Oncofusion

start / graph / exp-marker-order-at-lineage-switch-paired-flow-reanalysis

Experimentspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q45-immunophenotype

Which marker moves first? A re-analysis of flow files that already exist, to put an order on lineage switch

also called marker order at lineage switch · paired flow reanalysis · early switch detection

Every hospital that treats this leukaemia keeps the raw data files from every flow cytometry test it runs, including the ones taken between diagnosis and relapse. This experiment proposes going back through the files of patients who switched from a B-cell to a myeloid leukaemia and asking, marker by marker, which one changed first.

The same thing, for a scientist

**No new samples, no new patients, no new assay.** The materials are **archived flow cytometry standard (FCS) files** already held by every centre treating KMT2A-rearranged leukaemia: diagnosis, interim MRD timepoints, and relapse, from patients who underwent lineage switch after antigen-directed therapy. **The gap this fills.** Project EVOLVE assembled 75 switches plus **10 cases of "lineage drift" - incomplete immunophenotypic change - and did not report which markers had changed in them** [E1]. A 2026 review written expressly on detection of lineage switch issues definitions and monitoring recommendations rather than an order [E2]. The one paper whose title claims a harbinger immunophenotype, PMID 32876512, is **unreadable through all five routes** [E3]. So the ordering question is open, and it is open on data that already exists. **Design.** Retrospective multi-centre, using the EVOLVE consortium as the obvious recruiting frame. Inclusion: KMT2A-rearranged B-ALL with documented lineage switch, at least two flow timepoints between the last pre-switch sample and the switch sample. Per patient, per marker, per timepoint: percent-positive and MFI, re-gated to a common template, with **all files re-analysed centrally by one operator blinded to timepoint order** - because the per-centre variability found for NG2 [E4] applies to every marker here. **Markers, chosen for a reason:** CD19, CD22, CD10, CD34, CD38, CD45, CD58, NG2/CSPG4, CD133, CD33, CD13, CD15, CD65, CD117, **CD371**, **CD14, CD64, CD11b**, CD7, CD56. **Primary output:** for each marker, the median lead time between its crossing a positivity threshold and the morphologic/clinical switch, with a per-patient ordering and an uncertainty band. **The comparator that makes it honest.** A matched arm of KMT2A-rearranged B-ALL patients who received the same antigen-directed therapy and **did not** switch. Without it the analysis measures what happens before switch and cannot say whether it also happens in everyone else - and the CD371 study shows it very well might, with transient myelomonocytic switch in **up to 65.4% of day-15 samples** and 5-year EFS of **88.3%** [E5].

Go deeper
Pre-registered predictions, so the analysis can fail
  1. CD371 moves first, because it is already the most sensitive marker of transient myelomonocytic shift under chemotherapy pressure 5.
  2. Monocytic markers (CD64, CD14, CD11b) move before CD117. CD117 has the best myeloid specificity of the classical markers (0.86 against 0.36 for CD13 and 0.44 for CD33) but every CD117-negative AML in that series was acute monocytic, and monoblastic AML is the commonest switch destination - see gene-kit. A CD117-anchored panel should therefore be late, not early.
  3. NG2/CSPG4 does not fall at switch, because the rearrangement is retained through the switch by definition and NG2 is a direct fusion target gene. If NG2 does fall while FISH stays positive, that is independent support for hyp-ng2-tracks-the-transcriptional-programme-not-the-rearrangement from patient data rather than from a dish.
  4. CD19 loss is not the first event. If CD19 falls last or simultaneously with the myeloid gain, the switch is a programme change rather than an antigen-loss event, which is what this graph's gene-cd19 already argues on outcome data and has never shown on kinetics.

What would make the result actionable rather than merely true. A lead time long enough to intervene. Project EVOLVE gives a median of 1.5 months from immunotherapy to switch with 81.4% within six months 1 - so the window is short and a marker with a two-week lead is interesting and probably not useful, while one with a two-month lead changes management. The analysis should report lead time in weeks and state explicitly whether any marker clears a plausible action threshold, because a paper that reports an ordering without a lead time will be read as an early-warning system that it is not.

