Oncofusion

start / graph / ana-the-surfaceome-screen-for-kmt2a-rearranged-leukaemia-was-published-in-2022-and-this-graph-already-cites-it

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q138-kmt2a-aff1-surfaceome

The screen this beat was dispatched to run 'because it has never been run' was published in Haematologica in 2022, used the same instrument, and is cited in this graph already - in an entry filed under innate immunity

also called the surfaceome screen was already run · LAMP5 surfaceome precedent

This project asked for a search of every surface protein in KMT2A-rearranged leukaemia, on the grounds that nobody had done one. Somebody had: a 2022 paper filtered leukaemia expression data through exactly the same published surface-protein list and found LAMP5 - and this project already quotes that paper, but filed under inflammation rather than under targets, so nobody looking for antigens could find it.

The same thing, for a scientist

The dispatch brief for this beat stated that a fusion-specific surface-target screen for KMT2A::AFF1 'has never been run'. **It has.** Sarno and colleagues compiled a lineage-independent MLL-rearranged gene signature from public repositories and **filtered it through the in silico human surfaceome** - the Bausch-Fluck PNAS 2018 resource [E3] - producing a ranked list of predicted cell-surface proteins overexpressed in MLL-rearranged leukaemia. The top result, **LAMP5**, was shown to be a **direct target of the oncogenic MLL fusion protein**, and its depletion inhibited leukaemia growth in vitro and in vivo [E1]. That is the same design this beat was asked to invent, with the same instrument, three years earlier. **And this graph cites the paper**: `cell-state-innate-immune-lamp5` quotes the surfaceome-filtering sentence verbatim in its own summary [E2]. The finding was in the corpus and was unreachable from the antigen framing, because it sits under innate-immune signalling.

Go deeper
What the 2022 screen actually did, in its own terms

From the abstract, read in full: 'In order to identify novel targets for immunotherapy treatments, we compiled a lineage-independent MLL-r leukemia gene signature using publicly available data sets. Data from large leukemia repositories were filtered through the in silico human surfaceome, providing a list of highly predicted cell surface proteins overexpressed in MLL-r leukemias. LAMP5, a lysosomal associated membrane protein, is expressed highly and specifically in MLL-r leukemia. We found that LAMP5 is a direct target of the oncogenic MLL-fusion protein.' 1

Every element the brief asked for is present: surfaceome as the entry point, MLL-rearranged expression as the filter, immunotherapy targets as the stated purpose, and a demonstrated fusion-target relationship rather than a correlation.

Why the corpus could not see it, which is the reusable part

cell-state-innate-immune-lamp5 is an excellent entry and it is filed as a cell state. Its title is about innate-immune rewiring and NF-kB versus interferon. The surfaceome sentence is in its summary_expert, and the word 'antigen' does not appear in its id, title or tags. A beat inventory keyed on surface, antigen, immunotherapy, surfaceome, membrane returned 38 entries and did not return it 4.

This is the shape this project has recorded before as a gap list cannot see what was decided. The corpus held the refutation of its own gap claim, in an entry written for a different purpose, and the defect was in the retrieval path rather than in either entry. The fix is cheap and is applied here: the entry is now edge-linked to the antigen cluster, so the next antigen beat reaches it.

What the 2022 screen did NOT do, which is where this beat's contribution actually lies

It did not apply a normal-tissue filter. The abstract's specificity claim - 'expressed highly and specifically in MLL-r leukemia' - is specificity against other leukaemias, which is the axis their data could address. Nothing in it asks where LAMP5 sits in a normal child.

Measured here against the HPA consensus normal-tissue panel, LAMP5 reaches 297.8 nTPM in basal ganglia 5. That is the contribution this beat can legitimately claim: not a new screen, but the axis neither the 2022 screen nor the 2024 NG2 paper applied. A beat should not restate a published screen as its own; it should say what the published screen left open.

