Oncofusion

start / graph / gene-ar

Genestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q78-interactome-absent

AR / androgen receptor - the best-supported partner in the absence set by independent-publication count, and the reason a menin inhibitor is in prostate cancer trials

also called AR · NR3C4 · DHTR · Androgen receptor · Dihydrotestosterone receptor

The androgen receptor is the switch that drives prostate cancer. It physically holds onto the MLL complex, and blocking menin - the same drug target as in infant leukaemia - shuts that switch down. This is the one partner on the missing list with two separate labs behind it.

The same thing, for a scientist

AR (P10275, 920 aa) is the ligand-activated nuclear receptor that drives prostate cancer. It reaches this project as a **KMT2A partner with two independent publications** - the only partner in this beat's twelve that has more than one [E1]. Grasso et al. 2012 immunoprecipitated **endogenous AR from VCaP cells and blotted MLL, MLL2, ASH2L, UTX and ASXL1, with FOXA1 as a positive and EZH2 and CHD1 as negatives, and confirmed AR-MLL and AR-ASH2L by reverse immunoprecipitation**; siRNA against MLL or ASH2L inhibited androgen-stimulated KLK3/PSA expression by more than 7.5-fold [E2]. Velot et al. 2021 recovered KMT2A and AFF4 in an AR BioID proximity map in LAPC4 cells and separately co-immunoprecipitated AR with KMT2A in HEK293T [E3]. **And the consequence for this project is direct: Malik et al. 2015 showed that the menin-MLL interaction is required for AR signalling and that a menin-MLL inhibitor blocks castration-resistant prostate cancer growth** [E4]. The menin-KMT2A interface this graph is built around is therefore a validated drug target in a second disease, through this partner, and this graph had no AR entry.

Go deeper
Five records, two publications, and a correction to the dispatch count
IntAct ACpairtypedetectionexpansionsource
EBI-15988502AR-KMT2Aassociationanti-bait co-IPspokePMID 22722839
EBI-15988542AR-KMT2Aphysical associationanti-bait co-IPnonePMID 22722839
EBI-26965611AR-KMT2Aproximity (MI:2364)BioIDspokePMID 33640491
EBI-26966395AR-KMT2Aphysical associationanti-tag co-IPnonePMID 33640491
EBI-26965572AR-AFF4proximity (MI:2364)BioIDspokePMID 33640491

The dispatch brief recorded 1 record, 0 spoke. The REST endpoint returns 5 records across 2 baits and 2 publications, of which 3 are spoke-expanded and 2 are not 1. Both the record count and the spoke count were wrong in the brief, in opposite directions - so the partner is both better supported and more spoke-contaminated than the summary said. This is the largest single discrepancy in the beat and is treated in ana-the-unwritten-interactome-partners-are-better-observed-than-the-written-ones.

The pattern matters: each publication contributes one spoke row and one non-spoke row, because the spoke row is the screen output and the non-spoke row is the directed validation of it. Counting rows conflates a discovery with its own confirmation.

Grasso 2012 is an exome-sequencing paper, and its interaction figure is real

The mutational landscape of lethal castration-resistant prostate cancer reads like the last place to find a co-IP, and this is the failure mode the dispatch brief's record counts ran into: the abstract's headline is 50 exomes and a mutation rate. But the abstract itself states 'demonstrate interaction of the MLL complex with the AR, which is required for AR-mediated signalling', and Figure 3a is a proper endogenous co-IP panel with two negative controls 2.

So a sequencing paper contributed one of the better-controlled interaction experiments in this beat, and any classifier that grades partner sources by publication type - as this project's own bait-versus-prey filter nearly did (brief rule 92) - would have mislabelled it.

Why a prostate partner is a menin-hypothesis result and not a curiosity

Malik et al. 2015 is the load-bearing consequence 4. The menin-MLL complex is required for AR signalling in castration-resistant prostate cancer, and a small-molecule menin-MLL inhibitor blocked CRPC growth in vitro and in vivo. That is the same interface, the same chemistry family, and the same argument this project makes for infant leukaemia, working in a disease with an entirely different driver.

