the cohort [E1]. **So the lineage switch is not what clears the oncoprotein from chromatin.** One switch happened without a menin inhibitor and the oncoprotein stayed; one happened under a menin inhibitor and it went. **In a deposit with no vehicle arm, this pair of pairs is the closest thing to a control that exists**, and it points at the drug [E2]. **Every reason to distrust that reading, listed.** The two patients carry **different fusion partners** - KMT2A::AFF1 for p279/p318 and KMT2A::ENL for 148752/152985. They come from the same institution but were processed in different batches two years apart. The timing of each aliquot relative to the switch is not recorded. n is one patient per arm. **This is a two-by-one observation, not a controlled comparison**, and it is recorded as the strongest available reading of an uncontrolled design rather than as evidence of drug effect.
What this does to a claim already in this corpus
A prior beat concluded from the p279/p318 pair that the fusion's target set is almost completely replaced across a lineage switch, using a Jaccard of 0.117 on its own promoter universe. On this beat's universe and calling rule the same pair gives a Jaccard of 0.400 - the 12th-highest of all 351 pairs in the cohort, the top 3.4%. The two measurements are not in conflict about the data; they differ in universe and calling rule, which is exactly the property this corpus has recorded before as measuring the ruler. What is robust across both is the comparison to other pairs, and on this beat's ruler the paired patient sits in the top 3.4% of pairs for similarity rather than at the bottom. That is the opposite ordering and it is a direct challenge to the earlier entry's headline, delivered here rather than left latent. Note the ranks are not strictly comparable either, because the earlier beat ranked against a handful of pairs and this one against 351.
| script | analyses/q144-unbounded-magnitude/run.py |
|---|---|
| status | run |
| result summary | Untreated switch p279->p318 keeps 1894 of 2045 calls (Jaccard 0.400); menin-inhibited switch 148752->152985 falls 345 to 0 (Jaccard 0.000). |
same patient pairs
{
"same": [
{
"i": "148752",
"j": "152985",
"spearman": 0.030782521129153646,
"ols_slope": 0.03545288730376175,
"sma_slope": 1.1201416440768663,
"A_ratio": 1.1201416440768663
},
{
"i": "p279",
"j": "p318",
"spearman": 0.3584977339530413,
"ols_slope": 0.3879438417416486,
"sma_slope": 1.1188379966850006,
"A_ratio": 1.1188379966850006
}
],
"unrelated_median_spearman": 0.23128981407491167,
"unrelated_median_A_ratio": 1.5416909691585254
}- The patient whose oncoprotein went to zero relapsed as AML anyway, so complete clearance of the oncoprotein from chromatin is not sufficient for disease control - the central worry for menin inhibitors, visible here in the one patient where the clearance is complete.
- Is there any KMT2A::AFF1 patient profiled before and during a menin inhibitor? That single sample would separate the drug from the partner in this comparison and none exists in either deposit reachable here.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In GSE252378 the untreated lineage-switching pair p279/p318 calls 2045 and 1894 oncoprotein-target promoters with a Jaccard of 0.400 and a score Spearman of 0.358, while the menin-inhibitor pair 148752/152985 calls 345 and 0 with a Jaccard of 0.000 and a Spearman of 0.031.
GSE252378 - KMT2A oncoproteins induce epigenetic resistance to targeted therapies (171 GSMs, AutoCUT&RUN / AutoCUT&Tag, hg38 fragment BEDs and bigWigs) · Janssens DH, Henikoff S (Fred Hutchinson Cancer Center / Van Andel Institute) · NCBI Gene Expression Omnibus 2024What this evidence cannot show
One patient per arm. Different fusion partners between the two pairs, different processing batches, and no recorded timing of aliquots relative to switch or dose. Promoters rather than the authors' intervals; within-sample shares, no spike-in.
Evidence that argues against this (1)
- The zero in 152985 could be a property of that patient's AML rather than of the drug, and the one counter-example - an untreated switch that keeps its oncoprotein - is in a different fusion partner. A single patient in each arm cannot separate drug from partner from patient, and no design in this deposit can. Janssens DH et al. 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The paper reports that in the 152985 sample, which lineage switched during menin-inhibitor treatment, the oncoprotein and menin are reduced to undetectable levels, while ENL persists on a subset of oncoprotein-target loci including a GMP-like program.
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 the article text and abstract served by PubMed Central (PMC11522475). The persistence of ENL was not re-measured in this beat.
Evidence that argues against this (1)
- The paper's own reading is that ENL persistence rather than oncoprotein loss is what maintains the leukaemia, which means the zero oncoprotein call in 152985 is compatible with complete on-target drug effect AND with disease progression - so it should not be read as a pharmacodynamic success without the clinical outcome, which was relapse. Janssens DH et al. 2024
the untreated switch is the counter-case that makes the zero a drug effect rather than a lineage effect
on this beat's universe and calling rule the paired patient is the MOST similar pair in the cohort, not the least
the corrected pre-/post-treatment labelling
the oncoprotein goes to zero only in the menin-inhibited pair
the untreated lineage-switched AML, re-measured from fragment counts
both paired patients
the zero-call sample is the menin-inhibited one and is withheld rather than scored as zero overlap