Oncofusion

start / graph / subtraction-synthesis-how-general

Mechanismweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q11b-subtraction

How general is subtraction across KMT2A fusions? Common enough to matter, too heterogeneous to be one thing, and mostly invisible to the way this project searches

also called subtraction synthesis · gain by subtraction generality · target selection consequences

Across seven KMT2A leukemia partners, losing something contributes in about half — but the losses are of three different kinds, and only one of them can be spotted by looking at the protein sequence. The project's way of hunting for drug targets can see that one kind and is blind to the other two.

The same thing, for a scientist

Systematic audit of seven KMT2A partners for loss-of-function contributions. Result: one partner is subtraction-driven with a positive in-cis restoration result (MLLT10), three are 'both' (ELL, AFDN, KMT2A-PTD), two are addition (MLLT1, EPS15), one is not determinable (MLLT3), and the AFF1 control is confirmed as addition. Revised 2026-09-19 after a counter-argument sweep: in none of the four 'loss' partners has the loss been shown to be REQUIRED for transformation, and each of the three sub-types has a published challenge (Chen 2013, Liedtke 2010, Dorrance 2008). The headline is not the count but the taxonomy: subtraction occurs by discarded module, by sequestration of an endogenous protein, and by silencing of the second allele. Only the discarded-module sub-type is detectable by domain-fate analysis, which is the only method this project's target selection has used, and the two invisible sub-types have as much causal evidence behind them as the visible one.

Go deeper
The answer to the question asked

How general is subtraction? Loss contributes in four of seven audited partners, but it is not one mechanism. Calling all four 'subtraction' is a category error of the same shape as the one this beat was sent to correct.

PartnerVerdictSub-typeLoss shown causal?
MLLT10subtractiondiscarded module (PZP 22-198)yes — in-cis restoration abolishes CALM-AF10 transformation
KMT2A-PTDbothallelic (wild-type allele silenced)yes — restoration is selectively lethal in patient blasts
AFDNbothsequestration (cytoplasmic AF6)partly — loss of function shown, requirement for transformation not
ELLbothsequestration (EAF1) + discarded N-terminal EAF siteno
MLLT3not determinablediscarded module (YEATS 1-138), absent in ~92% of patientsuntested
MLLT1additionnone — YEATS retained in 84%, required for leukemogenesisrefuted
EPS15additionnone demonstrated; premise of EH loss itself disputedrefuted
AFF1 (control)additionnone — SIAH degron at 268-290 retainedrefuted

The frequency is high enough that the retained-half framing cannot be treated as a class property. It is not high enough, and the evidence is not uniform enough, to invert the framing either.

The premise this beat was given, and what happened to it

The dispatch held that MLLT1 and MLLT3 both discard most or all of the YEATS reader, and that if the YEATS parallel held for the two commonest lymphoblastic partners it would not be an edge case but the rule.

The parallel is half right, and the half that is wrong is the one that carried the strategic weight.

  • MLLT3: ~92% of MLL-AF9 patients break at or after AF9 residue 376, excluding YEATS. The sequence fact holds. The causal claim has never been tested.
  • MLLT1: ~84% of MLL-ENL patients break at or before ENL residue 5, so YEATS is retained, and deleting it from MLL-ENL reduces stem-cell frequency and lengthens latency. The reader is a capability the fusion exploits, not a constraint it escapes. This graph's own breakpoint record already said the exon-2 junction retains most of YEATS; the brief contradicted the graph, and the graph was right.

So the load-bearing inference — 'it holds for the two commonest lymphoblastic partners, therefore it is the rule' — fails. AFF1 and MLLT1 are the two commonest lymphoblastic partners, and both are addition.

Two near-identical paralogues, fused to the same 5' partner, break at opposite ends of the protein and the reader is load-bearing in one direction and discarded in the other. The field states this as an open question, not a solved one. Until it is solved, MLLT3 should be recorded as not determinable, and this entry deliberately does so rather than borrowing the MLLT10 result to fill the gap.

The real finding: the search method, not the biology, was the limit

The most consequential result of this beat was not about any partner. It is that the question 'what did the fusion discard?' is the wrong question, or at least only a third of the right one.

