Oncofusion

start / graph / concept-minimally-expressed-in-normal-haematopoietic-cells-is-not-a-safety-claim

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

'Not expressed on normal haematopoietic cells' is a claim about blood, and it has been read for thirty years as a claim about the body

also called the haematopoietic-negative scope error

The reassurance most often given for NG2 as a leukaemia target is that it is absent from normal blood cells. That is true, and it is the wrong comparison: a therapy for leukaemia is allowed to damage blood cells, and what matters is every other tissue - where NG2 is abundant.

The same thing, for a scientist

The founding NG2 paper's title states that the antigen *'is not expressed on the cell surface of normal hematopoietic cells'* [E1]. The 2024 Blood paper restates it as *'minimally expressed in healthy hematopoietic cells'* [E2]. Both are correct. Both are scoped to **the one compartment an anti-leukaemia therapy is permitted to ablate.** The compartment that decides tolerability is everything else, and there NG2 is a pericyte and oligodendrocyte-precursor antigen at 124.5 nTPM in normal blood vessel and 246 TPM in tibial artery [E3]. The same 1996 abstract in fact reports NG2 **on** cell lines derived from non-haematopoietic tissues - the sentence is in the paper and is not usually carried forward with the title [E1]. **The generalisable point: a safety statement inherits the scope of the panel it was measured on.** An antigen screened against a blood panel can only be declared absent from blood. This project has recorded the same shape elsewhere - a database fact is opposable on scope - and it is the reason this beat's screen scores every candidate against 51 tissues rather than against marrow.

Go deeper
How to state a normal-expression negative so it cannot be over-read

Name the panel, in the claim, not in the caveat:

  • Wrong: 'NG2 is not expressed on normal cells.'
  • Still wrong: 'NG2 is not expressed on normal cells (measured in haematopoietic tissue).' - the scope is in a parenthesis that does not survive quotation.
  • Right: 'NG2 is absent from normal haematopoietic cells and present on pericytes and oligodendrocyte precursors.'

The test is whether the sentence still misleads after its parenthetical is deleted, because that is how sentences travel. This project has recorded before that a warning in a field does not protect the prose 4.

Facts and figures
ruleA normal-expression negative inherits the scope of the panel it was measured on. Name the panel inside the claim, and test whether the sentence still misleads once its parenthetical is deleted.
worked examples
{
  "NG2/CSPG4": {
    "scoped_negative": "absent from normal haematopoietic cells (1996)",
    "out_of_scope_positive": "pericytes, oligodendrocyte precursors; blood vessel 124.5 nTPM, tibial artery 246.21 TPM"
  },
  "FOLR1": {
    "scoped_negative": "a fusion-specific gene in a CBFA2T3::GLIS2 model versus matched normal progenitors",
    "out_of_scope_positive": "choroid plexus 2,860.9 nTPM, salivary gland 206.7, lung 182.4, kidney 142.6"
  },
  "LAMP5": {
    "scoped_negative": "specific to MLL-rearranged among leukaemias",
    "out_of_scope_positive": "basal ganglia 297.8 nTPM, nucleus accumbens 323.89 TPM"
  }
}
How this could bite us back
  • Applied indiscriminately, this concept rejects every antigen, because almost everything is somewhere. The point is not that out-of-scope expression disqualifies a target but that it must be stated so a reader can weigh it - FOLR1 went into a CAR-T programme with choroid-plexus expression and that may well have been the right decision.
  • This entry itself makes a scoped claim - three worked examples, all mRNA, all adult - and would be over-read if quoted as 'fusion antigens are unsafe'.
Still open
  1. How many of this graph's stated target-safety claims are scoped to a panel that is not named in the claim itself? A sweep of the 37 target- entries would answer it and has not been run.
  2. For NG2 specifically, is there a primary-tissue immunohistochemistry series that would replace the mRNA argument with a protein one? The pericyte literature is large and was not read on this beat.
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The founding NG2 report states that the human NG2 homologue recognised by antibody 7.1 is not expressed on the cell surface of normal haematopoietic cells, and in the same abstract reports that it was found on cell lines derived from non-haematopoietic tissues.

    PMID 8562938
    What this evidence cannot show

    1996, one monoclonal antibody, flow cytometry on haematopoietic cells and haematopoietic cell lines. The non-haematopoietic observation is on cell lines, not on primary normal tissue, so it is weaker than the negative it sits beside.

    Evidence that argues against this (1)
    • The authors were addressing a diagnostic question, not a therapeutic one, so scoping the panel to blood was the correct design for their purpose. The error is in later re-use, not in the paper. Blood 1996
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A 2024 Blood paper describing NG2 as a direct MLL-AF4 target gene states that NG2 expression is a hallmark of MLLr B-ALL and is minimally expressed in healthy haematopoietic cells.

    PMID 39093982DOI
    What this evidence cannot show

    Read from the abstract. The paper's subject is glucocorticoid resistance, not target safety, so the sentence is background rather than a safety assessment.

    Evidence that argues against this (1)
    • The paper does not propose NG2 as an immunotherapy target, so criticising its safety framing would be unfair; the scope problem arises when the sentence is quoted in a target-selection context. Blood 2024
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    CSPG4/NG2 reaches 124.5 nTPM in normal blood vessel in HPA and 246.21 TPM in tibial artery in GTEx v8, and exceeds 5 nTPM in 31 of 51 HPA normal tissues.

    What this evidence cannot show

    Adult bulk tissue mRNA in both atlases; neither measures infant tissue and neither measures surface protein.

    Evidence that argues against this (1)
    • Abundant mRNA in a tissue does not establish that the protein is reachable by a circulating therapeutic; pericytes sit behind the endothelium. Human Protein Atlas 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    This project's own analysis applied a 51-tissue normal panel to the whole surfaceome and found that every one of the 15 fusion-enriched candidates fails or is amber on it.

    What this evidence cannot show

    Cell-line and bulk-tissue mRNA with pre-registered thresholds; the 5 nTPM clean threshold is a judgement and a looser one would pass several of the 15.

    Evidence that argues against this (1)
    • If a 51-tissue mRNA panel rejects every candidate including CD19 on its first pass, the panel may be too strict rather than the candidates too dirty - which is exactly what the CD19 control failure showed and what the lymphoid-load correction addresses. Oncofusion Knowledge Graph 2026
Connected to