The bounded absence, both symbols, with the querytranslations

Searched 2026-09-23 with the project's safe esearch, which raises on an empty translation rather than returning a false zero:

queryquerytranslationcount
surfaceome[tiab] AND KMT2A[tiab]"surfaceome"[Title/Abstract] AND "KMT2A"[Title/Abstract]0
surfaceome[tiab] AND MLL[tiab]"surfaceome"[Title/Abstract] AND "MLL"[Title/Abstract]1
surfaceome[tiab] AND leukemia[tiab]"surfaceome"[Title/Abstract] AND "leukemia"[Title/Abstract]15
surfaceome[tiab] AND leukaemia[tiab]"surfaceome"[Title/Abstract] AND "leukaemia"[Title/Abstract]2

The single record under the legacy MLL operand is the paper that refutes the brief, and the modern KMT2A operand returns zero. A beat that had queried only KMT2A would have written a confident, false bounded absence. This is the third consecutive cycle in which the legacy operand carried the load-bearing paper 6.

Facts and figures
scriptanalyses/kmt2a-aff1-surface-target-screen/run.py
statusrun
result summaryThe KMT2A-r surfaceome screen was published (Haematologica 2022, LAMP5) using the same instrument; this graph already cited it under a cell-state id. The brief's 'never been run' premise is false. The unapplied axis is normal tissue, not the surfaceome.
querytranslations
surfaceome[tiab] AND KMT2…0.000surfaceome[tiab] AND KMT2A[tiab]: 0surfaceome[tiab] AND MLL[…1.000surfaceome[tiab] AND MLL[tiab]: 1surfaceome[tiab] AND leuk…2.000surfaceome[tiab] AND leukaemia[tiab]: 2surfaceome[tiab] AND leuk…15.0surfaceome[tiab] AND leukemia[tiab]: 15
querytranslations — the numbers behind the chart
{
  "surfaceome[tiab] AND KMT2A[tiab]": {
    "translation": "\"surfaceome\"[Title/Abstract] AND \"KMT2A\"[Title/Abstract]",
    "count": 0
  },
  "surfaceome[tiab] AND MLL[tiab]": {
    "translation": "\"surfaceome\"[Title/Abstract] AND \"MLL\"[Title/Abstract]",
    "count": 1
  },
  "surfaceome[tiab] AND leukemia[tiab]": {
    "translation": "\"surfaceome\"[Title/Abstract] AND \"leukemia\"[Title/Abstract]",
    "count": 15
  },
  "surfaceome[tiab] AND leukaemia[tiab]": {
    "translation": "\"surfaceome\"[Title/Abstract] AND \"leukaemia\"[Title/Abstract]",
    "count": 2
  }
}
How this could bite us back
  • Correcting a brief's premise is only useful if the correction itself is checkable. The querytranslations are pasted in full for that reason; a reader who suspects this entry is over-claiming can re-run all four queries.
  • This entry risks reading as a criticism of cell-state-innate-immune-lamp5, which would be wrong. That entry is accurate and well-cited; the failure is that nothing connected it to the antigen question, and the remedy is an edge, not a rewrite.
  • The brief that carried the false premise also carried a measured, correct inventory of the corpus's antigens. A beat that treats a whole brief as unreliable because one premise failed will discard the parts that were right.
Still open
  1. How many of this corpus's other stated gaps are refuted by an entry it already holds under a different type prefix? The LAMP5 case was found by accident while reading a citation-reuse report, not by any systematic check.
  2. What are the other fourteen records under surfaceome[tiab] AND leukemia[tiab]? At least one may be a second MLL-relevant screen, and none was read on this beat.
  3. Should an entry that reports a screen carry a required field naming the screen's design, so that a screen reported inside a cell-state entry is still retrievable by a beat looking for screens?
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A published screen filtered large leukaemia expression repositories through the in silico human surfaceome to obtain predicted cell-surface proteins overexpressed in MLL-rearranged leukaemia, identified LAMP5 as highly and specifically expressed there, showed LAMP5 is a direct target of the oncogenic MLL fusion protein, and showed LAMP5 depletion inhibits leukaemia growth in vitro and in vivo.

