SMARCA4 (BRG1) - a tumour suppressor by mutation and a maintenance requirement by knockdown, and in KMT2A-rearranged lines the dependency data say it is not selective at all

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• Resistance by paralogue: SMARCA2 occupies the same slot. A SMARCA4-selective agent invites SMARCA2 upregulation, which is why the clinical molecules are dual - and dual inhibition removes the intended selectivity.
• Essential-function toxicity: SMARCA4 knockdown reduced LSK colony formation and LSK proliferation in the anchor paper, so the normal-HSPC window is not as clean as 'leukaemia-specific' suggests.
• The SMARCA4-loss tumour-suppressor literature is a direct safety question: chronic pharmacological inhibition of an ATPase whose genetic inactivation drives rhabdoid and small-cell carcinoma is a second-malignancy hypothesis that a paediatric programme has to answer.
• Delivery and population: the only clinical trial of this class in leukaemia (NCT04891757) has a minimum age of 16 years, so infant KMT2A-rearranged disease is outside every current protocol.
• SWI/SNF inhibition impairs PU.1-dependent B-cell and monocyte populations and induces leukopenia in an immunocompetent model - and the index disease is a B-lineage leukaemia in infants whose B-cell compartment is the tissue at risk. That is on-target toxicity in the same lineage as the tumour, which is the hardest kind to dose around, and it was invisible in the immunodeficient xenograft models where FHD-286's efficacy was established.

1. Does SMARCA4 occupancy overlap KMT2A-fusion occupancy genome-wide in a KMT2A-rearranged blast? Shi et al. report that most direct MLL-AF9 target genes do NOT respond to Brg1 loss, which predicts the two occupancies are largely disjoint - nobody has measured it.
2. Is the reported resistance of KMT2A-R B-ALL to BRG1 inhibition a lineage effect, a line-selection effect, or real? It is currently one preprint against one peer-reviewed myeloid paper.
3. Does acute degradation of SMARCA4 (degron or PROTAC) phenocopy shRNA in a KMT2A-rearranged line, or does the slow kinetics of knockdown allow SMARCA2 compensation that a degrader would not?
4. FHD-286 inhibits both BRG1 and BRM, so the MLL1-rearranged differentiation result is a BAF-ATPase result and not a SMARCA4 result. Does a SMARCA4-selective degrader reproduce it? Until that is done, the synthetic-lethal logic in this area - inhibit BRM where BRG1 is lost - cannot be applied to the leukaemia setting.
5. Why does a CRISPR viability screen place SMARCA4 in the anti-selective half while pharmacological BRG1/BRM inhibition drives differentiation specifically in MLL1-rearranged AML? The most economical explanation is that differentiation and fitness are different endpoints, but the alternative - that the drug's effect is largely BRM-driven - has not been excluded.
6. Does SWI/SNF supply accessibility for MEIS1 in a KMT2A-rearranged B-lineage cell? The PU.1-recruits-SWI/SNF mechanism names RUNX1, LMO2 and MEIS1 as the factors that lose binding, and MEIS1 is a canonical index-fusion output - yet the experiment is AML, and KMT2A-rearranged B-ALL lines are reported resistant to BRG1 inhibitors. Accessibility at MEIS1-bound sites under FHD-286 in a KMT2A::AFF1 line is the direct test, and it would also say whether the resistance is target-level or programme-level.
7. Does SMARCA4's bromodomain matter for anything in this disease? It is the only acetyl-lysine reader module in this tranche and every clinical-stage molecule in the class is an ATPase inhibitor. A bromodomain-dead SMARCA4 would say whether the reader fold contributes to the maintenance requirement or is dispensable to it.
8. Is SMARCC1 - not SMARCA4 - the BAF subunit this project should be writing about? It is the only pan-fusion BAF member in the AP-MS analysis and the only one found in every one of the seven baits, while SMARCA4 appears in one. The beat that enumerated BAF chose the ATPase because it is the druggable one, which is a different criterion from being the one the fusion network actually contains.

1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   SMARCA4 is UniProt P51532, 1,647 aa, primary gene name SMARCA4, synonyms BAF190A BRG1 SNF2B SNF2L4.

   UniProtKB P51532 (SMCA4_HUMAN) · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   Query gene_exact:SMARCA4 AND organism_id:9606 AND reviewed:true run 2026-09-20; exactly one reviewed record; primary gene name and length read back off the same fetched record.