AFDN discards nothing functional — the junction sits before RA1 at residue 39 — and is nevertheless one of the clearest loss mechanisms in the set, because the fusion drags endogenous cytoplasmic AF6 into the nucleus and releases RAS. ELL shows the same shape with EAF1 and nuclear speckles. KMT2A-PTD discards nothing at all and has the strongest patient-level loss evidence of any lesion here, because its loss is a silenced allele.

Three substrates, three detection methods:

Sub-typePhysical substrateDetected byFound this cycle
discarded moduleprotein sequencedomain-fate analysisMLLT10, MLLT3 (candidate)
sequestrationsubcellular locationimmunofluorescence of the endogenous proteinAFDN, ELL
allelictranscriptional state of the second alleleallele-specific expressionKMT2A-PTD

The sequestration question had never been asked in this graph. It was asked of two partners in this beat and returned positive for both. That ratio is the finding.

What this means for target selection1 deeper

What does not change. The menin strategy for KMT2A::AFF1 is untouched. Menin binds a retained motif, the interaction is structurally resolved and clinically validated, and the AFF1 control came through the audit intact — the one plausible hidden subtraction (degron loss) was refuted on coordinates. interface-menin-kmt2a-mbm1 and the revumenib programme stand.

What changes.

  1. Retained-module scanning is incomplete by construction, and now the shape of the gap is known. It cannot see sequestration or allelic silencing. Adding the two missing questions to the per-fusion dossier template costs nothing and is the single cheapest improvement available (ana-partner-subtraction-triage).
  2. Partner-matched therapy needs breakpoint-level, not gene-level, stratification. The starkest case is MLLT1: a YEATS inhibitor such as SGC-iMLLT would be given on t(11;19) genotype to roughly one in six patients whose fusion does not contain the target at all. Gene-level partner identity is the wrong resolution.
  3. Loss-driven partners need a triage branch, not a target. When a mechanism is a loss, ask first what kind of loss. A silenced allele is restorable with approved drugs; a deleted domain is not restorable with anything. KMT2A-PTD is actionable for exactly this reason and MLLT10 is not, despite both being subtraction.
  4. The menin-benefit question is now sharper and still unanswered. If subtraction contributes in four of seven partners, menin inhibition — which acts on the retained cassette — addresses only part of the mechanism in those patients. Pooled trial analysis cannot show this. Re-cutting per-partner response depth in AUGMENT-101 could.
A note on how this graph checks itself

Three factual corrections came out of this audit, and all three were of statements already in the graph, not of the outside literature.

  • mech-retained-half-framing-audit states the PZP in-cis add-back has not been done. It was done in 2021, in a paper the graph already cites in a different entry.
  • fusion-kmt2a-ell describes the discarded ELL N-terminus as a Pol II/DSIF-contacting region and lists the retained elongation-stimulatory region among the fusion's modules; the primary literature characterises the discarded region as the dual EAF1/EAF2 docking site and shows the elongation domain to be dispensable for transformation.
  • protein-eps15-domains assigns fate 'lost' to all three EH domains; the founding clinical report says the fusion contains almost all of AF1p, and the EH-lacking architecture is reported elsewhere as a novel variant.

None of these are edited here — they belong to other beats. The pattern is worth naming: the graph's errors are increasingly not missing facts but stale inferences that no later entry re-examined, and two of the three were visible from citations the graph already held.

Counter-argument sweep, 2026-09-19: what the audit itself got wrong

This entry's own method - re-examining conclusions using citations already in reach - has now been turned on it, and it changes three of the eight rows.

RowAs writtenPublished challenge
MLLT10, 'subtraction, loss shown causal'in-cis restoration abolishes CALM-AF10 transformationMLL-AF10 and CALM-AF10 transformation is Dot1l-dependent for initiation AND maintenance, by knockout and by EPZ004777 [E1 counter, Chen 2013]. A sufficient addition mechanism already exists
AFDN, 'both, sequestration'endogenous AF6 pulled from cytoplasm, RAS releasedthe AF6 RA1 domain is the minimal region SUFFICIENT for immortalisation and short-latency leukaemogenesis, Ras-binding-dead mutants still transform, heterologous RA domains and activated Ras fusions do not substitute, and RA1 acts by SELF-ASSOCIATION [E3 counter, Liedtke 2010]
KMT2A-PTD, 'both, allelic, loss shown causal'wild-type allele silenced, restoration selectively lethalMll(PTD/-) mice die at birth; HoxA up-regulation is dominant, present with and without the wild-type allele; increased pluripotent CFU-GEMM occurs only WITH wild-type Mll [E4 counter, Dorrance 2008]

What this does to the headline. 'Loss contributes in four of seven audited partners' becomes 'a loss is documented alongside the transforming mechanism in four of seven, and in none of them has the loss been shown to be required'. That is a materially weaker claim and the entry's strategic recommendations should be read against it.