That cuts both ways and the entry should say so.

  • For the hypothesis: the menin-MLL interface is not a leukaemia idiosyncrasy. It is a general transcriptional dependency, independently drugged, with an independent phenotype. A second disease validating the same interface is the strongest kind of external support this project can get.
  • Against the hypothesis: if menin-MLL is required for AR signalling, then a menin inhibitor is an anti-androgen in every androgen-responsive tissue, including a growing child's. No paediatric menin-inhibitor trial has reported gonadal or pubertal endpoints, and the population is infants and young children who will be followed for decades.

A third reading is separately documented: menin also enhances AR-independent proliferation and migration in prostate cancer cells 5, so menin's role there is not exhausted by the AR contact.

Facts and figures
uniprotP10275
length AA920
gene symbolAR
protein nameAndrogen receptor (Dihydrotestosterone receptor, Nuclear receptor subfamily 3 group C member 4)
uniprot resolutionaccession:P10275, reviewed human, one record; primary gene name read back as AR (entry ANDR_HUMAN), synonyms NR3C4, DHTR; length 920 aa checked against UniProt 2026-09-21. The two-letter symbol was resolved by accession, never by symbol search.
symbol collision checkdata/symbol-collisions.json (74 rows) greped 2026-09-21 for AR as a whole-word token: no row existed. A row IS proposed in this beat's memo - AR is a two-letter token that matches as an abbreviation for aldose reductase (AKR1B1), allergic rhinitis, androgen receptor, aspect ratio and the element argon, and it is also a UniProt alias of TCF20 ('Protein AR1').
brief stated record count1
brief stated spoke count0
rest record count5
rest spoke count3
therapeutic consequencea menin-MLL inhibitor blocks CRPC growth (PMID 25822367) - the menin-KMT2A interface validated as a drug target in a second disease
shown for fusionNO - nobody has looked in any KMT2A fusion, in any lineage. Every system is prostate or HEK293T.
fetched2026-09-21
subnuclear localizationUnestablished in haematopoietic cells; the AR nuclear-foci and condensate literature is prostate. Tissue-bounded absence, not an unmeasured one.
localization shown for fusionEXPRESSION YES in MLL-AF9; LOCALISATION NO in any fusion. AR expression has been measured in a murine MLL-AF9 acute myeloid leukaemia model, where high-AR leukaemia-initiating cells cause more severe disease and AR is higher in female LICs (PMID 40991369) - that is MLL-AF9, murine and myeloid, NOT KMT2A::AFF1 and not human infant B-lineage. Nobody has asked where AR is (cytoplasm versus nucleus) in any KMT2A-rearranged cell: the bounded localisation query returned 0. Kept separate from this entry's existing shown_for_fusion, which concerns the AR-KMT2A/AFF4 INTERACTION.
baits
[
  "KMT2A (Q03164)",
  "AFF4 (Q9UHB7)"
]
interaction with kmt2a
{
  "n_intact_records": 4,
  "n_spoke_expanded": 2,
  "n_source_publications": 2,
  "intact_acs": [
    "EBI-15988502",
    "EBI-15988542",
    "EBI-26965611",
    "EBI-26966395"
  ],
  "interaction_types": {
    "MI:0914 association": 1,
    "MI:0915 physical association": 2,
    "MI:2364 proximity": 1
  },
  "systems": [
    "Homo sapiens VCaP prostate carcinoma (endogenous)",
    "Homo sapiens LAPC4 prostate carcinoma (BioID)",
    "Homo sapiens HEK293T"
  ],
  "reciprocal_direction_published": true,
  "functional_arm": "siRNA against MLL or ASH2L inhibits R1881-stimulated KLK3/PSA by >7.5-fold at 24 h, two siRNAs each, p<0.001",
  "negative_controls_in_source": [
    "EZH2 (negative)",
    "CHD1 (no interaction observed)"
  ],
  "shared_pdb_entries": []
}
interaction with aff4
{
  "n_intact_records": 1,
  "n_spoke_expanded": 1,
  "n_source_publications": 1,
  "intact_acs": [
    "EBI-26965572"
  ],
  "interaction_types": {
    "MI:2364 proximity": 1
  },
  "system": "Homo sapiens LAPC4 prostate carcinoma cell",
  "independent_corroboration": 0
}
