Oncofusion

start / graph / ana-cd133-is-five-times-higher-in-normal-kidney-than-in-bone-marrow-and-the-safety-filter-had-omitted-every-surface-antigen

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

PROM1/CD133 is 5.27x higher in normal kidney than in bone marrow - worse than MEN1 on this project's own safety metric - and it was absent from the filter because that filter scored 22 chromatin targets and not one surface antigen

also called CD133 normal tissue ratio 5.27 · PROM1 kidney 19.5 nTPM vs marrow 3.7 · the surface antigens were never in the HPA filter · hpa-normal-tissue-surface-antigens

CD133 is the most promising surface handle on this leukaemia and the one an antibody drug would aim at. Asking where it sits in a normal body shows it is about five times more abundant in kidney than in the bone marrow you are trying to treat, which is a worse ratio than the drug target this project already flags for hormone-gland side effects.

The same thing, for a scientist

**This project built a normal-tissue safety filter and left out the entire class of targets a drug would actually bind.** `ana-hpa-normal-tissue-target-filter` scored **22 candidates plus two controls** on Human Protein Atlas consensus tissue RNA as *highest normal tissue / bone marrow*. Every one is intracellular or chromatin - MEN1, KMT2A, DOT1L, MEIS1, CDK9. **Not one surface antigen is in it, including PROM1/CD133.** That omission became load-bearing twice over. `gene-prom1` records CD133 as the most completely worked-out marker in the disease - essential for leukaemic growth, driven by **direct MLL-AF4 binding** through an intragenic H3K79me2/3 enhancer. And cycle 135's q149 made it clinically central from the other side: within KMT2A::AFF1 infant ALL, **PROM1 is ON in 24 of 25 HOXA-positive cases, and 11 of 13 PROM1-negative cases are IRX1-positive** - so a CD133-directed agent should be expected to select for the IRX1 subgroup. **Run on the identical panel, metric, identifier guard and controls, so the two runs read on one axis [E1]:** | gene | bone marrow nTPM | highest normal tissue | ratio | |---|---:|---|---:| | **PROM1 / CD133** | **3.7** | **kidney 19.5** | **5.27** | | TAPT1 *(PROM1's KEE partner)* | 24.2 | liver 28.5 | 1.18 | | LAMP5 | 2.1 | cerebral cortex 41.1 | **19.57** | | CSPG4 / NG2 | **0.0** | lung 18.1 | *undefined* | | CD19 *(artefact control)* | 5.9 | spleen 70.8 | 12.0 | | ACTB *(reference)* | 3397.8 | lymph node 4107.2 | **1.21** | | ALB *(reference)* | 0.0 | liver 198523.8 | *undefined* | 🔵 **The method reproduces.** ACTB came back at **1.21** against the earlier run's **1.2**, and ALB reproduced as **undefined with marrow 0.0 and liver 198,523.8** - the same two reference points, unchanged. The comparison to the 22 chromatin targets is therefore a like-for-like one, not a new incomparable metric. 🟠 **PROM1's 5.27 places CD133 above MEN1 (3.3), DPY30 (2.7) and DOT1L (2.4), and just below MLLT1 (6.6).** The pre-registered S1 band was <=1.5 (where KMT2A sits at 1.0) and S2 was the MEIS1 17.2 / MEF2C 11.6 band. **It landed between them: INTERMEDIATE.** So this is not a disqualifying number and it is not a reassuring one - it says a CD133 agent acts on normal kidney at roughly five times the level it acts on the compartment being treated. 🔴 **And the off-tissue is real parenchyma, which is the part that matters.** CD19's ratio is **12.0**, but its top tissue is **spleen** - CD19 is a B-cell antigen and spleen and lymph node are full of B cells, so CD19's number is **entirely the lymphocytes inside those organs**. PROM1's top tissue is **kidney**, which is not a lymphoid organ. **The same metric that is meaningless for CD19 is meaningful for PROM1**, and the reason is which tissue the signal is in, not how big the ratio is.

Go deeper
What this does to the surface-antigen hypothesis

hyp-no-surface-antigen-is-both-fusion-driven-and-normal-tissue-clean-in-kmt2a-aff1 asserts an empty intersection between 15 fusion-enriched and 3 parenchymally clean surface proteins among 169 candidates, and argues the disjointness is mechanistic: the fusion de-represses a developmental programme, and developmental programmes are used elsewhere in the body.

