In-silico digest of 81 modelled KMT2A fusion junctions: the peptide that would prove the fusion protein exists
also called junction peptide detectability · fusion-specific tryptic peptide · VDFKQTYSNEVHCVEEILK
A fusion protein shares almost all of its sequence with two normal proteins in the same cell, so the only piece that can prove the fusion itself is present is the short stretch that straddles the join. We computed that stretch for 81 possible KMT2A fusion joins and found it does exist and is measurable - but for the commonest KMT2A::AFF1 architecture the enzyme normally used to chop proteins up cuts exactly at the join, so the proof peptide only appears when that cut is missed.
The same thing, for a scientist
Bottom-up proteomics identifies a protein from its peptides. For a fusion oncoprotein every retained-region peptide is shared with a wild-type parent expressed in the same nucleus, so the only attributable species is a junction-spanning peptide. This analysis builds chimeric coding sequences at the nucleotide level (so phase-1 and phase-2 exon boundaries produce their real hybrid codon rather than a guessed residue) for 81 modelled junctions across KMT2A::AFF1, KMT2A::MLLT3 and the reciprocal AFF1::KMT2A, digests each with six protease specificities at 0-2 missed cleavages, and tests every junction-spanning peptide for mass, length and uniqueness against the full reviewed human proteome (UP000005640, 20,416 sequences). **The headline is that the peptide exists**: 47 of 48 in-frame junctions yield a detectable, proteome-unique tryptic junction peptide within two missed cleavages. **The obstacle is where the cut falls.** Five of the eleven KMT2A coding-exon boundaries in the breakpoint region end in lysine - including residue 1406, the boundary in the only deposited KMT2A::AFF1 fusion protein sequence [E1] - so trypsin cleaves exactly at the junction and the fully-cleaved digest contains a wild-type KMT2A peptide (VDFK) and a wild-type AFF1 peptide (QTYSNEVHCVEEILK) and nothing fusion-specific. The junction peptide for canonical KMT2A::AFF1 is **VDFKQTYSNEVHCVEEILK**, 19 residues, [M+H]+ 2281.12, unique across the human proteome, and obtainable only at one missed cleavage. Glu-C in phosphate buffer recovers it fully cleaved as **FKQTYSNE**, [M+H]+ 1016.47. For KMT2A::MLLT3 the same KMT2A K1406 gives tryptic **VDFKDPNR**, [M+H]+ 990.50, again at one missed cleavage - and Glu-C is useless there because AF9's poly-serine tract produces a 64-residue 6,284 Da peptide. The reciprocal AFF1::KMT2A junction (AFF1 346 | KMT2A 1407) gives a **fully cleaved** tryptic 32-mer at [M+H]+ 3474.62, so by this measure the reciprocal product is easier to see by standard proteomics than the direct fusion.
How the canonical junction was fixed without trusting recall
Exon numbering is a live trap here. RefSeq NM_001197104.2 (MANE Select) has 36 exons; the Meyer/Marschalek DCAL consensus used clinically numbers 38; older papers number lower again. The same physical breakpoint therefore has at least three names, and the graph's own fusion-kmt2a-aff1 entry already carries a note about it.
So the architecture was anchored on a deposited sequence instead. GenBank AAC37520.1 / EMBL L22179.1, 'MLL-AF4 der(11) fusion protein', 2311 aa, conceptually translated from the RS4;11 cell line 1. An exact-prefix match against today's Q03164 fails after 29 residues because the 1993 deposit carries 41 extra residues in its KMT2A half, so the junction was located by 12-mer sliding-window identity. It falls at:
...IPADGVHRIRVDFK | QTYSNEVHCVEEILK...
KMT2A 1-1406 AFF1 347-1210
The independent check is that both of those positions are exactly a RefSeq coding-exon boundary with phase 0 (KMT2A exon 9 ends at protein 1406.00; AFF1 exon 3 ends at protein 346.00). A junction inferred from a 1993 conceptual translation landing on two independent 2024 exon boundaries is not a coincidence.
One further numbering fact fell out and is worth recording: the RefSeq MANE translation of KMT2A is 3,972 aa and UniProt canonical Q03164 is 3,969 aa. The three extra residues (GTE) sit at position ~1604 - downstream of every reported breakpoint - so breakpoint-region numbering is identical in both, and every residue number above 1604 is not.