    PMID 33910331DOI
    What this evidence cannot show

    MLL-rearranged as a class, not KMT2A::AFF1 specifically - the signature is explicitly lineage-independent and partner-agnostic. Abstract read in full; the supporting figures were not read on this beat, so the ranked surfaceome list itself has not been compared with this beat's fifteen.

    Evidence that argues against this (1)
    • The screen's specificity claim is against other leukaemias, not against normal tissue. LAMP5 at 297.8 nTPM in basal ganglia is the axis it did not address. Human Protein Atlas 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    This graph's entry cell-state-innate-immune-lamp5 already cites PMID 33910331 and states in its own summary that filtering leukaemia expression repositories through the in silico human surfaceome identified LAMP5 as a direct fusion target.

    What this evidence cannot show

    A statement about this repository's own contents on one date. The entry was read directly, not inferred from an index.

    Evidence that argues against this (1)
    • The entry is not wrong and is not mis-scoped; a cell-state filing for a signalling mechanism is defensible. The defect is that no edge or tag connected it to the antigen cluster, which is a graph-topology problem rather than an entry problem. Haematologica 2022
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The in silico human surfaceome is a published resource, the instrument the 2022 MLL-rearranged screen used.

    The in silico human surfaceome. · Proc Natl Acad Sci U S A 2018
    PMID 30373828
    What this evidence cannot show

    A machine-learning prediction over sequence and annotation features, not an experimental surface-capture census. Title verified by efetch; note that the phrase does not retrieve the paper under a tiab search for the exact quoted title, so it must be found by author or by the single-word token.

    Evidence that argues against this (1)
    • Because both the published screen and this beat's screen rest on computational surface prediction (theirs on this resource, this beat's on UniProt topological annotation), a shared false negative would look like independent confirmation of an absence. UniProt 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    A keyword inventory of this corpus over surface, antigen, immunotherapy, surfaceome and membrane returned 38 entries and did not return cell-state-innate-immune-lamp5.

    What this evidence cannot show

    The inventory tool matches on id OR folder, not on full text, and its own output warns that it is a match and not a census. A full-text grep for 'surfaceome' across node files would have found the entry; the point is that the tool a beat is instructed to use did not.

    Evidence that argues against this (1)
    • The tool prints an explicit warning that it under-returns classes where it matched under half, and it did so here for ana- and gene- prefixes. A beat that heeded that warning and widened the keywords might have found it. Oncofusion Knowledge Graph 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    LAMP5 reaches 297.8 nTPM in basal ganglia in the HPA consensus normal-tissue panel, its highest value outside the lymphohaematopoietic tissues.

    What this evidence cannot show

    Adult post-mortem bulk brain tissue mRNA. LAMP5 is named for brain-and-dendritic-cell-associated LAMP (BAD-LAMP), so high neural expression is expected from the gene's own nomenclature and is not a surprise finding.

    Evidence that argues against this (1)
    • Brain expression does not by itself preclude a CAR-T target if the antigen is not surface-accessible behind the blood-brain barrier; the 2022 paper's authors may have considered and discounted it in text this beat did not read. Haematologica 2022
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PubMed on 2026-09-23 returns 0 records for surfaceome[tiab] AND KMT2A[tiab] and 1 for surfaceome[tiab] AND MLL[tiab]; the single record under the legacy symbol is PMID 33910331.

    What this evidence cannot show

    Bounded by titles and abstracts only. A surfaceome analysis presented as a supporting figure in a paper about something else would be invisible to this query, and this project has recorded that exact failure before.

    Evidence that argues against this (1)
    • The 15 records under surfaceome[tiab] AND leukemia[tiab] were not all read on this beat, so a second MLL-relevant screen could sit among them unexamined. Oncofusion Knowledge Graph 2026
Connected to