start / graph / ana-q144-the-fusion-target-promoter-sets-of-two-patients-overlap-only-1-6-fold-above-what-two-random-sets-of-the-same-size-would-give

AnalysisstrongQ144-unbounded-magnitude

Two KMT2A-rearranged patients share oncoprotein-target promoters at a median Jaccard of 0.132 against an independence expectation of 0.069 - an enrichment of only 1.63x, with 5 of 351 pairs at or below chance

Two children's leukaemias share about an eighth of their fusion-bound genes. That sounds low, but two random lists of the same length would already share a fifteenth, so the real overlap is only about 1.6 times better than chance.

The same thing, for a scientist

**A Jaccard index is not a result until the chance overlap is beside it.** Two sets of sizes a and b drawn independently from a universe of N already overlap at an expected Jaccard of `(ab/N) / (a + b - ab/N)`, and for the set sizes seen here that expectation is **0.069** [E1]. **Measured over 351 pairs** on the **4646** common promoters: observed median Jaccard **0.132**, expected **0.069**, **ratio 1.63** (10th-90th percentile of the ratio 1.24 to 2.55). **5 of 351 pairs sit at or below the independence expectation** [E1]. Depth-matched, the observed median falls to **0.090** [E1]. **This reframes a number this corpus has already reasoned from.** A prior beat compared one patient's B-ALL against its own lineage-switched AML at Jaccard **0.117** and two unrelated infant B-ALLs at **0.149**, and concluded that one patient's two states share *less* than two different patients do. **Neither number was published with a chance floor, and one cannot be computed for them from the entry alone**, because the expectation depends on that beat's own set sizes and promoter universe, which its entry does not record. What can be said is that the floor measured on THIS universe and THESE set sizes is 0.069, that raw Jaccards in the 0.1 range are routinely within a small factor of such a floor, and that the comparison between 0.117 and 0.149 is a comparison of two enrichments over a shared baseline [E2]. **The direction of that earlier finding is not overturned - it is re-scaled**, and its argument needs the null printed beside it. **What survives, and it is the load-bearing part.** The enrichment is **real**: the median pair is 1.63x above chance and only 5 of 351 pairs fall to it. Fusion-target promoter sets are not independent draws. They are also nowhere near a shared program: **the overwhelming majority of one patient's oncoprotein-target promoters are not another patient's.**

Go deeper

The general rule this is an instance of

Any set-overlap statistic between two called sets inherits the calling rule's false-positive rate twice over. If a fixed fraction of each set is a fusion-independent call - and the fusion-negative CD34+ control says a substantial fraction is, at this interval scale - then both sets contain a shared contaminant AND a private random component, and a raw Jaccard cannot separate those from a shared biological program. Printing the independence expectation is the minimum; a permutation over the calling procedure itself would be better and was not done here.

Facts and figures

script	analyses/q144-unbounded-magnitude/run.py
status	run
result summary	Median Jaccard 0.132 vs independence 0.069 (ratio 1.63); 5 of 351 pairs at or below chance.

jaccard

{
  "universe_promoters": 4646,
  "observed_median": 0.1320754716981132,
  "expected_under_independence_median": 0.06895315294989768,
  "observed_over_expected_median": 1.6314772603470553,
  "observed_over_expected_p10": 1.2380866107435564,
  "observed_over_expected_p90": 2.5455576757206546,
  "n_pairs_at_or_below_null": 5,
  "n_pairs": 351,
  "note": "Expected Jaccard for two independent sets of the observed sizes in a universe of this many measurable promoters.  A raw Jaccard near 0.1 can be entirely this."
}

Still open

What does a coverage-preserving permutation null give for these Jaccards? That is the number that would turn the 1.6x ceiling into an estimate.
What were q142's set sizes and promoter universe? Without them its Jaccards cannot be placed against a chance floor, and the entry should record them.

