Score the HSPC-like fraction on paired pre-treatment and on-treatment samples from KMT2A-rearranged patients receiving a menin inhibitor — a flow-first design with a single-cell confirmatory arm
also called the persister-compartment experiment · CD34+CD19- fusion-positive fraction under revumenib
Take marrow or blood from children on a menin inhibitor, before treatment and again a few weeks in, and count what fraction of the remaining leukaemia cells look like blood stem cells rather than B cells. If that fraction grows, the drug is leaving behind the cells that cause relapse; if it shrinks, the drug is clearing exactly the right cells. Either answer changes how the drug should be used.
The same thing, for a scientist
**Design.** Paired pre-treatment and on-treatment (day 28 ± 7) bone marrow and, where available, peripheral blood from KMT2A-rearranged patients on a menin inhibitor, with the **primary readout by flow cytometry** and a **confirmatory single-cell arm** on a subset. **Primary readout (flow, deliberately):** the proportion of fusion-positive blasts that are **CD34+ CD19−**, with fusion status established per-sample by FISH or RT-PCR on sorted fractions. Flow is chosen as primary because the endpoint has to be measurable in a paediatric clinical laboratory to matter, because the samples are small, and because a per-cell transcriptome at MRD-level blast counts is unreliable. **The flow gate is an operational proxy for the signature and the experiment must validate the proxy, not assume it** — which is what the confirmatory arm is for. **Confirmatory arm:** scRNA-seq (and scATAC where cell number permits) on 8–10 paired samples, scoring each cell against the published HSPC-like signature, with per-cell fusion-spanning read detection to establish which HSPC-like cells are leukaemic. **Controls that decide whether the result means anything:** (1) a **chemotherapy-only** paired arm from the same institution, because a rising stem-like fraction under *any* effective therapy is the null and would otherwise be read as a drug-specific effect; (2) **pre-treatment technical replicates** from a second aspirate, because the endpoint is a fraction of a small population and its sampling variance has never been measured; (3) normal paediatric marrow, to establish where the CD34+CD19− gate sits in the absence of disease. **Effect size and n.** To detect a 2-fold change in a fraction whose baseline is on the order of 1–10% of blasts, with paired within-patient comparison and an assumed coefficient of variation of 0.5 on the log-ratio, roughly **25–30 paired patients** gives 80% power at α = 0.05. That is a multi-site number and this experiment is written as a correlative study on an existing trial, not as a standalone.
Why this is a correlative add-on and not a new trial
Menin inhibitors are in active paediatric and infant trials, and paired diagnosis / on-treatment marrows are collected as a matter of course for MRD. The incremental cost is an extra flow tube and a biobanked viable aliquot, and the single-cell arm can be run on stored viably frozen cells.
That also fixes the biggest weakness of the hypothesis it tests: no public dataset currently has paired single-cell samples across menin-inhibitor exposure, and nothing in a reanalysis can create one. This experiment exists because the reanalysis route is closed, and the entry should say so rather than proposing a desk study that cannot be done.
What each outcome licenses, written before the data
| result | reading | what changes |
|---|---|---|
| fraction rises ≥ 2-fold in most later-relapsing patients | hypothesis supported | combination design should add an agent active on stem-like cells; diagnostic fraction becomes a candidate stratifier |
| fraction falls ≥ 2-fold | hypothesis refuted, and the drug clears the hard compartment | strongest available argument for frontline use in infants |
| no change beyond 1.5-fold | compartment not differentially affected | look elsewhere — niche, fusion-independent persistence, pharmacokinetics |
| flow gate and single-cell signature disagree | the proxy is invalid | the clinical readout must be abandoned even if the biology is real |
The last row is the one most likely to be quietly dropped if it happens, which is why it is in the table.