other mll complex partners in same experiment
[
  "KMT2D",
  "ASH2L",
  "KDM6A",
  "ASXL1",
  "FOXA1"
]
localization
{
  "subcellular": "LIGAND-SWITCHED between cytoplasm and nucleus, and that switch IS the localisation fact. UniProt P10275: Nucleus (ECO:0000269 PubMed:12958311, 15634333, 17587566, 19244107, 19345326, 25091737) and Cytoplasm (12958311, 17587566, 19244107), with the note that AR is predominantly cytoplasmic unliganded and translocates on ligand binding, and can also enter the nucleus unliganded in the presence of RACK1. Detected at the promoters of target genes (PMID 25091737).",
  "subnuclear": "NOT ESTABLISHED IN ANY HAEMATOPOIETIC CELL. Ligand-induced AR nuclear foci and condensate behaviour are extensively reported, and every record examined here is prostate or a prostate-derived line. Recorded as a TISSUE-BOUNDED absence.",
  "how_shown": "immunofluorescence and fractionation across the coded citations; chromatin occupancy for the promoter statement; HPA antibody immunofluorescence",
  "tissue_of_evidence": "PROSTATE, overwhelmingly, plus HEK293T for the interaction work. UniProt's own tissue comments are heart and skeletal muscle, and basal and stromal cells of the prostate. Nothing in the localisation literature is leukaemic.",
  "hpa_agrees_with_the_unliganded_state": "HPA v25.1 gives AR a single main location, CYTOSOL, reliability Supported. That is exactly what the ligand model predicts for cells cultured without androgen, and it is a useful independent confirmation that the resting address is cytoplasmic rather than nuclear.",
  "why_this_matters_here": "This graph's AR entry exists because the menin-MLL interface is AR-dependent in prostate cancer. If that dependence is to be reasoned about in leukaemia, the relevant question is not whether AR is present but whether it is NUCLEAR - i.e. whether the ligand is there. In AML the two published papers disagree about whether androgen signalling does anything, which is precisely a question about occupancy rather than abundance.",
  "node": "localization-partner-baseline-compartment-is-often-miscalled"
}
cell type
{
  "haematopoietic_answer": "AML BLASTS AND LEUKAEMIA-INITIATING CELLS - and, importantly, NOT normal blood. HPA Blood Atlas records AR as Not detected in immune cells, distribution Not detected. Against that, AML cells express high levels of AR across cytogenetic risk groups, induced by IL-6 and correlating with poor overall survival (PMID 40935826); and in a murine MLL-AF9 AML model, high-AR leukaemia-initiating cells produce more severe disease than low-AR LICs, with AR higher in female LICs, a difference CONFINED TO THE LICs AND ABSENT FROM NORMAL BONE MARROW CELLS (PMID 40991369).",
  "the_shape_of_this_result": "Normal-blood-negative and leukaemia-positive is a tumour-restricted expression pattern, which is the most interesting shape a cell-type field can take in this project. It was not in this entry before: the entry contained no mention of AML at all.",
  "the_two_papers_disagree": "PMID 40991369 frames AR signalling as a target in AML. PMID 40935826 finds high AR expression in AML but reports that androgens had no effect on the survival of primary AML cells in vitro. High expression is agreed; dependence is not. Do not cite either alone.",
  "tissue_label_required": "Brief rule 73. Every mechanistic AR statement elsewhere in this entry is PROSTATE or HEK293T, and the AML results above are myeloid and mostly murine. Neither is infant B-lineage KMT2A::AFF1."
}
cell state
{
  "state": "LEUKAEMIA-INITIATING CELL state in murine MLL-AF9 AML, where AR is higher in LICs than in normal marrow and higher in female than in male LICs; and an IL-6-induced state in human AML, where IL-6 signalling raises AR (PMID 40935826).",
  "sex_as_a_state_variable": "Both papers make SEX a variable of the leukaemic state rather than of the host: AR-high female LICs behave differently from AR-low male LICs regardless of recipient sex (PMID 40991369), and male sex is an independent adverse risk factor in 4,281 intensively treated AML patients (PMID 40935826). This graph has no other entry in which sex is a cell-state variable."