What survives intact, and it is the part that matters. The claim about the project's SEARCH METHOD is untouched by any of these challenges. Sequestration and allelic silencing are real phenomena whatever their causal weight, and a retained-module scan cannot see either of them. Adding the two missing questions to the per-fusion dossier template still costs nothing, and the recommendation to stratify by breakpoint rather than partner gene is unaffected - indeed it is strengthened, since an independent series found half of its AF9 breakpoints outside the published cluster regions [E2 counter, Langer 2003].

And the self-audit lesson repeats on itself. The Liedtke result was already cited in this entry's own E3 caveat as a tension, and was not promoted into the verdict. That is the same failure mode - a stale inference that no later reader re-examined - that this entry identified in three other entries.

Facts and figures
partners audited7
recommended template changeAdd two questions to every per-fusion dossier: (1) does the fusion relocate any endogenous protein and does that remove a function; (2) is the untranslocated allele of either partner silenced. Both were never asked before this beat; both returned positives the first time they were asked.
cheapest high value next stepAllele-specific KMT2A expression analysis in translocation-driven KMT2A-r blasts from existing transcriptomic cohorts, to test whether allelic subtraction extends beyond PTD
verdict counts noteCounts unchanged, causal weight downgraded 2026-09-19. A loss is documented alongside the transforming mechanism in four of seven partners; in none has it been shown to be required. Challenges: Chen 2013 (PMID 23138183) for MLLT10, Liedtke 2010 (PMID 20395419) for AFDN, Dorrance 2008 (PMID 18617636) for KMT2A-PTD.
verdict counts
subtraction1.000subtraction: 1not_determinable1.000not_determinable: 1control_confirmed_addition1.000control_confirmed_addition: 1addition2.000addition: 2both3.000both: 3
verdict counts — the numbers behind the chart
{
  "subtraction": 1,
  "both": 3,
  "addition": 2,
  "not_determinable": 1,
  "control_confirmed_addition": 1
}
verdicts
{
  "fusion-kmt2a-mllt10": "subtraction",
  "KMT2A-PTD": "both (addition + allelic silencing)",
  "fusion-kmt2a-afdn": "both (addition + sequestration)",
  "fusion-kmt2a-ell": "both (addition + sequestration, discarded-module component untested)",
  "fusion-kmt2a-mllt3": "not determinable",
  "fusion-kmt2a-mllt1": "addition",
  "KMT2A::EPS15": "addition",
  "fusion-kmt2a-aff1": "addition (control, confirmed)"
}
subtraction subtypes
[
  {
    "name": "discarded-module",
    "substrate": "protein sequence",
    "detected_by": "domain-fate analysis",
    "instances": [
      "MLLT10 PZP (causal)",
      "MLLT3 YEATS (untested)"
    ],
    "visible_to_current_project_method": true
  },
  {
    "name": "sequestration",
    "substrate": "subcellular location of the endogenous counterpart",
    "detected_by": "immunofluorescence of the endogenous protein, with fusion silencing as rescue",
    "instances": [
      "AFDN / AF6 (causal for the loss of function)",
      "ELL / EAF1 (documented, untested)"
    ],
    "visible_to_current_project_method": false
  },
  {
    "name": "allelic",
    "substrate": "transcriptional state of the second allele",
    "detected_by": "allele-specific expression",
    "instances": [
      "KMT2A-PTD (causal in patient blasts)"
    ],
    "visible_to_current_project_method": false
  }
]
premise audit
{
  "premise": "MLLT1 and MLLT3 both lose most or all of the YEATS reader, so the subtraction mechanism is the rule rather than an edge case",
  "outcome": "half refuted; MLLT3 upheld as a sequence fact with causality untested, MLLT1 refuted outright (YEATS retained in 84% and required for leukemogenesis)",
  "strategic_consequence": "The two commonest lymphoblastic partners are AFF1 and MLLT1, and both are addition. The inference that subtraction is the rule for lymphoblastic disease does not survive."
}
restoration experiments ever performed
[
  "CALM-AF10 PZP add-back (Klein 2021) — positive",
  "MLL-ENL YEATS deletion, i.e. the reverse direction (Hu 2022) — showed the domain is required, not dispensable",
  "MLL-PTD wild-type allele re-expression via DNMT/HDAC inhibition (Whitman 2005) — selectively lethal",
  "Counter-reading: MLL-AF10/CALM-AF10 Dot1l knockout and EPZ004777 (Chen 2013) - shows the retained OM-LZ/DOT1L addition mechanism is necessary and sufficient for both fusions, which the PZP restoration result must be interpreted on top of"
]
How this could bite us back