That hypothesis did not mention PROM1, so its central claim had never been tested against the surface antigen most likely to matter. It now has its first quantitative datum from that antigen, and the datum is consistent with it: CD133 is fusion-driven (direct MLL-AF4 binding, published) AND not normal-tissue clean (kidney 5.27x marrow). That is the hypothesis's predicted shape.

🟠 But one row argues the other way and should not be buried: TAPT1 sits at 1.18, marrow-selective, despite sharing the intragenic enhancer that drives PROM1. So being switched on by the same element does not force an off-compartment profile, which weakens the mechanistic version of the hypothesis even as the PROM1 row supports its empirical version.

What this cannot do, carried forward unchanged

Verbatim from the earlier run, because it still holds: mRNA is not protein and expression is not dependence. A tissue can express a gene highly and be unaffected by inhibiting it, and express it little and be exquisitely sensitive. This ranks where to look for toxicity; it does not predict toxicity.

Two further limits specific to this run:

  1. HPA consensus values are ADULT tissue and the population of concern is infants. CD133 is a developmental stem/progenitor antigen, so adult kidney may understate infant expression in exactly the organ that carries the signal. This was pre-registered as S3(c) and it does not resolve.
  2. A ratio is not a dose. 19.5 nTPM in kidney is a modest absolute level - three orders of magnitude below ACTB's 3,397.8 in marrow. The ratio says where the drug is acting relative to the compartment; it says nothing about whether 19.5 nTPM of CD133 is enough protein for an antibody-drug conjugate to injure a nephron.
Facts and figures
scriptanalyses/hpa-normal-tissue-surface-antigens/run.py
statusrun
run date2026-09-24
sourceHuman Protein Atlas consensus tissue RNA (nTPM)
n tissues16
method noteidentical panel, metric, identifier guard and controls to analyses/hpa-normal-tissue-target-filter so the runs are comparable
control reproductionACTB 1.21 vs the earlier run's 1.2; ALB undefined in both
preregistered branch firedINTERMEDIATE - between the S1 (<=1.5) and S2 (>=6.0) bands
prior run band
CCNT10.400CCNT1: 0.4KMT2A1.000KMT2A: 1ACTB1.200ACTB: 1.2DOT1L2.400DOT1L: 2.4MEN13.300MEN1: 3.3MLLT16.600MLLT1: 6.6MEF2C11.6MEF2C: 11.6MEIS117.2MEIS1: 17.2
prom1
{
  "uniprot": "O43490",
  "ensembl": "ENSG00000007062",
  "length_aa": 865,
  "bone_marrow_ntpm": 3.7,
  "highest_normal_tissue": "kidney",
  "highest_normal_ntpm": 19.5,
  "ratio_worst_to_marrow": 5.27
}
comparators same run
{
  "TAPT1": 1.18,
  "LAMP5": 19.57,
  "CD19": 12,
  "CSPG4": "undefined (bone marrow 0.0)",
  "ACTB": 1.21,
  "ALB": "undefined (bone marrow 0.0)"
}
prior run band — the numbers behind the chart
{
  "MEIS1": 17.2,
  "MEF2C": 11.6,
  "MLLT1": 6.6,
  "MEN1": 3.3,
  "DOT1L": 2.4,
  "ACTB": 1.2,
  "KMT2A": 1,
  "CCNT1": 0.4
}
measurements
[
  {
    "quantity": "highest normal tissue / bone marrow, HPA consensus RNA",
    "value": 5.27,
    "unit": "ratio of nTPM",
    "subject": "PROM1/CD133 (ENSG00000007062), kidney 19.5 nTPM over bone marrow 3.7 nTPM",
    "universe": "a 16-tissue HPA consensus panel, identical to the 22-target run; not the whole body"
  },
  {
    "quantity": "control reproduction against the earlier run",
    "value": {
      "ACTB_now": 1.21,
      "ACTB_earlier": 1.2
    },
    "unit": "ratio of nTPM",
    "subject": "ACTB, the shared reference gene",
    "universe": "the same 16-tissue panel in both runs, establishing the two are comparable"
  }
]
How this could bite us back
  • Selection for the target-negative subgroup - q149 found PROM1 ON in 24 of 25 HOXA-positive KMT2A::AFF1 cases while 11 of 13 PROM1-negative cases are IRX1-positive. A CD133-directed agent therefore treats the HOXA arm and spares the IRX1 arm, and should be expected to select for it - an escape route present at diagnosis rather than acquired. Any CD133 trial in this disease needs PROM1 and IRX1 measured at baseline, not just PROM1.
  • Renal exposure - CD133 mRNA is 5.27x higher in normal kidney than in bone marrow on this project's own metric, and kidney is not a lymphoid organ so the signal is parenchymal rather than an artefact of infiltrating leukocytes. An antibody-drug conjugate or CAR against CD133 carries renal exposure that a marrow-only view does not show. mRNA is not protein and this is not a toxicity prediction - it is a statement about where to look first.
  • Adult reference for an infant disease - HPA consensus values are adult tissue. CD133 is a developmental stem/progenitor antigen, so adult kidney may understate expression in an infant, and the same applies to every other tissue in the panel. The direction of this bias is unknown and it is the single biggest limit on using these numbers for a paediatric go/no-go.
Still open
  1. Has any anti-CD133 agent shown renal toxicity in humans? That is the evidence that would settle what a 5.27x kidney ratio means, and it is a clinical-literature question this run did not ask.
  2. Score the remaining fusion-enriched surface antigens from the 169-candidate surfaceome screen on this same panel. PROM1 was omitted; the other 14 fusion-enriched proteins presumably were too, and the hypothesis's 'empty intersection' rests on a filter that never saw them.
  3. TAPT1 shares PROM1's intragenic enhancer and is marrow-selective at 1.18. Does co-regulation by one element really not constrain tissue distribution, or is TAPT1's promoter contact weaker than gene-prom1 implies?
  4. Is CD133 protein, rather than mRNA, present on normal kidney at a level an ADC could engage? HPA has immunohistochemistry for exactly this and it was not read in this run.
Every source behind this entry (1)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.