}
localization bounded absence
{
  "date": "2026-09-21",
  "tool": "NCBI E-utilities esearch, db=pubmed",
  "shape": "gene clause AND localisation clause AND KMT2A/MLL-fusion clause, one [tiab] tag per term (brief rule 94), every cytogenetic term quoted (brief rule 95)",
  "querytranslation_fusion_clause": "\"KMT2A\"[Title/Abstract] OR \"MLL\"[Title/Abstract] OR \"MLL1\"[Title/Abstract] OR \"MLL-AF4\"[Title/Abstract] OR \"MLL-AF9\"[Title/Abstract] OR \"MLL-ENL\"[Title/Abstract] OR \"KMT2A-rearranged\"[Title/Abstract] OR \"MLL-rearranged\"[Title/Abstract] OR \"t 4 11\"[Title/Abstract] OR \"t 9 11\"[Title/Abstract] OR \"MLL fusion\"[Title/Abstract]",
  "count": 0,
  "controls": {
    "fusion_clause_alone": 8743,
    "gene_clause_alone": 27755
  },
  "reading": "Counted in TITLES AND ABSTRACTS only; supporting figures were not searched (brief rule 80)."
}
How this could bite us back
  • If menin-MLL is required for AR signalling, a menin inhibitor is a partial anti-androgen in every androgen-responsive tissue. In infants and young children on months of therapy, that is a developmental-endocrine liability with no published endpoint anywhere in the class.
  • The same finding is the strongest external validation this project's central interface has. Both readings come from the same paper and the entry keeps both, because quoting only the supportive one is how a hypothesis stops being testable.
  • AR is a two-letter token. Any automated enumeration that resolves 'AR' from text rather than from an accession will pick up aldose reductase, allergic rhinitis and argon - and TCF20's own UniProt alias is 'Protein AR1'.
  • If AR supports leukaemia-initiating cells in MLL-AF9 (PMID 40991369) but primary human AML blasts do not depend on androgen in vitro (PMID 40935826), an AR-directed combination could look inactive in every short-term assay this project would run and still matter for relapse - the LIC compartment is the one such assays sample worst.
Still open
  1. Does a clinical menin inhibitor - revumenib, ziftomenib or bleximenib rather than the tool compound used in 2015 - suppress AR signalling at the exposures children receive? If it does, an anti-androgen effect is on the label of a paediatric drug and nobody has looked for it.
  2. Have any paediatric menin-inhibitor trials collected gonadal, pubertal or growth endpoints? The population is infants and young children followed for decades, and the AR result predicts an endocrine signal.
  3. Is the AR-AFF4 proximity row real? It is a single spoke-expanded BioID hit with no corroboration, and AFF4 rather than AFF1 - which matters, because this graph's SEC structural evidence is all AFF4 imported to AFF1 by paralogy.
  4. AR is high in murine MLL-AF9 leukaemia-initiating cells and absent from normal blood, and the menin-MLL interface is AR-dependent in prostate cancer. Is AR NUCLEAR in a KMT2A-rearranged blast, and does a menin inhibitor change its chromatin occupancy? Cytoplasmic AR would make the prostate mechanism inapplicable in leukaemia in one immunofluorescence.
  5. Two 2025 papers agree that AML blasts express high AR and disagree about whether they depend on it (PMID 40991369 versus 40935826). Does the disagreement track the model - murine MLL-AF9 LICs in vivo versus human primary blasts in vitro - and is the KMT2A-rearranged subset ever analysed separately?
Every source behind this entry (9)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    IntAct's REST record holds five AR interactions against core baits from two independent publications: AR-KMT2A spoke-expanded association and AR-KMT2A non-spoke physical association from PMID 22722839, and AR-KMT2A spoke-expanded BioID proximity, AR-KMT2A non-spoke anti-tag co-immunoprecipitation and AR-AFF4 spoke-expanded BioID proximity from PMID 33640491.