| design | Paired pre-treatment and day-28 (+/- 7) bone marrow, and peripheral blood where available, from KMT2A-rearranged patients receiving a menin inhibitor. Primary readout by flow cytometry (proportion of fusion-positive blasts that are CD34+CD19-), with fusion status confirmed by FISH or RT-PCR on sorted fractions. Confirmatory scRNA-seq (plus scATAC where cell number allows) on 8-10 paired samples, scoring cells against the published HSPC-like signature with per-cell fusion-spanning read detection. Run as a correlative add-on to an existing trial or as a single-centre investigator-led cohort. |
|---|---|
| readout | log2 ratio of the HSPC-like fraction (on-treatment / pre-treatment), per patient, correlated with duration of response |
| controls | (1) a chemotherapy-only paired arm from the same institution - a rising stem-like fraction under any effective therapy is the null; (2) pre-treatment technical replicates from a second aspirate, to measure the sampling variance of the fraction, which has never been reported; (3) normal paediatric marrow to locate the CD34+CD19- gate in the absence of disease; (4) per-cell or per-fraction fusion detection, without which regenerating normal haematopoiesis is indistinguishable from leukaemic persistence. |
| cost tier | med |
| timeline | 18-30 months if run as a correlative add-on with existing biobanked material; longer if prospective consent is required |
| n target | 25-30 paired patients for 80% power to detect a 2-fold within-patient change at alpha 0.05, assuming CV 0.5 on the log-ratio; 8-10 of these for the single-cell arm |
| tests hypothesis | hyp-the-hspc-like-fraction-is-the-menin-inhibitor-persistent-compartment |
| why not a reanalysis | No public dataset holds paired diagnosis and on-treatment single-cell samples across menin-inhibitor exposure in KMT2A-rearranged leukaemia. The 25-patient HTAN resource has no drug-treated arm and GSE159608's VTP-50469 arm is cell lines, bulk, and H3K4me3 only. |
- The most likely failure of this experiment is not a null but a confident false positive: CD34+CD19- cells rise on treatment because normal haematopoiesis is recovering, the fusion control is skipped or underpowered, and a marrow-recovery artefact is published as drug-induced stem-like persistence. That is why fusion detection is listed as a control and not as an option.
- If the result is positive it will immediately be proposed as a stratifier, and the fraction is highest in the youngest infants - the patients least able to tolerate the intensification a high-risk label brings. The experiment should be designed and reported so that the prognostic claim requires a separate, properly powered analysis rather than falling out of this one.
- What IS the sampling variance of a 1-10% blast subfraction between two aspirates from the same patient on the same day? Nobody has published it, and the entire experiment's effect size is interpreted against it.
- Is day 28 the right on-treatment timepoint? Menin-inhibitor differentiation responses can be slow and a single early timepoint may sample the wrong part of the curve; a three-timepoint design costs more and may be necessary.
- Can per-cell fusion-spanning read detection be made sensitive enough at typical 3' scRNA-seq coverage, or does this arm require targeted amplification of the fusion junction?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The HSPC-like fraction is defined and detectable in KMT2A-rearranged patient samples as CD34+CD19- blasts expressing GATA2 and HMGA2, identified by single-cell multiomics in a 25-patient cohort, which gives this experiment both its flow gate and its transcriptomic signature.
Single-cell multiomics reveals increased plasticity, resistant populations, and stem-cell-like blasts in KMT2A-rearranged leukemia · Chen C, Yu W, Alikarami F, et al. (senior author Tan K) · Blood 2022What this evidence cannot show
CD34+CD19- is the marker description given for the HSPC-like population in that study; it is not a validated clinical gate and the paper did not propose it as one. Using it as the primary readout is this experiment's own design choice and is why the confirmatory single-cell arm exists. No sample in that cohort was under menin inhibition.
Evidence that argues against this (1)
- CD34+CD19- in a treated marrow is also the phenotype of normal regenerating haematopoiesis, which is abundant exactly at the on-treatment timepoint this experiment samples. Without per-cell or per-fraction fusion detection the primary readout would measure marrow recovery and report it as leukaemic persistence - which is the single most likely way this experiment produces a confidently wrong positive. Xu J et al. 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Menin inhibitors are in clinical use and trials in KMT2A-rearranged acute leukaemia including paediatric patients, so paired pre- and on-treatment marrow sampling is already routine and this study can be run as a correlative add-on.
Menin Inhibition With Revumenib for KMT2A-Rearranged Relapsed or Refractory Acute Leukemia (AUGMENT-101) · Issa GC, Aldoss I, Thirman MJ, et al. · Journal of Clinical Oncology 2025What this evidence cannot show
Cited to establish that the clinical setting and the sampling exist, not for any biological claim. This trial did not report single-cell or stem-like-fraction correlatives. Whether any active trial's biobanking consent permits this reanalysis is a regulatory question this entry cannot answer.
Evidence that argues against this (1)
- Correlative add-ons to registrational trials are slow, and a sponsor has limited incentive to run an analysis whose most likely informative outcome is that its drug leaves the relapse-causing compartment behind. An investigator-led cohort at a single large centre may reach an answer sooner even at lower n. Syndax Pharmaceuticals 2025
paired pre/on-treatment flow with a confirmatory single-cell arm; scores the hypothesis primary falsifier in both directions
designed as a correlative add-on to an existing menin-inhibitor trial
measures the fraction directly and validates the CD34+CD19- flow gate against the published transcriptomic signature
the resource that supplies the baseline fraction and the signature, and whose lack of a drug arm is why this must be an experiment rather than a reanalysis