Facts and figures
designretrospective multi-centre re-analysis of archived flow cytometry standard (FCS) files from KMT2A-rearranged B-ALL patients with documented postimmunotherapy lineage switch, plus a matched non-switching comparator arm exposed to the same antigen-directed therapy; all files re-gated centrally by one operator blinded to timepoint order
readoutper marker, per patient, per timepoint: percent-positive against a common threshold and MFI; primary output is median lead time in weeks between each marker crossing threshold and morphologic switch, with a per-patient ordering and uncertainty band
controlsmatched non-switching KMT2A-rearranged patients on the same therapy; central blinded re-gating; FISH or molecular confirmation of rearrangement retention in the switched sample
cost tierlow
timeline12-18 months, dominated by multi-centre data-sharing agreements rather than analysis
markers
[
  "CD19",
  "CD22",
  "CD10",
  "CD34",
  "CD38",
  "CD45",
  "CD58",
  "NG2/CSPG4",
  "CD133",
  "CD33",
  "CD13",
  "CD15",
  "CD65",
  "CD117",
  "CD371",
  "CD14",
  "CD64",
  "CD11b",
  "CD7",
  "CD56"
]
prospective predictions
[
  "CD371 moves first",
  "monocytic markers (CD64, CD14, CD11b) move before CD117",
  "NG2/CSPG4 does not fall at switch",
  "CD19 loss is not the first event"
]
tests hypotheses
[
  "hyp-ng2-tracks-the-transcriptional-programme-not-the-rearrangement"
]
How this could bite us back
  • An early-warning panel that fires on reversible induction-phase drift would cause harm. Transient myelomonocytic switch occurs in up to 65.4% of day-15 samples in one BCP-ALL subgroup whose 5-year EFS is 88.3%; a rule that escalated those children to transplant on a phenotypic signal would be a net negative. The non-switching comparator arm is what prevents this and must not be dropped for feasibility.
  • Detecting the switch earlier does not currently buy anything. Median survival after switch is 4.8 months with remission rates under 40% and no established salvage, so this experiment should be run alongside, not ahead of, work on what to do about a switch that has been detected.
  • The result will be over-interpreted as a validated monitoring panel. It is a retrospective ordering on selected, heterogeneous, multi-era archived files, and it needs prospective validation before it enters a protocol - the same derivation-set optimism that inflates every flow score in this beat.
  • If the ordering turns out to be centre-specific because of reagent and instrument drift, publishing a consensus order would export one laboratory's configuration as a clinical standard.
Still open
  1. Does the EVOLVE full text already report which markers changed in its 10 lineage-drift cases? This must be checked first and may rescope the whole experiment.
  2. Can archived FCS files from multiple centres and eras be harmonised well enough to compare marker kinetics, or is the analysis only valid within a centre?
  3. Is there any marker whose change precedes morphologic switch by long enough to act on, given a median of 1.5 months from immunotherapy to switch?
  4. Does NG2 fall at lineage switch while FISH stays positive? That would support hyp-ng2-tracks-the-transcriptional-programme-not-the-rearrangement from patient data rather than from a dish.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Project EVOLVE evaluated 75 cases of lineage switch after antigen-targeted therapy and additionally described 10 cases with incomplete changes in immunophenotype, referred to as lineage drift. Among the 70 B-ALL-derived cases, lineage switch emerged at a median of 1.5 months (range 0-36.5) after immunotherapy with 81.4% presenting within the first 6 months of the most proximal immunotherapy; 45 of 70 (64.3%) involved KMT2A rearrangements; remission rates were under 40% and median overall survival after lineage switch was 4.8 months, with similarly poor outcomes for lineage drift.

    PMID 40193715DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read in this session. **The abstract does not state which markers had changed in the 10 lineage-drift cases**, which is the precise content of this experiment's question - the full text may, and the first step of this experiment is to read it. A retrospective case collection with no denominator, so it also cannot supply the non-switching comparator arm this design requires; that has to be assembled separately.

    Evidence that argues against this (1)
    • If the EVOLVE full text already reports the marker changes in its drift cases, this experiment is partly redundant and should be rescoped to the ordering and lead-time question only. Reading the full text before designing anything is not optional. Kovach AE et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review

    In lineage switch the immunophenotype at relapse differs from that at diagnosis with the underlying genetic drivers conserved, confirming a clonal relationship. Lineage switch is most common, and was first recognised, in B-ALL with KMT2A rearrangement, which often relapses as AML. Increasing use of antigen-targeted therapies applying selective antigen pressure may increase the incidence of lineage switch across subtypes. Patients have dismal outcomes and optimal therapies remain unknown; the authors present definitions and recommendations for immunophenotypic and molecular monitoring.