    What this evidence cannot show

    One database on one day, REST endpoint (brief rule 93). Each publication contributes one spoke row and one non-spoke row, which are the screen output and its own directed validation - so five records are two independent experiments plus their confirmations, not five findings. The AFF4 arm rests on a single spoke-expanded proximity row and is much weaker than the KMT2A arm. The dispatch brief for this beat recorded 1 record and 0 spoke rows; both figures were wrong.

    Evidence that argues against this (2)
    • Three of the five rows are fractions of single depositions, and the fractions are very large. Re-enumerated at the endpoint on 2026-09-21: PMID 33640491 contributes 420 IntAct rows carrying only TEN distinct interaction ACs. The AR-KMT2A proximity row EBI-26965611 is one of 182 pairs sharing that one AC, and the AR-AFF4 row EBI-26965572 is one of 31 sharing another; both carry intactMiscore 0.27. The Grasso spoke row EBI-15988502 likewise expands into seven pairs - AR against ERG, KDM6A, ASXL1, FOXA1, ASH2L, KMT2D and KMT2A. So 'five records' is two directed co-immunoprecipitations plus three prey-list positions, and the AFF4 arm in particular is one prey identification in one of ten depositions, with no directed validation anywhere. EMBL-EBI IntAct 2026
    • The 756 denominator is also endpoint-specific, and AR is one of the accessions where the two endpoints diverge most. Measured the same day: IntAct REST 756 records for P10275 against PSICQUIC id:P10275 420 (56%). Because the shortfall is protein-specific - 82% recovered for Q13309, 26% for O60477 - the sentence 'the only partner in this beat's twelve that has more than one publication' is a statement about what one endpoint surfaces for each accession, not a comparison on equal footing. EMBL-EBI IntAct and PSICQUIC 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In human VCaP castration-resistant prostate cancer cells (not haematopoietic, no KMT2A fusion), endogenous AR was immunoprecipitated and MLL, MLL2, ASH2L, UTX and ASXL1 all co-precipitated while CHD1 did not, with FOXA1 as a positive control and EZH2 as a negative; reverse immunoprecipitation confirmed AR-MLL, AR-MLL2, AR-ASH2L and AR-FOXA1, and siRNA against MLL or ASH2L inhibited R1881-stimulated KLK3/PSA expression by more than 7.5-fold at 24 hours (two siRNAs each, p < 0.001).

    PMID 22722839DOI
    What this evidence cannot show

    VCaP prostate carcinoma cells; no haematopoietic system, no KMT2A fusion, no paediatric context (brief rule 73). Full text read this session via Europe PMC fullTextXML (PMC3396711). The interaction figure is a secondary panel of an exome-sequencing paper. The paper's 'MLL2' is subject to brief rule 66's ambiguity - IntAct maps it to KMT2D (O14686) and that assignment was not independently verified here; 'MLL' unambiguously means KMT2A.