   ▸Evidence that argues against this (1)

   • The protein-name field is itself a trap: SMARCA4's reads 'BRG1-associated factor 190A' and SMARCA2's reads 'BRG1-associated factor 190B'. Reading the protein name back to check 'is this BRM?' returns a string containing BRG1 either way. Only gene_primary and the 1,647-vs-1,590 length separate them. UniProt Consortium 2026
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   SMARCA4 depletion in MLL-AF9 mouse cells and in human THP-1 and MV4-11 cells causes loss of expansion, self-renewal and the undifferentiated state, rescued by human SMARCA4 cDNA; SMARCA4 is described as sustaining leukaemogenic transformation in MLL-rearranged leukaemia despite a tumour-suppressor mutational spectrum in solid tumours.

   SWI/SNF Subunits SMARCA4, SMARCD2 and DPF2 Collaborate in MLL-Rearranged Leukaemia Maintenance. · Cruickshank VA, Sroczynska P, Sankar A, Miyagi S, et al. · PLoS One 2015
   PMID 26571505DOI
   ▸What this evidence cannot show

   shRNA knockdown throughout, no knockout or degron; MV4-11 is a KMT2A::AFF1 line but is myeloid, so the KMT2A::AFF1 B-lineage disease this lab studies is not represented. The paper's own leukaemia-specificity control is mixed: c-Kit-enriched bone marrow proliferation was unaffected, but LSK colony number and LSK proliferation WERE reduced by the same knockdowns - so the normal-tissue window is narrower than the abstract implies. Read in full via PMC4646637.

   ▸Evidence that argues against this (1)

   • The same paper shows Jurkat and T-ALL1 (T-ALL, no KMT2A rearrangement) also require full SMARCA4 expression, which removes fusion specificity from its own result. Xiu Y et al. 2018
3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Brg1 is required for MLL-AF9/NrasG12D AML maintenance in vivo and in human NOMO-1 cells, and the requirement is explicitly not restricted to MLL fusion subtypes: BCR-ABL B-ALL, KASUMI-1, JURKAT, REH and K-562 are also sensitive.

   Role of SWI/SNF in acute leukemia maintenance and enhancer-mediated Myc regulation. · Shi J, Whyte WA, Zepeda-Mendoza CJ, Milazzo JP, et al. · Genes Dev 2013
   PMID 24285714DOI
   ▸What this evidence cannot show

   shRNA competition assays in mouse and human lines; the 'not restricted to MLL fusion subtypes' statement rests on supplementary panels rather than a matched quantitative comparison. Read in full via PMC3877755.

   ▸Evidence that argues against this (1)

   • HEL and CMK AML lines were much less affected in the same panel, so the requirement is not universal across AML either - which means neither 'KMT2A-specific' nor 'all leukaemia' is right, and the determinant is unknown. Xiu Y et al. 2018
4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

   SMARCA4 ranks 16,333 of 17,728 for KMT2A-rearranged selectivity with difference +0.0791 (KMT2A-r mean -0.4469, other leukaemia -0.5260, pan-essentiality -0.3636); SMARCA2 ranks 11,581 (+0.019) with pan-essentiality +0.0285.

   Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap Chronos via Open Targets) · Oncofusion knowledge graph, analyses/depmap-kmt2a-selective-scan-full/run.py · DepMap Chronos gene effect via Open Targets GraphQL 2026
   ▸What this evidence cannot show

   11 index and 43 comparator leukaemia lines, none infant-derived; Cellosaurus fusion annotation may be incomplete, biasing toward the null; DepMap release version not stated by the API. A positive difference is an absence of selectivity, NOT an absence of dependency - the KMT2A-r mean of -0.4469 is a genuine dependency.

   ▸Evidence that argues against this (1)

   • A single-gene CRISPR knockout cannot see a dependency buffered by a paralogue. Because SMARCA2 can substitute for SMARCA4 in the complex, this rank is uninformative about ATPase-module dependency as a whole. Fiskus W et al. 2024
5. E5weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.preprint

   KMT2A-R B-ALL cell lines are resistant to the BRG1 inhibitors BRM014 and FHD-286 while Ph-like B-ALL lines are markedly sensitive, and elevated BRG1 expression predicts shorter overall survival in Ph-like but shows the opposite trend in KMT2A-R B-ALL.