  • Two of the three subtraction sub-types have no inhibitor-shaped handle. A more complete mechanistic picture here reduces rather than expands the druggable surface for several partners, and the honest output of this beat is partly a narrowing of the target space.
  • The recommendation to stratify by breakpoint rather than by partner gene is right mechanistically and costly practically: infant KMT2A-r cohorts are already too small, and `ana-target-safety-triage-v2` has already run into a constraint gating on its own statistical power. Stratifying further will make that worse before it makes anything better.
  • If subtraction contributes in most partners but menin inhibitors still work clinically across partners, the mechanistic account and the clinical result are in tension, and the mechanistic account is the one with less evidence behind it. This beat should not be read as predicting that menin inhibitors will fail in subtraction-driven partners; it predicts only that the benefit may differ, and that nobody has looked.
  • This entry's own E3 caveat already recorded the Liedtke tension and the verdict table did not reflect it. An audit that catches stale inferences in other entries while carrying one of its own is exactly the failure it was written to name, and the next reader should assume the same is true of this revision.
Still open
  1. Is the untranslocated KMT2A allele silenced in translocation-driven KMT2A-r leukemia, as it is in PTD? This is answerable from existing transcriptomic cohorts and would move the allelic sub-type from one lesion to a class property.
  2. Do all cytoplasmic-partner MLL fusions (AFDN, EPS15, GAS7, SEPT family) sequester their endogenous counterparts? If sequestration is a structural-class property it can be predicted rather than discovered, and this graph has never asked the question of any of them.
  3. Does menin-inhibitor benefit differ between addition-driven and subtraction-driven partners? Four of seven partners have a subtraction component, menin inhibition acts only on the retained cassette, and every trial analysis to date pools partners.
  4. Why do AF9 and ENL — close paralogues fused to the same 5' partner — break at opposite ends of the protein? Functional selection and recombination-landscape accident predict different things, and separating them requires no new biology, only comparison of MLLT1 and MLLT3 intronic architecture.
  5. How many other conclusions in this graph are stale inferences that a later entry could overturn using citations the graph already holds? Three such corrections came out of this single beat.
  6. Has any loss mechanism in any KMT2A fusion been shown to be REQUIRED for transformation, rather than documented alongside it? After the 2026-09-19 counter sweep the answer appears to be no. If that holds, the correct headline for this beat is about the project's blind spot, not about subtraction being a partner class.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Restoring the PZP domain in cis to CALM-AF10 prevents transformation of HSPCs in vitro and in vivo, abrogates leukemogenesis, downregulates Hoxa genes and reverses cellular transformation, with the authors concluding that disruption of PZP function in the fusion directly leads to transformation.

    The role of the PZP domain of AF10 in acute leukemia driven by AF10 translocations · Klein BJ, Deshpande A, Cox KL, Xuan F, Zandian M, Barbosa K, Khanal S, Tong Q, Zhang Y, Zhang P, Sinha A, Bohlander SK, Shi X, Wen H, Poirier MG, Deshpande AJ, Kutateladze TG · Nature Communications 2021
    PMID 34226546DOI
    What this evidence cannot show

    CALM-AF10, not KMT2A::MLLT10. Verified against PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • The founding member of the discarded-module class is fully explained by an addition mechanism. MLL-AF10 and CALM-AF10 transformation is dependent on Dot1l by conditional knockout and by the selective inhibitor EPZ004777, for in vitro transformation and for in vivo initiation and maintenance, with suppression of Hoxa-cluster and Meis1 expression. The retained OM-LZ recruits DOT1L; DOT1L is necessary. The PZP-restoration result adds a constraint on top of a mechanism that is already sufficient, which is a weaker claim than 'subtraction-driven'. Chen L et al. 2013
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The YEATS domain is retained in 84.1% of MLL-ENL patients and is crucial for MLL-ENL-mediated leukemogenesis, with deletion impairing fusion-protein binding and acetyl-binding point mutations decreasing stem-cell frequency and increasing leukemia latency; approximately 92% of MLL-AF9 patients break at or after AF9 residue 376 and exclude the YEATS domain.