The lysine problem, stated exactly
Of the eleven KMT2A coding-exon boundaries between protein 1052 and 1565 that could serve as a 5' donor, the last residue is:
| RefSeq exon | last residue | residue | phase | trypsin cuts here? |
|---|---|---|---|---|
| 3 | 1052 | Q | 0 | no |
| 4 | 1111 | D | 1 | no |
| 5 | 1189 | C | 2 | no |
| 6 | 1211 | K | 1 | yes |
| 7 | 1337 | S | 1 | no |
| 8 | 1362 | K | 0 | yes |
| 9 | 1406 | K | 0 | yes |
| 10 | 1444 | E | 0 | no |
| 11 | 1493 | K | 0 | yes |
| 12 | 1525 | W | 0 | no |
| 13 | 1565 | K | 1 | yes |
The KMT2A breakpoint cluster region is lysine-rich, so this is not bad luck at one exon: five of eleven candidate donors put a tryptic site at the junction, and the deposited architecture is one of them. Across all 48 in-frame modelled junctions, 28 (58.3%) have no fully-cleaved tryptic junction peptide.
The consequence is specific and fixable, not fundamental. A standard search with max missed cleavages = 0 cannot find KMT2A::AFF1 no matter how deep the run. A search allowing 1-2 - which most do by default - can, but the fully-cleaved forms dominate the ion current, so the fusion-specific species is a minority product of an already sub-stoichiometric protein.
What to run instead
Three changes each recover the junction, and they are not interchangeable between partners:
- Glu-C in phosphate buffer (cleaves after E and D) gives KMT2A::AFF1 FKQTYSNE at [M+H]+ 1016.47 with zero missed cleavages - a near-ideal PRM target. It fails for KMT2A::MLLT3, where the AF9 poly-Ser tract yields a 64-residue, 6,284 Da peptide that no standard method will see.
- Trypsin with one missed cleavage, searched as a targeted PRM/MRM transition list rather than as a discovery search. For KMT2A::AFF1 the target is VDFKQTYSNEVHCVEEILK ([M+H]+ 2281.12); for KMT2A::MLLT3 it is VDFKDPNR ([M+H]+ 990.50), which is short, basic and well-behaved.
- Chymotrypsin gives KMT2A::AFF1 KQTYSNEVHCVEEIL at [M+H]+ 1791.86 at one missed cleavage - a useful orthogonal confirmation, since a single peptide identification at a junction is exactly the kind of result that needs a second protease.
Every peptide quoted here was tested against the full reviewed human proteome and is absent from it.
The corollary about the reciprocal product
The reciprocal AFF1::KMT2A junction (AFF1 1-346 | KMT2A 1407-3969) produces MPSQSVEEDCEAENVWEMGGLGILTSVPITPR, 32 residues, [M+H]+ 3474.62, proteome-unique, at zero missed cleavages. It is a perfectly ordinary tryptic peptide.
So the protein this field describes as the hardest to attribute - because anti-MLL-C antibodies cannot separate it from wild-type KMT2A after taspase1 cleavage - is, by mass spectrometry, the easier of the two to prove present. The attribution problem is a property of the reagent class, not of the molecule. That is the single most useful thing this computation produced.
What this analysis does not establish
It says a peptide exists and is in range. It does not say anyone would find it in a real sample. Three things stand between the two:
- Abundance. KMT2A and its fusions are low-abundance nuclear proteins; a single junction peptide from an unenriched lysate is far below typical discovery depth. This argues for immunoprecipitation-then-PRM, not for reanalysing existing whole-proteome runs and calling a null informative.
- Patient-specific junctions. The breakpoint varies, so a PRM assay built on KMT2A 1406 | AFF1 347 is wrong for a patient who breaks elsewhere. The 81-junction table exists so that the target can be recomputed per patient from a diagnostic RT-PCR result.
- Modified forms. A phosphorylated junction peptide has a different mass and would be missed by an unmodified-only transition list. Since the motivating question is fusion phosphorylation, the transition list must include the phospho-forms.
