Oncofusion

start / graph / hyp-the-hspc-like-fraction-is-the-menin-inhibitor-persistent-compartment

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q139-single-cell

The HSPC-like blast fraction is the compartment that survives menin inhibition, and its size at diagnosis predicts depth and durability of response

also called stem-like persister hypothesis for menin inhibitors · does revumenib shrink or enrich the HSPC-like fraction

Menin inhibitors produce deep remissions in this leukaemia that usually do not last. We think the cells that survive the drug are the small stem-cell-like group already present at diagnosis, and that counting that group before treatment would predict which children relapse. The prediction is falsifiable in both directions and the data to test it largely already exists.

The same thing, for a scientist

Menin inhibitors in KMT2A-rearranged leukaemia produce a characteristic clinical shape: high response rates, deep remissions, and relapse. Separately, single-cell work identifies an **HSPC-like blast fraction** enriched in the youngest KMT2A-rearranged patients, with downregulated steroid-response pathways, an immunosuppressive circuit, and — in the pan-leukaemia generalisation — prognostic value for drug response and survival. **We hypothesise that these are the same phenomenon: the HSPC-like fraction is the menin-inhibitor-persistent compartment, and its fractional size at diagnosis predicts depth and durability of response.** The mechanism proposed is *not* that the fraction is fusion-independent. It is that the fraction's **differentiation is not the fusion's to release**: a menin inhibitor works by allowing arrested blasts to differentiate, and a cell already positioned above the lymphoid commitment point has further to go and fewer differentiation cues available, so it is the slowest to clear and the likeliest to repopulate. **The hypothesis has a serious competing account and one datum already cutting against it.** The pan-leukaemia HSPC-like signature sits under a HOXA/AP-1/CEBPA network, which would predict the fraction is MORE menin-dependent, not less — so the drug should preferentially remove it. And this corpus's own reanalysis found the menin-inhibitor response in two KMT2A::AFF1 lines **did not touch HOXA or MEIS1**, which weakens the HOXA-dependence limb on both sides. The hypothesis is therefore `speculative`, and its value is that the two predictions are opposite and cheap to separate.

Go deeper
The prediction, stated so it can be scored

Primary prediction. In KMT2A-rearranged leukaemia sampled before and during menin-inhibitor therapy, the HSPC-like fraction (defined by the published single-cell signature, or by CD34+CD19− blasts carrying the fusion) increases as a proportion of residual blasts by at least 2-fold between diagnosis and the first on-treatment timepoint, in at least half of patients who subsequently relapse.

Secondary prediction. Diagnostic HSPC-like fraction above the cohort median is associated with shorter duration of response, hazard ratio ≥ 1.5.

Direction matters and the opposite result is informative. If the fraction falls ≥ 2-fold, the HOXA-dependence account wins and the hypothesis is refuted, and that would itself be a useful finding — it would mean the menin inhibitor is preferentially clearing the compartment that chemotherapy cannot, which is the best possible argument for moving it into frontline infant therapy.

A null — the fraction neither rises nor falls beyond 1.5-fold — means the compartment is not differentially affected and the drug's failure lies elsewhere, which points the next cycle at the niche or at fusion-independent persistence rather than at cell state.

Why this is worth asking now rather than after another cycle of mechanism

Three things converged this cycle:

  1. The 25-patient KMT2A-r single-cell multiome is deposited and, until now, uncited by this project — so the diagnostic-fraction half of the question has data already.
  2. GSE159608 contains a menin-inhibitor arm in SEM and RS4;11 with matched vehicle. That is not single-cell and not patient material, but it is a chromatin readout under the drug in the index fusion, which the corpus believed did not exist.
  3. This corpus already holds resistance-lineage-switch-under-menin-inhibition — lineage switch is a documented resistance route for this drug class. A lineage switch under a menin inhibitor requires a lineage-plastic cell to survive the menin inhibitor, which is this hypothesis restated as an observed clinical event. That is the strongest indirect support available and it is already in the graph.
Facts and figures
statementIn KMT2A-rearranged leukaemia, the HSPC-like blast fraction is the compartment that persists under menin inhibition, and its fractional size at diagnosis predicts the depth and durability of menin-inhibitor response.
originliterature-gap
prior probability0.3
predictions
[
  "The HSPC-like fraction, as a proportion of residual fusion-positive blasts, rises at least 2-fold between diagnosis and the first on-treatment timepoint in at least half of patients who later relapse.",
  "Diagnostic HSPC-like fraction above the cohort median associates with shorter duration of menin-inhibitor response, HR >= 1.5.",
  "Patients who undergo lineage switch on a menin inhibitor have a larger diagnostic HSPC-like fraction than those who relapse without switching.",
  "In vitro, sorted CD34+CD19- fusion-positive blasts show higher menin-inhibitor IC50 than the bulk CD19+ blast population from the same patient."
]
experiments
[
  "exp-hspc-like-fraction-before-and-during-menin-inhibition"
]
analyses
[
  "ana-the-htan-kmt2a-r-single-cell-multiome-is-the-reanalysable-resource"
]
falsifiers
[
  "The HSPC-like fraction, scored with the published signature on paired diagnosis and on-treatment single-cell samples from KMT2A-rearranged patients on a menin inhibitor, FALLS by 2-fold or more as a proportion of residual fusion-positive blasts in the majority of patients. That is the HOXA-dependence prediction and it kills this hypothesis outright.",
  "Sorted CD34+CD19- fusion-positive blasts from a KMT2A::AFF1 patient sample show a menin-inhibitor IC50 within 1.5-fold of the bulk CD19+ blasts from the same patient, in a 7-day ex vivo assay. Equal sensitivity means the compartment is not a persister compartment.",
  "Diagnostic HSPC-like fraction shows no association with duration of response in a menin-inhibitor-treated cohort of at least 40 KMT2A-rearranged patients (95% CI for the hazard ratio spanning 1.0 with an upper bound below 1.5).",
  "The HSPC-like fraction is found NOT to carry the fusion transcript in the deposited single-cell data - in which case it is normal residual haematopoiesis, the whole hypothesis is about the wrong cells, and the observation that it expands under therapy becomes trivially expected."
]
How this could bite us back
  • If this hypothesis is right, adding a menin inhibitor to frontline infant therapy deepens early response without changing the compartment that causes relapse - which would look like success on every early endpoint (MRD at day 33, complete remission rate) and change nothing at 5 years. That is the most expensive possible way to be wrong and the trials currently being designed would not detect it.
  • A diagnostic HSPC-like fraction used to escalate therapy would escalate in the youngest infants, who carry the most of it and tolerate intensification worst. A predictive biomarker that selects the least fit patients for the most toxic treatment needs a benefit estimate before it needs a cut-off.
  • If the fraction is instead the most menin-sensitive compartment, then the drug is best used exactly where this hypothesis says it is worst, and a lab that had adopted this hypothesis would have argued against the correct trial design.
Still open
  1. Does any paired diagnosis / on-treatment single-cell dataset exist for a patient on a menin inhibitor? If not, this hypothesis's primary falsifier cannot be scored from public data and the entry should say so prominently.
  2. Can the HSPC-like signature be scored on flow cytometry rather than scRNA-seq? A CD34/CD19/CD33 gate is available in every paediatric flow laboratory and a transcriptomic-only predictor will not be used clinically.
  3. If the fraction does persist, is it because it is fusion-independent or because it is differentiation-incompetent? Those need different next drugs - the first needs a second target, the second needs a differentiation cue.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    An HSPC-like blast population exists in the blood of younger KMT2A-rearranged patients, contains leukaemic blasts, is CD34+CD19-, expresses GATA2 and HMGA2, and is accompanied by increased lineage plasticity and downregulated steroid-response pathways in the most immature blasts of patients under 6 months.

    PMID 34864916DOI
    What this evidence cannot show

    Nothing in this study is under menin inhibition or any targeted therapy; the resistance in its title is to chemotherapy and immune control. The link to menin inhibitors is entirely this hypothesis's own inference. Mixed KMT2A partners, 25 patients.

    Evidence that argues against this (1)
    • The same body of work reports the HSPC-like signature under a HOXA/AP-1/CEBPA regulatory network, which predicts the OPPOSITE of this hypothesis: a HOXA-driven compartment should be the most menin-dependent and therefore the first to be cleared by a menin inhibitor, not the last. Chen C et al. 2025
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Lineage switch is a documented resistance route under menin inhibition in KMT2A-rearranged leukaemia, which requires a lineage-plastic cell to have survived the drug.

    PMID 34864916DOI
    What this evidence cannot show

    This is an inference joining two sources, not a single reported result: the plasticity substrate is from the single-cell paper and the clinical resistance route is from this graph's own resistance entry and the case literature behind it. Nobody has reported a single-cell profile of a leukaemia switching lineage UNDER a menin inhibitor. Stated as reasoning, not as a measurement.

    Evidence that argues against this (1)
    • Lineage switch under menin inhibition is reported in a small number of cases and could reflect the natural history of KMT2A-rearranged disease under any effective pressure rather than anything specific to the drug or to a stem-like compartment. The corpus's own Project EVOLVE evidence puts most documented switches after CD19-directed immunotherapy, not after menin inhibition. Silbert SK et al. 2025
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    The HOXA-dependence limb of the competing account is weakened by this corpus's own reanalysis, which found the menin-inhibitor transcriptional response in two KMT2A::AFF1 lines did not move HOXA or MEIS1.

    ana-the-menin-inhibitor-response-is-nine-fold-fusion-selective-and-does-not-touch-hoxa-or-meis1-in-either-kmt2a-aff1-line · Oncofusion Knowledge Graph · Iqbal Lab oncofusion knowledge graph repository 2026
    What this evidence cannot show

    An internal analysis entry, cited as an internal analysis entry. It concerns two KMT2A::AFF1 cell lines and a transcriptional readout, not the HSPC-like fraction and not patient material, so it weakens the HOXA-dependence prediction without supporting this hypothesis positively. I did not re-derive its result.

    Evidence that argues against this (1)
    • Two cell lines are not the disease, and a transcriptional non-response of HOXA in an established line says little about a rare stem-like compartment in a patient. A reader should treat this as a reason not to assume HOXA dependence rather than as evidence against it. Oncofusion Knowledge Graph coordinator 2026
Connected to