Oncofusion

start / graph / ana-q204-the-ten-reported-menin-inhibitor-escape-substitutions-are-all-one-nucleotide-away-which-the-sources-could-not-have-shown-otherwise

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.king

The ten reported menin-inhibitor escape substitutions are each one nucleotide change from the MEN1 sequence - which the sources could not have shown otherwise - so the useful outputs are a 25-variant validation panel and a burden threshold, not the headline

also called MEN1 escape mutational reachability · menin inhibitor escape validation panel

Every drug-resistance change reported so far in the drug target, menin, needs only one letter of DNA to change. That sounds alarming, but it is partly built in: the lab method used to find these changes can only make one-letter changes, and patients' cells change one letter at a time. So it does not show that two-letter escapes do not exist. What it does give is a list of 25 other one-letter changes at the same spots that nobody has tested, which any new drug should be tried against, and a rough rule for when a tumour is large enough that some resistant cells are likely to be there before treatment starts.

The same thing, for a scientist

From the canonical MEN1 coding sequence (Ensembl ENST00000450708, asserted to translate to UniProt O00255, 610 aa) and the reported escape set - M327I/V/T and G331D shared across five inhibitors, C334R, E368K/V and V372A inhibitor-specific (PMID 42103719), with T349M, M327I, G331R and G331D described as patient mutants (PMID 39276940) - all ten substitutions are reachable by a single nucleotide change, through twelve distinct routes (eight transitions, four transversions). The literature residue numbers are the 615-aa isoform; the +5 offset to canonical numbering was accepted only because it is the single offset in 0-10 that places the expected wild-type residue at all six sites. The single-nucleotide result is partly true by construction: base-editor screens create single-nucleotide substitutions and patient mutations arise one step at a time, so neither source could have shown a two-nucleotide escape. The informative outputs are (1) 25 single-nucleotide missense neighbours at the same six sites that are unreported, as a validation panel for any new inhibitor, and (2) an assumption-driven burden threshold: in a pure-birth expansion the expected number of independent escape events is S*u*(N-1) with S=12 routes, which exceeds one at N of about 8e7 cells for an assumed per-site per-division rate u=1e-9. The rate is swept, not measured.

Go deeper
What was computed

Each of the six escape sites was read as its codon in the canonical CDS, and all nine single-nucleotide neighbours of that codon were translated 1. The six sites and codons: M327 (ATG), G331 (GGC), C334 (TGT), T349 (ACG), E368 (GAG), V372 (GTA) in literature numbering 2. Every reported substitution appears among the neighbours. The remaining missense neighbours - 25 across the six sites - have not been reported as escapes in the sources read 1. Whether any of them confers resistance is unknown; none was tested here.

Why the single-nucleotide result is not a finding about biology

A base-editor screen can only introduce single-nucleotide changes, so a two-nucleotide escape cannot appear in it. Patient mutations arise one step at a time, so a two-step escape needs two independent events and is correspondingly rarer. Both sources therefore select a set that is single-nucleotide by design or by probability. The right reading is that no mutational barrier is known to protect any of the reported escapes, not that none exists for escapes outside the editor's reach 3.

The burden threshold, with its assumption stated

If an expansion to N cells takes N-1 divisions and each division mutates a specific nucleotide at rate u, the expected number of independent events producing any of the 12 escape routes is 12u(N-1). It reaches one at N of about 8.3e8 (u=1e-10), 8.3e7 (u=1e-9) and 8.3e6 (u=1e-8). The rates are a sweep and not a measurement; survival of a resistant clone (drift, fitness cost, niche) is not modelled, so this is an upper bound on pre-existence. A leukaemia that mutates far less than this range would over-call it 4. Separately, the corpus already records that the relapse clone in infant KMT2A-rearranged ALL can persist as a minor subclone, so low-frequency clones are the expected state, not an exception.