    PMID 42362807DOI
    What this evidence cannot show

    **A narrative review, cited as the strongest available evidence that no marker ordering has been published and as the source of current monitoring definitions.** Abstract read in full from raw efetch XML on 2026-09-20; **full text not read**, and the monitoring recommendations it contains may well specify which markers to watch - in which case this experiment should adopt them rather than re-derive them.

    Evidence that argues against this (1)
    • Inferring an absence from what an abstract does not mention is weak, and this experiment's justification rests partly on that inference. A reader with the full text should check the monitoring recommendations before treating the marker-order question as open. Silbert SK et al. 2025
  3. E3weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    A case series titled 'Aberrant leukemia-associated immunophenotype as potential harbinger of lineage switch in KMT2A-rearranged leukemia' exists and could not be read. The PubMed record carries no abstract. All five routes of this project's full-text test were run on 2026-09-20: Europe PMC core returned pmcid null with isOpenAccess N and a zero-character abstract; the cross-check on efetch db=pubmed found no ArticleId of IdType pmc, making routes 2, 3 and 4 structurally unavailable; and resolving the DOI with a browser User-Agent returned HTTP 403 with 25 characters after tag stripping.

    PMID 32876512DOI
    What this evidence cannot show

    **Title only; nothing about the paper's content is asserted beyond its title.** Bibliographic fields fetched from PubMed efetch on 2026-09-20, not from recall. A case series, so even if readable it would not supply an order with statistical support. Any investigator with Taylor & Francis access should read it before starting, because it may already contain the candidate harbinger markers this design would otherwise rediscover.

    Evidence that argues against this (1)
    • If this series does report a harbinger immunophenotype, the experiment's framing changes from discovery to validation - a better and cheaper study. Its unreadability is a reason to seek access, not a reason to proceed as though it does not exist. Kovach AE et al. 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In 148 patients under 24 months of age (86 under 12 months) from various centres in Brazil, NG2 could be used for MLL rearrangement prediction but the accuracy of the prediction varied greatly on a per-centre basis, and the authors concluded that proper staff training and standardised sampling procedures are essential when receiving samples from multiple centres.

    PMID 21444110DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read; the abstract gives no per-centre numbers so the spread cannot be quantified. Cited here to justify **central blinded re-gating of every FCS file**, which is the design decision this finding forces: a multi-centre marker-order study that accepted each centre's own gating would produce ordering differences that are laboratory effects.

    Evidence that argues against this (1)
    • Central re-gating cannot repair differences in reagent clone, fluorochrome, instrument configuration or sample handling between centres and eras, and archived FCS files from a decade of practice will differ in all four. The ordering may be recoverable within a centre and not across centres, which would shrink the usable sample sharply. Buldini B et al. 2024
  5. E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Among 1,812 children with newly diagnosed BCP-ALL in AIEOP-BFM ALL 2009, 160 (8.8%) were CD371-positive; during induction these cases showed a transient myelomonocytic switch in up to 65.4% of samples at day 15 and an inferior early response to chemotherapy (P < .001), yet 5-year event-free survival was 88.3%. CD371 was the most sensitive marker of transient myelomonocytic switch, whose recognition the authors describe as essential for proper multiparametric flow MRD assessment.

    PMID 38215390DOI
    What this evidence cannot show

    Abstract read in full from raw efetch XML on 2026-09-20; full text not read. **The transient myelomonocytic switch here is chemotherapy-associated and largely reversible - it is NOT the postimmunotherapy lineage switch this experiment studies**, and the two must not be conflated. The cohort is enriched for DUX4 fusions, a favourable genotype, and largely excludes infants by protocol. It is cited to justify the non-switching comparator arm and the CD371-moves-first prediction, not as a result about KMT2A-rearranged disease.

    Evidence that argues against this (1)
    • Because a reversible myelomonocytic shift is common and benign, any marker that 'moves first' in the switching cohort may move equally often in the non-switching cohort. If the comparator arm shows that, the honest conclusion is that no early marker is specific, and the experiment will have produced a negative that is more useful than a spurious early-warning panel. Buldini B et al. 2024
Connected to