The dispatch brief scoped this deposit's oncoprotein-score arm at two patients and a cell line; it is 27 patient samples across six fusion partners, and the scoping error was inherited from a beat that only needed three of them
I was told the relevant part of this dataset covered two children and one lab cell line. Counting every record showed 27 patient samples with the same complete set of antibodies, which makes the planned comparison an order of magnitude stronger than expected.
The same thing, for a scientist
🟠 **CORRECTED 2026-09-24 (cycle 131): the independent-patient count is 24, not 25.** This entry found **two** same-patient pairs and missed a third. GEO's `!Sample_description` declares *paired with* for **three** pairs: `p279`/`p318` and `148752`/`152985` (Texas Children's) **and `R3595`/`R3652` (Seattle Children's)**. 27 samples minus three pairs is **24 independent patients**. Verified by the king directly from the deposit's sample metadata, parsing records on `!Sample_title` (GEO's brief format emits the title before the accession). Found by q147. --- **The brief's premise, quoted:** *"the AFF1 arm is two patients plus a line, not three."* That is **true of the AFF1 antibody arm** - `SEM`, `p279`, `p318`, `p454`, of which p279/p318 are one infant - and it is the arm cycle q142 analysed. It is **not** true of the KMT2A N-over-C arm, which is the arm the oncoprotein score is built from [E1]. **Measured by iterating all 171 sample records:** 27 samples carry a complete `KMT2A_N1/N2/C1/C2` set. Every one of the 27 is complete - there is no partial set to discard [E1]. | partner | complete N/C sample sets | |---|---:| | KMT2A-AFF1 | 8 | | KMT2A-ENL | 7 | | KMT2A-AFDN | 4 | | KMT2A-AF10 | 3 | | KMT2A-AF9 | 3 | | KMT2A-EPS15 | 2 | Four institutions contribute. **Three** of the 27 samples pair off as one patient in two disease states (`p279`/`p318` and `148752`/`152985` at Texas Children's, `R3595`/`R3652` at Seattle Children's), so the **independent-patient count is 24**, not 27 *(corrected in cycle 131; this entry originally said two pairs and 25)* - a distinction that matters for any test whose n is a patient count [E1]. **Why the error propagated.** q142 needed the AFF1 antibody, which exists in only four samples, so its inventory was correct **for its beat** and was carried forward as a description of the deposit. The general form: *an inventory scoped to one antibody is not an inventory of the deposit*, and the brief that quotes it inherits the scope silently, because nothing in the sentence says which antibody it was counted under.
What this changes about what is answerable
A magnitude-transfer test on two patients and a line has n = 3 and cannot separate a biological difference from a technical one. At 27 samples with 24 independent patients, the same test has a between-patient distribution and - because N1/N2 and C1/C2 are two antibodies rather than replicates - a within-patient technical distribution to compare it against. The instrument for the null was in the deposit the whole time.
| script | analyses/q144-unbounded-magnitude/run.py |
|---|---|
| status | run |
| result summary | 27 complete N/C sample sets, 24 independent patients (three GEO-declared same-patient pairs), six fusion partners, four institutions. |
| brief premise corrected | the deposit's N/C arm is 27 samples, not 2 patients and a line |
| corrected cycle 131 | result_summary said 25 independent patients; the entry found two same-patient pairs and missed R3595/R3652 at Seattle Children's. 27 - 3 = 24. |
- Which nine of the paper's 36 profiled samples are absent from this deposit's complete-N/C set, and were they dropped for quality or for privacy?
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset
GSE252378 contains 27 patient samples with a complete KMT2A N1/N2/C1/C2 set spanning six fusion partners and four institutions, of which 24 are independent patients (three GEO-declared same-patient pairs: p279/p318, 148752/152985, R3595/R3652).
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
Counted by iterating every one of the 171 columns of the GEO series matrix and parsing !Sample_title, !Sample_description and !Sample_characteristics_ch1; no sampling. 'Independent patient' rests on the reciprocal 'paired with' statements in !Sample_description for p279/p318 and 148752/152985; no other pairing is declared and none was inferred.
Evidence that argues against this (1)
- The paper reports 36 samples including four cell lines, more than the 27 complete N/C sets found here, so either some profiled samples were not deposited with a complete set or some sit outside this accession. The 27 is a floor on what this deposit supports, not a reconstruction of the published cohort. Janssens DH et al. 2024
inventory of the deposit's N/C arm
corrects the scope carried forward from the AFF1-antibody arm