Oncofusion

start / graph / ana-q143-148752-and-152985-are-one-patient-before-and-during-revumenib-not-two-treated-patients

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q143-menin-under-drug-in-infants

148752 and 152985 are one patient before and during revumenib, not two treated patients - so the drug-free arm this corpus said did not exist is sample 148752

The two infant samples this project believed were both taken while a menin-blocking drug was being given are in fact one child, sampled twice: once at diagnosis before the drug, and once at relapse while on it. That means the before-and-after comparison the project had written off as impossible is available.

The same thing, for a scientist

GEO `!Sample_characteristics_ch1` for all twenty records of samples 148752 and 152985 reads `cell type: infant B-ALL/AML lineage-switching sample <id> treated with the Menin inhibitor`, and this corpus therefore recorded both as drug-exposed. The paper says otherwise. Janssens et al. write that 46 oncoprotein-binding sites were detectable in the B-ALL sample **at diagnosis**, that ENL was enriched in 148752 **prior to treatment with the menin inhibitor** and in 152985 **during treatment**, and in Methods that the AML samples were **collected upon relapse**; the two are repeatedly called **patient-matched**. 148752 is therefore the drug-naive diagnostic aliquot and 152985 the revumenib-exposed relapse aliquot of the **same child**. The GEO field names the patient's clinical course and is applied to every aliquot from that patient; it is not an exposure annotation.

Go deeper
The three sentences that settle it

All three read today from the PMC full text of the paper 1.

  1. "We compared the B-ALL and AML patient-matched samples (designated 148752 and 152985, respectively), and while 46 oncoprotein-binding sites were detectable in the B-ALL sample at diagnosis, treatment with the menin inhibitor reduced the genome-wide MLL N and C terminal signal to undetectable levels."
  2. "prior to treatment with the menin inhibitor, ENL was significantly enriched over both the B-ALL-specific genes and GMP-like program in the 148752 B-ALL sample, and during treatment, in the 152985 AML sample, ENL was only significantly enriched over the GMP-like program."
  3. Methods, Sample Information: "...including the samples that underwent a lineage-switching event in response to B-ALL directed treatment and during treatment with the menin inhibitor, as well as the two corresponding patient-matched AML samples collected upon relapse."
What the GEO record actually says, read raw with line numbers1 deeper

GSE252378_family.soft 2. For GSM8371001 the block is:

4830: ^SAMPLE = GSM8371001
4831: !Sample_title = 148752_Dot1L_R1_(240423_DJ_Hs_148752_Dot1L_R1_041624)
4832: !Sample_geo_accession = GSM8371001
4838: !Sample_source_name_ch1 = infant B-ALL lineage-switching sample 148752 treated with the Menin inhibitor
4841: !Sample_characteristics_ch1 = cell type: infant B-ALL lineage-switching sample 148752 treated with the Menin inhibitor
4842: !Sample_characteristics_ch1 = group: KMT2A-ENL

The string is identical on all ten 148752 records and all ten 152985 records, and !Sample_description adds "(paired with 152985)" / "(paired with 148752)" - which is the only place in the metadata that hints the two are one patient. Nothing in GEO says which aliquot was drawn before the drug. The reading that produced the error was not careless; the field is genuinely ambiguous and it resolves the wrong way.

The off-by-one, for the record

The first draft of this entry's own pre-registration cited line 4842 for the cell type: string. 4842 is the group: line. It was corrected by printing absolute line numbers rather than re-extracting, which is the procedure this project adopted after a 16.4% metadata-defect claim turned out to be an off-by-one in the opposite direction.

What changes because of this
  • The dispatch brief for this beat opened "There is NO drug-free arm from these patients ... Do not design a treated-versus-untreated comparison. It does not exist." It exists, and it is within one patient.
  • ana-the-infant-menin-inhibitor-occupancy-experiment-exists-and-it-is-kmt2a-enl-not-aff1 describes these as "two primary infant lineage-switching patient samples ... profiled under a menin inhibitor". That should read one patient, sampled at diagnosis before revumenib and at relapse during it. It also says "two replicates each" for the KMT2A channels; those are four different antibodies (N1 Millipore 05-764, N2 CST D2M7U, C1 Santa Cruz H10, C2 Millipore 05-765), not replicates, which is a stronger control and a different one.
  • The comparison is still not a clean pharmacodynamic pair - see ana-q143-drug-exposure-and-clonal-replacement-are-completely-confounded-across-this-pair.
Facts and figures
correction ofana-the-infant-menin-inhibitor-occupancy-experiment-exists-and-it-is-kmt2a-enl-not-aff1
corrected claimtwo primary infant lineage-switching patient samples profiled under a menin inhibitor
should readone patient, two aliquots: 148752 = diagnostic infant B-ALL drawn before revumenib; 152985 = relapse infant AML drawn during revumenib
second correctionthe KMT2A N1/N2/C1/C2 libraries are four different antibodies, not two replicates
patientone child, Texas Children's Hospital, KMT2A::MLLT1 (group KMT2A-ENL)
statusrun
scriptanalyses/menin-under-revumenib-in-an-infant-patient-pair/PREREGISTRATION.md
settled by
{
  "pmid": "39472576",
  "read": "full text at PMC11522475, not abstract only"
}
How this could bite us back
  • This entry makes a patient-matched before/after menin inhibitor pair available in primary infant material, which is rare enough that it will be quoted. It is one patient, the two aliquots are different lineages, and the later one is a relapse clone - so it is a case report with genomics attached, not a controlled experiment.
  • Anyone re-using the GEO annotation to build a treated-versus-untreated cohort across this series will mislabel the diagnostic aliquots of both lineage-switching patients.
Still open
  1. How long had this child been on revumenib when 152985 was drawn, and at what dose? Neither is in the paper, the GEO record or the repository.
  2. Was 148752 drawn before any therapy at all, or before the menin inhibitor but after B-ALL-directed treatment? The paper says 'at diagnosis' but the same patient group is described as switching 'in response to B-ALL directed treatment'.
  3. How many other GEO series annotate a patient's clinical course in a per-sample cell-type field, and how many of this project's deposit-level claims rest on that field?
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Janssens et al. state that 148752 is the patient-matched B-ALL sample at diagnosis and 152985 the AML sample collected at relapse during menin-inhibitor treatment.

    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 2024
    PMID 39472576DOI
    What this evidence cannot show

    Read from the PMC full text (PMC11522475), Results section 'Menin-inhibitor-resistant ENL-marked epigenetic lesions' and Methods 'Sample Information'. The paper never states the interval between the two draws, the dose, or how long the patient had been on drug when 152985 was taken.

    Evidence that argues against this (1)
    • The paper never says 148752 was drug-NAIVE in those words; it says 'at diagnosis' and 'prior to treatment with the menin inhibitor'. A patient can be at diagnosis and already on other therapy, and this child received B-ALL-directed treatment. So 148752 is drug-free with respect to the menin inhibitor only, which is the comparison this beat makes, and is not a treatment-naive sample in any broader sense. Janssens DH et al. 2024
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    The GEO characteristics field carries the same 'treated with the Menin inhibitor' string on all twenty records of both samples, including the diagnostic one.

    What this evidence cannot show

    Read from the raw SOFT family file with absolute line numbers, not re-extracted with a parser. The field is a patient-level annotation; GEO has no separate exposure field in this series.

    Evidence that argues against this (1)
    • An annotation that is wrong about exposure is not necessarily wrong about anything else: the same audit found 0 of 44 antibody/label disagreements in this deposit, so its fusion and antibody assignments are reliable and only the treatment word is misleading. Janssens DH et al. 2024
Connected to