    The ENL YEATS epigenetic reader domain critically links MLL-ENL to leukemic stem cell frequency in t(11;19) Leukemia · Hu H, Saha N, Yang Y, Ahmad E, Lachowski L, Shrestha U, Premkumar V, Ropa J, Chen L, Teahan B, Grigsby S, Marschalek R, Nikolovska-Coleska Z, Muntean AG · Leukemia 2022
    PMID 36435883DOI
    What this evidence cannot show

    The 92% AF9 figure comes from the companion review (Hu & Muntean 2023, PMID 37295550) by the same group, so the two statistics are not independent sources. Both verified against PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • The breakpoint statistics that carry the MLLT1-versus-MLLT3 argument are not independently replicated at the resolution used. An independent series of eleven t(9;11) children and two cell lines placed five patients' AF9 breakpoints OUTSIDE both previously described AF9 breakpoint cluster regions - in the centromeric region of intron 4, and in one case in intron 3. Nearly half of that small series therefore does not sit where the published architecture says it should. The '~92% break at or after AF9 residue 376' and '~84% at or before ENL residue 5' figures come from one group's review of reported breakpoints, and this entry already notes they are not independent of each other; they are also not independent of the breakpoint-calling conventions of the era in which those reports were written. Langer T et al. 2003
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-AF6-rearranged cells relocate endogenous AF6 from the cytoplasm to the nucleus with consequent RAS activation, and silencing the fusion restores cytoplasmic AF6 and significantly reduces RAS-GTP levels — a loss of function achieved by relocation rather than deletion.

    MLL-AF6 fusion oncogene sequesters AF6 into the nucleus to trigger RAS activation in myeloid leukemia · Manara E, Baron E, Tregnago C, Aveic S, Bisio V, Bresolin S, Masetti R, Locatelli F, Basso G, Pigazzi M · Blood 2014
    PMID 24695851DOI
    What this evidence cannot show

    Does not establish that sequestration is required for transformation, and sits in tension with Liedtke et al. 2010 showing that the fusion's own Ras binding is dispensable for oncogenesis. Verified against PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • The AFDN/AF6 sequestration positive - the single result that made question Q2 look productive - has a published alternative explanation that is an ADDITION, not a loss. Liedtke et al. identified the AF6 RA1 domain as the MINIMAL REGION SUFFICIENT for MLL-AF6-mediated myeloid progenitor immortalisation in vitro and short-latency leukemogenesis in vivo; single amino-acid substitutions known to abolish the AF6-Ras interaction did NOT abrogate MLL-AF6 oncogenesis; fusing MLL to heterologous RA domains of c-Raf1 or RalGDS, or directly to constitutively active K-RAS, H-RAS or RAP1, was not sufficient; and RA1 was shown to mediate efficient SELF-ASSOCIATION, leading the authors to conclude that constitutive MLL self-association is a more common pathogenic mechanism than previously appreciated. On that account the transforming event is oligomerisation of the MLL moiety, and the relocation of endogenous AF6 to the nucleus is a side effect of expressing a nuclear AF6 fragment rather than the mechanism. Both observations can be true; only one of them is shown to be required for transformation, and it is not the sequestration. Liedtke M et al. 2010
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL(PTD/WT) AML blasts express the PTD allele but not the wild-type allele; silencing is reversed by DNMT and HDAC inhibitors and wild-type MLL induction is associated with selective sensitivity to cell death, supporting a recessive gain-of-function model.