Every source behind this entry (2)

E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
Across 351 pairs of KMT2A-rearranged patient samples in GSE252378, oncoprotein-target promoter sets called by the published criterion overlap at a median Jaccard of 0.132 against an independence expectation of 0.069, a ratio of 1.63, with 5 pairs at or below the expectation.
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 2024
What this evidence cannot show
The independence expectation is analytic, not permutational, and assumes calls are exchangeable across the universe of 4646 promoters. They are not - promoters differ in coverage and in chromatin state - so the true chance floor is HIGHER than the analytic one and the enrichment reported here is an upper bound on the enrichment. Promoters, no spike-in.
Evidence that argues against this (1)
- An analytic independence null is the weakest of the available nulls. A permutation that preserved each promoter's coverage and each sample's call count would give a higher floor and a smaller enrichment, so the 1.6x quoted here is the most generous reading of the data and should be treated as a ceiling on the shared program, not an estimate of it. Janssens DH et al. 2024
E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
A prior beat in this corpus measured a Jaccard of 0.117 between one patient's B-ALL and its own lineage-switched AML, against 0.149 between two unrelated infant B-ALLs, and concluded that the fusion's target set moves more within a patient than between patients.
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 2024
What this evidence cannot show
Those figures were computed on a different promoter universe and a different calling rule from this entry's, so they are not directly comparable in level; what is comparable is that neither carried an independence null.
Evidence that argues against this (1)
- That beat's comparison is between two numbers computed the same way, so a common baseline cancels and the DIRECTION of its finding is unaffected by the missing null. What is not established is the SIZE: its entry does not record the set sizes or the promoter universe it used, so its own chance floor cannot be reconstructed from the corpus and no claim is made here about where 0.117 sits relative to it. Janssens DH et al. 2024

Connected to

related to →The B-ALL and the AML of one infant share no more fusion-occupied promoters than two unrelated infants do - lineage, not patient identity, sets where KMT2A::AFF1 sits
supplies the independence null that comparison was missing; direction unaffected, scale corrected
depends on →GSE252378 fragment BEDs: a complete KMT2A N-terminal / C-terminal antibody set in 27 patient samples across six fusion partners, deposited and until now unopened by this corpus
the 27-sample N/C arm
depends on →At promoter resolution the published N/C oncoprotein criterion calls 1043 promoters in fusion-negative CD34+ HSPCs - above the median across 27 KMT2A-rearranged patients - while the same-reagent controls in that sample call 66 and 11
the false-positive rate that contaminates both sets in every pair
related to →Scored in the published criterion's own currency, the number of oncoprotein-target promoters runs 0 to 2282 across 27 KMT2A-rearranged patient samples - 20.01x between the 10th and 90th percentiles - and the spread widens to 61.10x after depth matching
the same call sets, counted rather than intersected
← refutesOn one universe with one rule and a structure-respecting null, the two lineages of one infant overlap no more than two unrelated KMT2A::AFF1 infants do
the 1.6x is the cohort median against the cohort median's own null; the same-patient pair's excess is 1.33-1.46x, not 5.8x
← testsThe two beats' opposite answers are produced by the calling rule, not by the promoter universe: three universes give one answer, two rules give two
q144's absolute rule reproduced in three universes
← refutesThe 0.069 chance floor quoted beside the 0.400 same-patient Jaccard is the cohort MEDIAN's null, and this pair's own floor is 0.267
0.069 is that entry's cohort-median floor and is not the floor for the same-patient pair
← refutesGSE252378 carries an institution-level batch effect: unrelated pairs from the same hospital overlap more than cross-hospital pairs, and a cohort-median comparator absorbs that into the same-patient claim
the cohort median comparator is confounded with institution and the same-patient pair is same-institution by construction
← related toThe '0.400, keeps 1,894 of 2,045 calls' in the dispatch brief is two independent per-sample call counts read as a retention ratio; the intersection is 1,116 of a 2,812 union
the per-sample table that was misread; the table itself is correct
← refutesq142's exact comparison - p279 against its own switched AML versus p279 against an unrelated infant - replicates in q144's own universe with q144's own rule, and q144 never made it
q144's own universe and rule give 0.397 same-patient against 0.434 for the matched unrelated pair