    Evidence that argues against this (2)
    • This panel did not test the subunit the field subsequently identified as the point of contact. Malik et al. 2015 report that AR interacts with the MLL complex VIA THE MENIN-MLL SUBUNIT, that menin is higher in CRPC than in hormone-naive or benign prostate, and that a menin-MLL inhibitor blocks AR signalling and CRPC growth in vivo. Grasso's blots - MLL, MLL2, ASH2L, UTX, ASXL1, with FOXA1 and EZH2 as controls - contain no menin lane. So the experiment establishes that AR recovers the complex; it does not establish an AR-KMT2A binary contact, and on the best current account it is menin rather than KMT2A that AR is holding. The IntAct rows this entry counts are AR-KMT2A pairs, which is the curator's binarisation of that co-IP and not a measured contact. Malik R et al. 2015
    • Recovery of MLL-complex subunits in a prostate line is not specific to an AR pull-down. Using human autoantibodies in docetaxel-resistant prostate cancer cells, Ortiz-Hernandez et al. co-immunoprecipitate LEDGF/p75 with menin, MLL, JPO2, IWS1, ASK1, PogZ, c-MYC and HRP2 and confirm nuclear co-localisation - the same complex, a different bait, the same tissue. An interaction panel in this cell background recovers the MLL complex from more than one handle, which is a reason to want the reverse-IP and negative-control structure Grasso does supply, and a reason not to read the panel as AR-specific chromatin biology. Ortiz-Hernandez GL et al. 2021
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In androgen-dependent LAPC4 human prostate cells, a BioID proximity map of AR identified 267 non-redundant AR-associated candidates across unstimulated and androgen-stimulated conditions, among them KMT2A and AFF4; AR-KMT2A was separately confirmed by anti-tag co-immunoprecipitation in HEK293T.

    PMID 33640491DOI
    What this evidence cannot show

    Abstract read this session; the KMT2A and AFF4 identifications are taken from IntAct's curation of the study rather than from the paper's own text, which names KLF4 as its headline finding. LAPC4 prostate cells and HEK293T - no haematopoietic system, no KMT2A fusion (brief rule 73). Proximity labelling records neighbourhood, not contact, and both proximity rows are spoke-expanded.

    Evidence that argues against this (1)
    • 213 of the 267 AR-associated candidates in this study were not previously reported, which is a large novel fraction for a BioID map and means the false-positive rate on any single identification is unquantified. KMT2A's presence is corroborated by an independent co-IP; AFF4's is not. Vélot L et al. 2021
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The menin-MLL interaction is required for androgen receptor signalling in castration-resistant prostate cancer, and a small-molecule menin-MLL inhibitor blocked CRPC growth - making the menin-KMT2A interface this project is built around an independently validated drug target in a second, unrelated disease.

    PMID 25822367DOI
    What this evidence cannot show

    Abstract-level reading this session; the compound and the in vivo model were not verified from full text. Prostate cancer in adult men; no haematopoietic system and no KMT2A fusion (brief rule 73). New to this graph (checked with scripts/check-citation-reuse.py), which is itself notable - the strongest external validation of this project's central interface was not cited anywhere in 1,806 entries.

    Evidence that argues against this (1)
    • Menin also enhances AR-INDEPENDENT proliferation and migration in prostate cancer cells, so menin's role in that disease is not exhausted by the AR contact and a menin inhibitor's effect there is not purely an anti-androgen effect. Kim T et al. 2022
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Menin enhances androgen-receptor-independent proliferation and migration of prostate cancer cells, establishing a second, AR-independent menin function in the same disease.

    PMID 35014621DOI
    What this evidence cannot show

    Abstract only; prostate cancer cell lines. Included as the counterweight to reading the menin-AR axis as the whole of menin's role in that disease.