| script | analyses/junction-peptide-detectability/run.py |
|---|---|
| status | run |
| result summary | 81 junctions modelled, 48 in frame. 47/48 yield a detectable, proteome-unique tryptic junction peptide within 2 missed cleavages; only 20/48 yield one at zero missed cleavages. The canonical KMT2A::AFF1 junction is not among the 20. |
| detection window used | 600-4600 Da neutral monoisotopic, 6-50 residues |
| proteome background | UniProtKB reviewed human proteome UP000005640, 20,416 sequences, 2,178 candidate peptides checked |
| numbering note | KMT2A RefSeq MANE translation is 3972 aa vs UniProt canonical 3969 aa; the 3-residue insertion is at ~1604, downstream of all breakpoints |
inputs
[ "UniProt Q03164, P51825, P42568", "RefSeq NM_001197104.2, NM_005935.4, NM_004529.4", "UniParc UPI000011F81E (GenBank AAC37520.1)", "UniProtKB reviewed human proteome UP000005640" ]
canonical junction
{
"fusion": "fusion-kmt2a-aff1",
"kmt2a_last_residue": 1406,
"aff1_first_residue": 347,
"context": "...IPADGVHRIRVDFK | QTYSNEVHCVEEILK...",
"chimeric_length_aa": 2270,
"anchor": "GenBank AAC37520.1 (PMID 7689231), both boundaries phase-0 RefSeq coding-exon boundaries"
}junction peptides — the numbers behind the chart
[
{
"fusion": "fusion-kmt2a-aff1",
"protease": "trypsin",
"missed_cleavages": 1,
"peptide": "VDFKQTYSNEVHCVEEILK",
"length_aa": 19,
"monoisotopic_MH_plus": 2281.1169,
"unique_vs_reviewed_human_proteome": true,
"zero_missed_cleavage_peptide_exists": false
},
{
"fusion": "fusion-kmt2a-aff1",
"protease": "Glu-C (V8, phosphate buffer, cleaves after E and D)",
"missed_cleavages": 0,
"peptide": "FKQTYSNE",
"length_aa": 8,
"monoisotopic_MH_plus": 1016.4684,
"unique_vs_reviewed_human_proteome": true,
"zero_missed_cleavage_peptide_exists": true
},
{
"fusion": "fusion-kmt2a-aff1",
"protease": "chymotrypsin",
"missed_cleavages": 1,
"peptide": "KQTYSNEVHCVEEIL",
"length_aa": 15,
"monoisotopic_MH_plus": 1791.8582,
"unique_vs_reviewed_human_proteome": true,
"zero_missed_cleavage_peptide_exists": false
},
{
"fusion": "fusion-kmt2a-mllt3",
"protease": "trypsin",
"missed_cleavages": 1,
"peptide": "VDFKDPNR",
"length_aa": 8,
"monoisotopic_MH_plus": 990.5003,
"unique_vs_reviewed_human_proteome": true,
"zero_missed_cleavage_peptide_exists": false,
"note": "modelled as KMT2A 1-1406 | MLLT3 141-568; Glu-C fails here (64-residue, 6284 Da peptide from the AF9 poly-Ser tract)"
},
{
"fusion": "reciprocal AFF1::KMT2A",
"protease": "trypsin",
"missed_cleavages": 0,
"peptide": "MPSQSVEEDCEAENVWEMGGLGILTSVPITPR",
"length_aa": 32,
"monoisotopic_MH_plus": 3474.6225,
"unique_vs_reviewed_human_proteome": true,
"zero_missed_cleavage_peptide_exists": true,
"note": "modelled as AFF1 1-346 | KMT2A 1407-3969"
}
]kmt2a donor boundaries ending in lysine
[ 1211, 1362, 1406, 1493, 1565 ]
kmt2a donor boundaries examined
[ 1052, 1111, 1189, 1211, 1337, 1362, 1406, 1444, 1493, 1525, 1565 ]
- A negative junction-peptide result would be read as 'the fusion protein is not there', when the likeliest explanation is a search allowing zero missed cleavages. If this assay is used as a pharmacodynamic readout for a menin inhibitor, that failure mode reports drug efficacy where there is none.
- A PRM assay built on one junction is a patient-specific reagent. Deploying it as a general KMT2A-r test would systematically miss the ~60% of patients whose breakpoint is elsewhere, and the misses would not be random with respect to breakpoint subregion - which in this disease correlates with age and outcome.
- If a junction-peptide assay becomes the attribution standard, claims about KMT2A* - the 5'-truncated internal-promoter product - get worse, not better: it has no junction at all and would be scored negative by construction.
- Does the KMT2A::AFF1 junction peptide VDFKQTYSNEVHCVEEILK appear in any already-deposited proteomics dataset from a t(4;11) line? Re-searching public SEM/RS4;11 raw files with 2 missed cleavages and a fusion-augmented database is a zero-wet-lab experiment and nobody has reported doing it.
- Is the junction peptide phosphorylated? Its KMT2A half ends ...IRVDFK and its AFF1 half begins QTYSNE - a tyrosine and two serines within eight residues of the join. A phospho-form would carry the first site ever attributed to a fusion protein rather than inferred.
- Does the lysine-at-the-junction property hold across the other KMT2A partners (MLLT1, MLLT10, ELL, AFDN)? It is a property of the KMT2A donor, so it should - but the partner determines whether Glu-C or chymotrypsin rescues it, and AF9 already shows the rescue can fail.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A complete KMT2A::AFF1 fusion protein sequence derived from the RS4;11 cell line was deposited in 1993 and gives the junction as KMT2A residues 1-1406 joined to AFF1 residues 347-1210 in current canonical numbering.
