Oncofusion

start / graph / hyp-breakpoint-position-prognostic-interfant

Hypothesismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q9-landscapedisputed

Hypothesis: the KMT2A genomic breakpoint carries prognostic information that the field has been sitting on for twenty years

also called breakpoint prognostic hypothesis · 249-patient hypothesis

The exact place where the chromosome breaks differs between patients, and we think that difference changes how the disease behaves. It was tested once in thirty patients and never again, while two groups hold everything needed to test it properly.

The same thing, for a scientist

Frankfurt reports an age-dependent shift in breakpoint position within the KMT2A breakpoint cluster region across 3,401 patients, which implies the breakpoint is not random with respect to biology. Interfant-06 generated patient-specific genomic breakpoint information for 249 KMT2A-rearranged infants purely as MRD assay plumbing and holds their protocol-standardised outcomes. The hypothesis is that breakpoint position, and the retained/lost KMT2A domain content it determines, predicts MRD kinetics, relapse timing or relapse lineage independently of age, white cell count and MRD - and that this has gone untested only because the sequence and the outcome sit in different institutions.

Go deeper
The mechanistic argument, and where it is weakest1 deeper

For. The breakpoint determines which KMT2A domains are retained in the fusion. Breaks in intron 9 versus intron 11 differ in what they keep, in their repeat context (intron 9 is Alu-rich, intron 11 carries topoisomerase-II-associated features), and in the reciprocal product they produce. Frankfurt's age-dependent breakpoint shift 1 says the breakpoint tracks something real about the cell and the exposure in which the break occurred. If the breakpoint reports on the cell of origin, and the cell of origin determines the disease's behaviour, the breakpoint should carry prognostic information.

Against, and this is serious. Age at diagnosis already predicts outcome in infant ALL and is the variable the breakpoint shift is defined against. In a 0-12 month cohort the two may be collinear to the point of being one measurement. A positive association could therefore be age wearing a molecular costume. The analysis must be designed to survive that objection or it is not worth running - which is why breakpoint-age collinearity assessment on the Frankfurt data is a gating step before anything is requested from Interfant.

A second objection: n=249 with a binary breakpoint classification detects a moderate effect, not a subtle one. A null result at this sample size does not exclude a small real effect.

What makes it worth doing anyway

Three things.

First, cost. No samples, no patients, no assays, no reagents. The marginal cost is analysis time plus two data-access negotiations.

Second, either answer is publishable and decision-relevant. If breakpoint predicts outcome, risk stratification in infant ALL gains a variable available at diagnosis, and the platform argument for a single menin-interface drug weakens because different breakpoints may mean different biology. If it does not, a twenty-year-old open question closes and the field stops speculating.

Third, it is the entry ticket. A completed, useful analysis with Interfant and Frankfurt makes every subsequent request to either institution easier, and puts the lab inside the only consortium that controls access to this patient population.

Correction: this has been looked at once, and the original premise was wrong

An earlier version of this entry asserted that nobody had looked. That is false, and this project's own breakpoint entry already recorded the contrary.

Emerenciano and colleagues examined it in 2013: a clinical cohort of 545 children, with breakpoints localised in 30 of them, reporting that patients with poor outcome preferentially carried breakpoints within KMT2A intron 11 [E_prior].

The hypothesis survives in sharpened form. Thirty localised breakpoints cannot support a multivariable prognostic claim, the finding has never been replicated, and three successive large recombinome papers repeat it without adding survival data, with the same authors hedging it by 2023. So the correct claim is not that the question is unexamined but that it has been examined once, underpowered, and never at scale with protocol-standardised outcomes. That is still a real and cheap opportunity, and it is now stated accurately.

