Hypothesis: genetic and non-genetic escape from menin inhibition are substitutes, so MEN1 mutations will be rare in infants and the field's resistance biomarker will fail in children
A leukemia cell escapes a drug by whichever route is cheapest. If it can simply change its behaviour, it never needs to mutate. Infant leukemias carry almost no mutations, so they should escape by changing behaviour, and the genetic test the field relies on will come back clean while the child relapses.
The same thing, for a scientist
Resistance routes to menin inhibition are treated in the literature as an additive list. We hypothesise they are competing substitutes under selection, so their frequencies are negatively correlated across populations and models. The decision-relevant corollary: in KMT2A-rearranged infant ALL, whose genome is exceptionally mutationally silent [E3], MEN1 interface mutations should account for a substantially smaller share of menin-inhibitor failures than the 39% reported in adults post-revumenib [E1], while non-genetic and cell-state escape should account for a correspondingly larger share. Serial patient-derived xenografts from pretreated KMT2A::AFF1 ALL are already resistant with sustained on-target drug effect and without menin exposure [E2], which is what the substitution logic predicts. This is our inference; no study has measured resistance mechanism share in children.
The prediction that would change practice this year
MEN1 sequencing at relapse is becoming the standard reflex, and a negative result is being read as evidence that a differently-binding inhibitor or a re-challenge might work. If MEN1 mutation is rare in children for structural reasons rather than because the relapse is early, then a negative MEN1 result in a child carries almost no information, and treating it as reassuring is a clinical error.
This is checkable with samples that already exist. Paediatric patients treated on AUGMENT-101 and on the paediatric menin trials now open have banked relapse material. The question — what fraction of paediatric menin failures carry a MEN1 interface mutation — requires sequencing, not new patients, and the graph already records it as the single most decision-relevant unknown in the resistance dossier.
Distinguishing substitution from the trivial explanation
Infants have fewer mutations of every kind, so of course they will have fewer MEN1 mutations. The substitution hypothesis says something stronger: the share of resistance attributable to genetic escape falls faster than overall mutational burden does, because the non-genetic route absorbs the displaced demand. Concretely, if infant mutational burden is one tenth of adult and MEN1 mutations account for 39% of adult failures, the trivial account predicts roughly 4-6% in infants after accounting for the selection advantage that makes a driver mutation far more likely to be observed than a passenger. The substitution account predicts a lower figure still, together with a measurably higher rate of documented transcriptional or cell-state escape in the same samples.
Those two predictions are close together and separating them needs a decent sample size. Stating them separately matters anyway, because the trivial account alone already carries the clinical implication, and the clinical implication does not depend on the more interesting mechanism being true.
The second, harder claim: estimating the matrix
Beyond the paediatric prediction, the hypothesis asserts that the full route-by-route correlation structure is estimable and non-trivial — that PRC1.1 loss and MEN1 mutation co-occur less often than chance, that lineage switch and chromatin-layer escape share a precursor state and co-occur more often than chance, and that the resulting matrix ranks combination partners differently from the current layer-based heuristic.
We are markedly less confident about this part. Co-occurrence analysis in resistance screens is underpowered and confounded by screen design, and negative results will be uninterpretable. It is included because the matrix is the right object and because building it, even badly, makes explicit an assumption the field is currently making implicitly.
| statement | Genetic (MEN1 interface mutation) and non-genetic (cell-state, PRC1.1, transcriptional rewiring) escape from menin inhibition are competing substitutes rather than independent additive routes; consequently MEN1 interface mutations will account for a substantially smaller share of menin-inhibitor failures in KMT2A-rearranged paediatric and infant leukemia than the roughly 39% reported in adults after revumenib, with a correspondingly larger share of documented non-genetic escape in the same samples. |
|---|---|
| origin | analogy |
| source analogy | analogy-finance-portfolio-covariance-of-resistance |
| prior probability | 0.55 |
| prior reasoning | That MEN1 mutations will be less common in paediatric menin failures we put high, near 0.75, since it follows from mutational burden alone. That the share falls faster than burden does, which is the substitution claim proper, we put at about 0.4. That the full route correlation matrix is estimable and usefully non-trivial we put lowest, near 0.25, because the co-occurrence data are thin and confounded. The composite reflects the hypothesis as stated, which leads with the paediatric prediction. |
| decision value | Determines whether MEN1 sequencing is a usable resistance biomarker in the paediatric population this project serves, and whether a negative MEN1 result at relapse in a child should be treated as informative. That is an immediate clinical-interpretation question, not a long-term research one. |
predictions
[ "Sequencing of paediatric and infant menin-inhibitor failures finds MEN1 interface mutations in well under 20% of cases, against roughly 39% in adults after revumenib.", "The same paediatric samples show a higher rate of documented transcriptional or cell-state escape than adult samples, scored by a common signature.", "Across published CRISPR resistance screens, MEN1-mutant and PRC1.1-deficient resistance hits are mutually exclusive more often than chance predicts.", "In models, forcing cells into a low-plasticity state raises the observed rate of MEN1 mutation under drug pressure — the substitution mechanism run in reverse.", "A correlation-matrix-based combination ranking differs from the current layer-based ranking for at least one clinically available partner." ]
experiments
[ "exp-men1-mutant-fitness-cost-competition" ]
analyses
[ "ana-resistance-route-cooccurrence", "ana-menin-resistance-cell-state-signature" ]
falsifiers
[ "MEN1 interface mutation frequency in paediatric menin failures is similar to adults, showing that drug pressure selects the mutation efficiently regardless of baseline mutational burden.", "Genetic and non-genetic escape co-occur in the same samples more often than chance, making them complements rather than substitutes.", "Across resistance screens, route hits show no departure from independence in either direction.", "A correlation-matrix ranking reproduces the layer-based ranking exactly, in which case the extra machinery buys nothing." ]
- If children escape non-genetically, then every strategy in this graph built on anticipating and out-designing a resistance mutation is aimed at the adult version of the disease.
- A rare biomarker is worse than no biomarker if a negative result is read as reassurance. Publishing a low paediatric MEN1 mutation rate without simultaneously offering a non-genetic assay would leave clinicians with less usable information than they have now.
- The substitution logic implies there is always a cheapest route, so closing routes one at a time may never reduce the total escape rate, only change its composition.
- How many paediatric menin-inhibitor failures have been sequenced for MEN1 anywhere, and is the number already large enough to answer this without new sample collection?
- Does the substitution logic also predict that infants escape CD19-directed therapy by lineage switch rather than by CD19 mutation, which would be a second, independent test of the same idea on data that already exists?
- If plasticity pre-empts mutation, does successfully blocking plasticity convert a non-genetic escape problem into a genetic one, and is that a better problem to have?
- ← related toAdaptive, non-genetic resistance — the leukemia keeps the fusion and rewires how it is maintained
substitution framing weakened: interface mutants arise de novo in unbiased base-editor screens, so a quiet diagnostic genome constrains standing variation rather than resistance-mutation supply under selection
generated by treating resistance routes as substitutes in a portfolio
the paediatric prediction rests on the silent infant mutational landscape
predicts MEN1 sequencing fails as a resistance biomarker in children
- supports →Adaptive, non-genetic resistance — the leukemia keeps the fusion and rewires how it is maintained
predicts non-genetic escape dominates in infants
builds the route-by-route co-occurrence matrix and the assessment-completeness table
supplies the shared non-genetic escape signature needed to score route 4
if genetic and non-genetic escape are substitutes, a perfect mutation map predicts the wrong failure mode in infants
the expanded route list makes the substitution prediction testable across more than four routes