Facts and figures
scriptanalyses/q204-failure-mode-atlas/escape_reachability.py
outputanalyses/q204-failure-mode-atlas/escape_reachability.json
transcriptENST00000450708
observed escape substitutions10
needing one nt change10
distinct single nt routes12
transitions8
transversions4
unreported single nt missense at the six sites25
offset controlonly +5 of offsets 0-10 places the expected residue at all six sites
rate u is assumedtrue
N at which expected escape events equal one
{
  "u=1e-10": "8.3e8",
  "u=1e-9": "8.3e7",
  "u=1e-8": "8.3e6"
}
How this could bite us back
  • A next-generation inhibitor tested only against the ten known substitutions will look escape-proof against a set that was selected by the previous drug. Test it against the 25 neighbours and the two-nucleotide variants of the contact residues before the clinic.
  • Starting a single-escape-vulnerable agent as monotherapy at bulk burden invites a pre-existing escape clone if the assumed rate holds. Starting it at low burden or with an independent second leg reduces that, but a second leg protects only if the two lesions are independent, and one lesion that touches both legs removes the benefit.
  • If infant leukaemia mutates far less than the swept range, this logic over-calls pre-existing resistance and could push toward unnecessary combination toxicity in infants.
Still open
  1. Which of the 25 unreported single-nucleotide neighbours confer resistance? Saturation editing (base or prime) of the six sites against each clinical inhibitor would answer it.
  2. Do two-nucleotide substitutions at the contact residues confer resistance, which the existing screens could not generate?
  3. What is the per-site per-division mutation rate in infant KMT2A-rearranged ALL specifically? The threshold moves 100-fold across the swept range.
  4. Is an escape clone present at diagnosis in patients? The duplex-sequencing experiment in the corpus (exp-duplex-clone-tracking-under-menin-inhibition) would settle it for cooperating lesions, and the same assay should include the MEN1 pocket sites.
Every source behind this entry (4)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    All ten reported menin-inhibitor escape substitutions (M327I/V/T, G331D/R, C334R, T349M, E368K/V, V372A) are reachable from the canonical MEN1 codons by a single nucleotide change, through twelve distinct routes; 25 further single-nucleotide missense neighbours at the same six sites are unreported.

    What this evidence cannot show

    The escape set is the union of what two papers report; sites never selected under drug are not in it, so the 25 neighbours are only those at the six sites already known. The arithmetic is deterministic given the CDS; the claim about resistance applies only to the substitutions the sources report.

    Evidence that argues against this (1)
    • The sources cannot show a two-nucleotide escape if one exists, because base-editor screens create single-nucleotide changes by design. The single-nucleotide property is therefore partly a property of the screen.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The +5 offset between literature menin numbering and canonical (610-aa) numbering is the only offset in 0-10 that places the expected wild-type residue (M327, G331, C334, T349, E368, V372) at all six sites.

    What this evidence cannot show

    Confirms a known numbering convention (the corpus already holds this as a concept); adds the control, not a new fact.

    Evidence that argues against this (1)
    • The control tests six residues. It shows the +5 offset is right at those sites, not that every paper uses one convention for every residue; a paper numbering from a different isoform would break it silently, so each new residue should be checked the same way.
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CRISPR base-editor screening identified shared resistance mutations M327I/V/T and G331D and inhibitor-specific C334R, E368K/V and V372A across five menin inhibitors, and orthogonal in vitro and in vivo mutation generation under therapeutic pressure suggests these arise and affect all menin inhibitors.

    PMID 42103719DOI
    What this evidence cannot show

    Abstract read only (2026-10-14). The editors used and their editing windows were not read, so what the screen could not reach was not characterised beyond the general fact that base editors make single-nucleotide changes.

    Evidence that argues against this (1)
    • Screen coverage is limited to what the editors can reach; absence of other substitutions from the hit list is not evidence that they are tolerated.
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Menin patient mutants T349M, M327I, G331R and G331D occur adjacent to the MLL1 binding site and do not affect MLL1 binding, they generate steric clash with menin inhibitors, and G331D slows the dissociation of MLL1 from menin.

    Drug-resistant menin variants retain high binding affinity and interactions with MLL1 · Ray J, Clegg B, Grembecka J, Cierpicki T · J Biol Chem 2024
    PMID 39276940DOI
    What this evidence cannot show

    Abstract read only (2026-10-14). 'Might be particularly difficult to inhibit' is the authors' hedged inference. The burden threshold in this entry is a separate modelling step, not from this paper.

    Evidence that argues against this (1)
    • The binding and dissociation measurements are in vitro on an N-terminal MLL1 fragment, not on the full fusion in a cell; a slow off-rate in vitro does not by itself show that occupancy or drug response changes in a patient.
Connected to