Facts and figures
statementGenomic KMT2A breakpoint position, and the retained/lost domain content and reciprocal-product status it determines, carries prognostic information in infant KMT2A-rearranged ALL that is independent of age, white cell count and MRD. It has been tested once, in 30 patients with localised breakpoints, and never replicated at scale with protocol-standardised outcomes, because the breakpoint sequences (Interfant n=249; Frankfurt n=3,401) and the outcome data sit in different institutions and the breakpoint was treated as assay plumbing rather than as data.
originliterature-gap
prior probability0.35
gating unknownWhether Interfant-06 archived the patient-specific genomic KMT2A breakpoint sequences generated for MRD assay design. One email to the Princess Maxima Center coordinating centre resolves it.
cost tierlow
predictions
[
  "Breakpoint position within the KMT2A breakpoint cluster region will associate with end-of-induction MRD category, relapse timing or relapse lineage in the Interfant-06 KMT2A-rearranged cohort.",
  "Any association will be strongest for breakpoint classes that differ in retained domain content or in reciprocal-product status, rather than being a smooth function of genomic position.",
  "Cases predicted to lack a reciprocal product (for example spliced KMT2A::MLLT1 — CORRECTED 2026-09-19, this previously said MLLT10, which is the wrong partner: the reciprocal-free spliced population is MLLT1, 241 of 460 cases, whereas MLLT10 is not a spliced-fusion-enriched partner) will behave differently from reciprocal-expressing cases of the same fusion within the Frankfurt cohort.",
  "The effect, if present, will survive adjustment for age at diagnosis - and if it does not, the honest conclusion is that breakpoint is a proxy for age, which is itself a publishable clarification."
]
experiments
[]
analyses
[
  "ana-interfant-breakpoint-prognostic-reanalysis"
]
falsifiers
[
  "No association between breakpoint position (or derived domain content) and EFS, MRD category, relapse timing or relapse lineage in the Interfant-06 cohort after adjustment for age, WBC and prednisone response.",
  "Breakpoint position and age at diagnosis prove collinear within 0-12 months to the point that the effect cannot be separated, which makes the hypothesis untestable in infants rather than false.",
  "Reciprocal-negative and reciprocal-positive cases of the same fusion are indistinguishable in the Frankfurt cohort on every available variable.",
  "The Interfant-06 breakpoint sequences were not archived, which makes the hypothesis inaccessible rather than refuted - a distinct and more terminal outcome."
]
How this could bite us back
  • A positive result undermines the platform argument the lab makes to funders: if different breakpoints mean different biology, one menin-interface drug does not serve all KMT2A-rearranged patients.
  • A null result closes one of the lab's clearest differentiating angles, though it settles a long-standing question.
  • Age confounding could render the analysis uninterpretable rather than negative, which is worse than either clean outcome.
  • The whole hypothesis rests on an inference that breakpoint sequences were archived; if they were not, two collaboration requests and considerable goodwill are spent for nothing.
Still open
  1. Were the Interfant-06 breakpoint sequences archived, or only the derived MRD primers? Primers still encode approximate breakpoint position and would be a degraded but usable substitute.
  2. Should the collinearity assessment be run on the Frankfurt cohort alone first, as a zero-governance pilot, before approaching Interfant at all?
  3. Is the right primary endpoint survival, or MRD kinetics? MRD is measured earlier, has more events, and is less confounded by post-relapse therapy, so it may be the better-powered test of the same biology.
  4. Does breakpoint position predict menin-inhibitor sensitivity? Nobody can answer this yet because too few KMT2A-rearranged infants have received one, but it is the version of the question that would matter most to the lab's central hypothesis.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Frankfurt reports an age-dependent shift in the position of the breakpoint within the KMT2A breakpoint cluster region across a series that now totals 3,401 patients, establishing that breakpoint position is non-random with respect to patient biology.

    The MLL recombinome of acute leukemias in 2017 · Meyer C, ... Marschalek R · Leukemia 2017
    PMID 28701730DOI
    What this evidence cannot show

    A descriptive association in a referral cohort, not replicated in a population-based series with matched outcome data as far as this survey found. The age effect spans infants to adults; whether it is detectable within a 0-12 month window is unknown and is precisely the collinearity question.