Acute mixed-lineage leukemia t(4;11)(q21;q23) generates an MLL-AF4 fusion product. · Domer PH, Fakharzadeh SS, Chen CS, et al. · Proc Natl Acad Sci U S A 1993What this evidence cannot show
The deposited protein (GenBank AAC37520.1, 2311 aa, conceptual translation) carries 41 more residues in its KMT2A half than today's Q03164, so the junction is located by sliding-window identity rather than by exact prefix match. It is one cell line's breakpoint, not the population mode; the breakpoint distribution across 1,262 KMT2A::AFF1 cases is a separate question and is held in `fusion-kmt2a-aff1`.
Evidence that argues against this (1)
- The KMT2A breakpoint distribution is broad rather than single-valued, so no one junction is the architecture; roughly 39.5% of KMT2A breaks fall in one breakpoint-cluster subregion and 37.2% in another. A PRM assay built on a single junction is therefore wrong for a large minority of patients, which is why this analysis models 81 junctions rather than one. Meyer C et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Identifying a fusion protein by mass spectrometry requires a peptide that spans the junction, because every other peptide is shared with a wild-type parent; dedicated proteogenomic tooling exists for constructing and searching such peptides.
FusionPro, a Versatile Proteogenomic Tool for Identification of Novel Fusion Transcripts and Their Potential Translation Products in Cancer Cells. · Kim CY, Na K, Park S, et al. · Mol Cell Proteomics 2019What this evidence cannot show
A methods paper, not a KMT2A paper. It establishes that the junction-peptide approach is standard practice in proteogenomics and that the search infrastructure exists - which sharpens rather than softens the absence recorded in `absence-no-kmt2a-fusion-junction-peptide-reported-in-any-mass-spectrometry-dataset`, because the tooling was not the obstacle.
Evidence that argues against this (1)
- Junction-peptide identification from chimeric transcripts is treated in the methods literature as needing deliberate validation rather than as a routine output, which is consistent with the low yield of such identifications in practice and cautions against reading a null in an existing dataset as evidence of absence of the protein. Zahra S et al. 2026
- E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The canonical human KMT2A and AFF1 sequences used to build every chimera here are Q03164 (primary gene name KMT2A, 3969 aa) and P51825 (primary gene name AFF1, 1210 aa), both read back from the UniProt REST record at run time by the script.
What this evidence cannot show
Resolved by accession, not by gene-symbol search, precisely because `gene_exact:` matches synonyms in this API. The script asserts the primary gene name and records the length; it fails rather than proceeds if either is wrong. MLLT3 P42568 (568 aa) is resolved the same way.
Evidence that argues against this (1)
- UniProt canonical and RefSeq MANE Select disagree for KMT2A: the MANE translation of NM_001197104.2 is 3,972 aa against UniProt's 3,969, with three extra residues at ~1604. Any residue number above 1604 in this graph is ambiguous between the two conventions unless the source is stated. NCBI RefSeq 2026
computes the junction-spanning peptide (VDFKQTYSNEVHCVEEILK, [M+H]+ 2281.12) that would attribute a measurement to this protein
tryptic junction peptide VDFKDPNR at one missed cleavage; Glu-C fails because of the AF9 poly-Ser tract
quantifies the obstacle that entry names: retained-region peptides are shared, and the junction peptide needs a missed cleavage
- related to →KMT2A / MLL1 — the gene that breaks
five of eleven candidate 5' donor exon boundaries end in lysine, so trypsin cuts at the junction
AFF1 residue 347 is the first acceptor residue in the deposited RS4;11 fusion
the computation says the peptide is measurable; this entry records that nobody has reported one
- ← depends onThe protein nobody can attribute by antibody is the one mass spectrometry finds most easily
derived from the 81-junction in-silico digest
transitions precomputed by the in-silico digest
the proposed fix is the computed junction peptide run as PRM on the existing IP eluate
- ← analogous toJunction RT-PCR is the one routine assay in this disease that IS fusion-specific - and it is structurally blind to KMT2A*
the protein-level analogue: uniqueness by construction at the join, in peptide rather than amplicon
the mass-spectrometry route to the same attribution problem, which needs no new reagent
extends its single-architecture computation to all three annotated breakpoints
its canonical architecture is the 1406/1407 breakpoint; SEM's reagent matches 1362/1363, so the computed junction peptide is for a different fusion