    The MLL partial tandem duplication: evidence for recessive gain-of-function in acute myeloid leukemia identifies a novel patient subgroup for molecular-targeted therapy · Whitman SP, Liu S, Vukosavljevic T, Rush LJ, Yu L, Liu C, Klisovic MI, Maharry K, Guimond M, Strout MP, Becknell B, Dorrance A, Klisovic RB, Plass C, Bloomfield CD, Marcucci G, Caligiuri MA · Blood 2005
    PMID 15774615DOI
    What this evidence cannot show

    DNMT and HDAC inhibitors are pleiotropic, so the attribution to wild-type MLL specifically is not isolated. Verified against PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • The 'recessive gain-of-function' reading of MLL-PTD, which is the sole basis of the allelic sub-type, is contradicted in the genetically cleanest test of it. Dorrance et al. generated mice expressing Mll-PTD in the ABSENCE of wild-type Mll and compared them with Mll(PTD/WT). Mll(PTD/-) mice die at birth, unlike Mll(PTD/WT), Mll(WT/-) and Mll(WT/WT) animals, so removing the wild-type allele is not simply permissive. Both Mll(PTD/-) and Mll(PTD/WT) fetal liver cells showed increased HoxA expression and increased CFU-GM, which the authors read as the PTD acting in a DOMINANT gain-of-function fashion; and only Mll(PTD/WT) cells showed increased pluripotent CFU-GEMM progenitors, i.e. in selected tissues the PTD REQUIRES cooperation with wild-type Mll to produce the abnormality. Wild-type-allele silencing in patient blasts is a real observation, but the inference that the silencing is what makes the lesion oncogenic does not follow, and in the compartment that matters most for stemness the wild-type allele appears to be needed rather than dispensable. Dorrance AM et al. 2008
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The Siah-binding degron of Af4 maps to residues 268-290 (consensus PxAxVxP; robotic-mouse mutation V280A reduces Siah-1a affinity ~15-fold and abolishes Siah-mediated degradation), which lies far downstream of the AFF1 exon-4 fusion junction at approximately residue 53 and is therefore retained in KMT2A::AFF1.

    Mediation of Af4 protein function in the cerebellum by Siah proteins · Oliver PL, Bitoun E, Clark J, Jones EL, Davies KE · Proceedings of the National Academy of Sciences USA 2004
    PMID 15459319DOI
    What this evidence cannot show

    Mouse Af4 numbering, cerebellar context. The junction position at residue ~53 is computed from the Ensembl canonical AFF1 transcript CDS structure. The ~215-residue margin is far larger than any plausible alignment or isoform error. Verified against PubMed on 2026-09-18. CORRECTED 2026-09-24: the coordinates 268-290 are MURINE Af4. The human homologue is P51825 282-304 (22/23 identical, motif PTAYVRP 290-296) and the KMT2A::AFF1 junction is at human residue 347, so the degron is N-terminal to the junction and LOST, not retained.

    Evidence that argues against this (2)
    • The degron arithmetic settles a question about the forward fusion, and there is published evidence that the forward fusion is not where this particular mechanism lives. Bursen et al. transduced purified Lin-/Sca1+ progenitors and found that AF4.MLL alone induced pro-B ALL, B/T biphenotypic acute leukaemia or mixed-lineage leukaemia after transplant, whereas MLL.AF4 alone caused no disease over 13 months; they concluded that t(4;11) leukaemia may rest on two oncoproteins. The SIAH/degron hypothesis was formulated about the reciprocal product, so demonstrating that KMT2A::AFF1 retains the degron leaves the stability question open rather than closing it, and the audit's 'AFF1 control holds' verdict should be read as 'holds for the forward fusion', which is a narrower statement than the synthesis makes. Bursen A et al. 2010
    • The inference drawn from this evidence in the synthesis table - that the AFF1 subtraction hypothesis is refuted because the degron is retained - is withdrawn. Junction fixed at AFF1 residue 347: QTYSNEVH occurs exactly once in UniProt P51825 (1,210 aa), and the corpus independently reconstructed the junction peptide VDFK|QTYSNEVHCVEEILK from the siMA6 siRNA, recording that both KMT2A::AFF1 chimeras resume at QTYSNEVHCVEEILK so the AFF1 acceptor is constant across breakpoints. Degron located in human numbering by homology: mouse Af4 268-290 (KSAAMQQKPTAYVRPMDGQDQAP) aligns 22/23 to human P51825 282-304 (KSVAMQQKPTAYVRPMDGQDQAP), offset +14; the PxAxVxP motif PTAYVRP is unique in human at 290-296 and the robotic-equivalent valine is V294. The element ends at 304, 43 residues N-terminal to the junction. UniProtKB 2026
Connected to