    Evidence that argues against this (1)
    • The premise that 'nobody has looked' is wrong, and this materially changes the hypothesis's novelty and its prior. The Brazilian Collaborative Study Group of Infant Acute Leukaemia reported in 2013 that MLL breakpoint localisation correlates with outcome in infant leukaemia: in a cohort of 545 children aged 24 months or under (385 ALL, 160 AML) with genomic breakpoints mapped in a subset of 30 MLL-rearranged cases, patients with poor outcome preferentially carried breakpoints within MLL intron 11, and the authors concluded breakpoint localisation 'should be considered as a novel outcome predictor'. The finding points the same way as the hypothesis, which is good news, but it rests on 30 mapped breakpoints, has not been replicated in thirteen years, and is confounded with age and white cell count, which were themselves prognostic in the same cohort. The hypothesis should be reframed as a replication with adequate power rather than as an unexamined question. Emerenciano M et al. 2013
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Interfant-06 measured MRD in 249 infants with KMT2A-rearranged ALL by DNA-based PCR of rearranged KMT2A, immunoglobulin and/or T-cell receptor genes, alongside protocol-standardised outcome data, with end-of-induction MRD strongly prognostic (6-year DFS 60.2% negative, 45.0% intermediate, 33.8% high).

    PMID 33405950DOI
    What this evidence cannot show

    The paper analysed MRD, not breakpoint characteristics. That patient-specific breakpoint sequences were generated AND archived is inferred from the DNA-based KMT2A MRD methodology and is the single unverified premise of this hypothesis.

    Evidence that argues against this (1)
    • End-of-induction MRD is strongly prognostic and it is also a consequence of everything that preceded it, including breakpoint position if that matters. Using MRD as the outcome against which breakpoint is tested therefore risks conditioning on a mediator, which would attenuate exactly the association this hypothesis is looking for. Statistics in Medicine 2006
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Relapse in KMT2A-rearranged leukaemia is mechanistically heterogeneous, with very early relapse (<9 months) lacking both the drug-response gene mutations and the chemotherapy-exposure signature present in early relapse (9-36 months), implying a pre-existing determinant of outcome present at diagnosis.

    The genomic landscape of relapsed infant and childhood KMT2A-rearranged acute leukemia · Ahlgren L, ... Hagstrom-Andersson AK · Nature Communications 2025
    PMID 41062506DOI
    What this evidence cannot show

    n=36 discovery cohort with small subgroup denominators. The absence of an acquired coding driver in very-early relapse is consistent with a pre-existing genetic determinant such as breakpoint, but equally with an epigenetic or cell-state determinant; this evidence motivates the hypothesis without supporting it specifically.

    Evidence that argues against this (1)
    • Relapse timing categories — under 9 months, 9 to 36 months — are cut-points chosen on a continuous variable, and the absence of a chemotherapy-exposure signature in very early relapse is partly definitional, since those clones had less exposure. The heterogeneity is real; its division into these strata is a choice. Statistics in Medicine 2006
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review

    Breakpoint position determines retained KMT2A domain content, and the graph already records that different breaks plausibly give different biology, including the possibility that some breaks alter only the reciprocal product.

    MLL leukemia and future treatment strategies · Marschalek R · Archiv der Pharmazie 2015
    PMID 25740345DOI
    What this evidence cannot show

    A single-author review stating the Frankfurt group's interpretive position on reciprocal fusions and domain retention; it motivates the mechanistic argument but does not establish it. The claim that different breaks give different clinical biology is exactly what this hypothesis proposes to test, so citing an advocate for it is not independent support.

    Evidence that argues against this (1)
    • This is the graph citing itself. Recording that different breaks plausibly give different biology is a statement of the hypothesis, so using it as evidence for the hypothesis adds no information — and the reciprocal-product possibility it mentions remains untested here. Blood Cells, Molecules and Diseases 2010
  5. E_priorweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    Breakpoint position was tested against outcome in 2013 in a cohort of 545 children with breakpoints localised in 30, reporting that poor-outcome patients preferentially carried breakpoints within KMT2A intron 11.

    The distribution of MLL breakpoints correlates with outcome in infant acute leukaemia · Emerenciano M, Meyer C, Mansur MB, Marschalek R, Pombo-de-Oliveira MS · British Journal of Haematology 2013
    PMID 23432364DOI
    What this evidence cannot show

    Breakpoints were localised in only 30 of the 545 children, so any multivariable model is badly overfitted. Never independently replicated, and confounded with both partner gene and age.

    Evidence that argues against this (1)
    • The authors of the recombinome series, who repeat this claim, state by 2023 that whether breakpoint position affects clinical outcome is yet unclear, and perform no survival analysis of their own in any edition. Meyer C et al. 2023
Connected to