    Evidence that argues against this (1)
    • The better-evidenced prostate function of menin is the AR-DEPENDENT one, and the two have never been tested against each other. Malik et al. 2015 (Nat Med) make the menin-MLL interaction required for AR signalling, tie menin expression to CRPC and to poor overall survival, and show a menin-MLL inhibitor blocks castration-resistant growth in vivo. This item rests on one abstract-level report in Mol Cells. Recording an AR-independent menin function is right; giving it equal standing with the AR-dependent one would not be, and nothing in either paper excludes the simpler reading that the AR-independent effects are downstream of the same complex acting at non-AR loci. Malik R et al. 2015
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt P10275 records AR in both nucleus and cytoplasm and states that the receptor is predominantly cytoplasmic when unliganded and translocates to the nucleus on ligand binding, with RACK1 permitting unliganded nuclear entry.

    What this evidence cannot show

    Database comment on one day. Every coded citation is prostate or a transfected line; there is no haematopoietic experiment behind any of them. Brief rule 91: the citation behind the RACK1 statement (PMID 12958311) is a RACK1 interaction paper, so the localisation statement is a secondary observation inside it.

    Evidence that argues against this (1)
    • The cytoplasmic-unliganded / nuclear-liganded model may be an incomplete account of where the receptor acts. Non-genomic steroid-receptor signalling is reported in human prostate stem and progenitor cells, where the operative events are not nuclear translocation of the receptor; if a comparable non-genomic arm exists in blood, a nuclear/cytoplasmic immunofluorescence - the experiment this entry proposes - would return 'cytoplasmic' and be read as 'AR is not doing anything', which would be the wrong conclusion. Endocrinology 2019
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a murine MLL-AF9 acute myeloid leukaemia model, high-AR leukaemia-initiating cells produce more severe disease than low-AR LICs in both male and female recipients, AR expression is higher in female LICs, and the difference is confined to LICs and absent from normal bone marrow cells.

    PMID 40991369DOI
    What this evidence cannot show

    MURINE MLL-AF9, a myeloid fusion, read at abstract level. This is an EXPRESSION and phenotype result, not a localisation experiment - nothing in it says whether AR is nuclear in these cells. Brief rule 73 and rule 86: name MLL-AF9, never 'a KMT2A fusion', and never transfer it to KMT2A::AFF1 infant B-ALL.

    Evidence that argues against this (1)
    • A 4,281-patient AML analysis confirms high AR expression across cytogenetic risk groups but finds that androgens had no effect on the survival of primary AML cells in vitro, so the dependence claim does not travel with the expression claim. Leukemia 2025
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AML cells express high levels of androgen receptor across cytogenetic risk groups, AR expression is induced by IL-6 signalling in vitro and correlates with poor overall survival, yet androgens had no effect on the survival of primary AML cells in vitro; male sex was an independent adverse risk factor across 4,281 intensively treated patients.

    PMID 40935826DOI
    What this evidence cannot show

    Human AML but not KMT2A-stratified in anything read here, and myeloid rather than B-lineage. Abstract-level reading. A negative in-vitro survival result does not exclude an in-vivo or niche-dependent effect, which is where the Blood Advances paper's disagreement lives.

    Evidence that argues against this (1)
    • The murine MLL-AF9 study reaches the opposite therapeutic conclusion, reporting AR-dependent LIC behaviour with different mechanisms in the two sexes. Blood Adv 2025
  9. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Human Protein Atlas v25.1 gives AR a single main subcellular location, Cytosol, with reliability Supported, and records AR as not detected in immune cells in the Blood Atlas.

    What this evidence cannot show

    HPA's immunofluorescence lines are cultured without added androgen, so a cytosolic main call reports the UNLIGANDED state and is not evidence that AR cannot be nuclear. The Blood Atlas Not-detected call is RNA in normal peripheral blood and is not contradicted by high AR in AML blasts - different cells.

    Evidence that argues against this (1)
    • HPA's 'not detected in immune cells' is contradicted at protein level in the cells this project cares about: AML blasts express high AR across cytogenetic risk groups. A normal-peripheral-blood RNA atlas is the wrong denominator for a leukaemia graph, and citing the Not-detected call without that correction would produce exactly the error this entry is trying to avoid - concluding that AR is irrelevant in blood because it is absent from healthy blood. Leukemia 2025
Connected to