    In Human Protein Atlas consensus tissue RNA across a 16-tissue panel, PROM1 is 3.7 nTPM in bone marrow and highest in kidney at 19.5 nTPM, a ratio of 5.27; on the identical panel TAPT1 is 1.18, LAMP5 is 19.57, CD19 is 12.0, CSPG4 has bone marrow 0.0, and the controls ACTB (1.21) and ALB (undefined, marrow 0.0) reproduce the earlier run's published values of 1.2 and undefined.

    What this evidence cannot show

    Run 2026-09-24 (cycle 136). Identifier guard: HPA search is substring-based, so every row is filtered on an exact gene-symbol match, rejected if more than one exact row returns, and its Ensembl id recorded with provenance. PROM1's ENSG00000007062 was cross-checked against UniProt O43490 (gene PROM1, 865 aa, Prominin-1), which lists that Ensembl gene id. TAPT1, CSPG4, LAMP5 and CD19 Ensembl ids were taken from HPA's own exact-symbol row and are flagged HPA-sourced rather than independently verified. HPA consensus values are adult tissue. A 16-tissue panel is not the whole body - the highest-expressing tissue overall may not be in it.

    Evidence that argues against this (2)
    • The ratio metric is badly behaved at both ends and this run shows it twice. With bone marrow at 0.0 it is undefined (CSPG4, ALB), and for a haematopoietic antigen it is inflated by the lymphocytes inside lymphoid organs (CD19 scores 12.0 entirely on spleen and lymph node B cells). So a rank ordering by this ratio mixes genuine parenchymal liability with tissue cellular composition, and the number must always be read together with WHICH tissue carries the signal.
    • PROM1's absolute kidney level is 19.5 nTPM, which is low - CD133 is a well-documented normal kidney proximal-tubule and stem/progenitor antigen, so a ratio of 5.27 may simply be recovering textbook biology rather than revealing a new liability. The clinically relevant question is whether an anti-CD133 agent has shown renal toxicity in humans, which this run does not address and which would be the better evidence.
Connected to