   Selective Sensitivity of Ph-like B-ALL to BRG1 Inhibition Reveals a Novel Targeted Therapy Strategy. · Abhinav Ayyadevara VSS, Gaur S, Paik A, Perencsik R, et al. · bioRxiv 2025
   PMID 40672274DOI
   ▸What this evidence cannot show

   bioRxiv preprint, not peer-reviewed; B-lineage cell lines only; 'KMT2A-R B-ALL was resistant' is read from the abstract, and the number of resistant lines and their identities are not stated there. Treat as a flag to be re-checked on publication, not as a settled result. This is a direct challenge to a therapeutic premise and is deliberately graded weak.

   ▸Evidence that argues against this (1)

   • FHD-286 does kill MLL1-rearranged AML cells and prolongs xenograft survival in the myeloid setting, so if both results hold the determinant is lineage, not the rearrangement. Fiskus W et al. 2024
6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

   CORRECTED 2026-09-21, narrowed from 'any BAF subunit' to SMARCA4. No published experiment places SMARCA4 specifically on a KMT2A fusion protein, and in the one affinity-MS dataset that pulled down seven MLL fusions (PMID 29777171) SMARCA4 appears in a single bait column, MLL-CBP, and NOT in the MLL-AF4 column. The original broader claim -- that no BAF subunit has been placed on a KMT2A fusion -- was false: SMARCB1, SMARCC1 and SMARCC2 are all in the MLL-AF4 column of that same dataset. The PubMed count of 0 is accurate; the conclusion drawn from it was not.

   PubMed E-utilities esearch record counts: `BAF AND (MLL-AF4 OR MLL-AF9 OR KMT2A-AFF1) AND immunoprecipitation` = 0; positive controls through the identical code path `MLL-AF9 AND immunoprecipitation` = 28, `SWI/SNF AND immunoprecipitation` = 317, `leukemia` = 405,093; run 2026-09-20 · NCBI E-utilities, run for this beat on 2026-09-20 · PubMed 2026
   ▸What this evidence cannot show

   Bounded absence, and the bound was mis-read once already. Bounded to PubMed title/abstract/all-fields on 2026-09-20 over the terms shown. This caveat previously listed, as an uncovered case, 'supplementary mass-spectrometry tables not reflected in an abstract' -- which is precisely where the disproving data sat, in a file already on disk in this repository and in a paper cited by 51 entries here. Naming an escape hatch in a caveat does not close it, and this is rule 64 in its exact form: a literature declared empty while the data that disproves it sits in the same repo. The surviving SMARCA4-specific claim rests on a rank-truncated table (300 interactions per bait), so SMARCA4's absence from the MLL-AF4 column is a non-detection and not a demonstrated absence.

   ▸Evidence that argues against this (2)

   • A negative search is not a negative experiment. The closest positive in any fusion family is the EWSR1::FLI1 prion-like domain recruiting BAF, which was found by looking for it - nobody appears to have looked in KMT2A.
   • This repository's own analyses/mll-fusion-interactome-complexes/results.json places SMARCB1, SMARCC1 and SMARCC2 in the MLL-AF4 column of PMID 29777171, which refutes the original broad form of this claim outright. It also shows the complex-level enrichment is background-dependent: BAF scores 6 of 17 found against 0.82 expected, p = 9.4e-05 and q_BH = 1.9e-04 against the 20,431 proteome, but q = 0.0756 against a nuclear background of 5,754 and q = 0.1859 against the 4,600 proteins actually recovered. So BAF is NOT significant at a realistic background, and the analysis's own background_caveat says to read the ranking and treat the absolute p as an upper bound. The only pan-fusion BAF member is SMARCC1 - not SMARCA4.
7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   The only curated physical-interaction rows linking SMARCA4 to the menin axis are four IntAct rows from one mutation-directed BRET over-expression screen, annotated 'mutation with no effect', which this graph has already declined to treat as a menin partnership.

   Systematic discovery of mutation-directed neo-protein-protein interactions in cancer. · Mo X, Niu Q, Ivanov AA, Tsang YH, et al. · Cell 2022
   PMID 35512704DOI
   ▸What this evidence cannot show

   The paper is a screen of 17,792 pairs by BRET in live cells under co-expression; SMARCA4 sits in its null-effect class, meaning the tested variant did not change a signal, not that a partnership was demonstrated. No orthogonal method, no co-IP, no mapped region.

   ▸Evidence that argues against this (1)

   • A null-effect BRET row is genuinely ambiguous: it is equally consistent with a real constitutive association that the mutation does not modulate. It is weak evidence in either direction, which is why the correct action is to record it and not to build on it. Xiu Y et al. 2018
8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   FHD-286, a selective dual BRG1/BRM inhibitor, induces differentiation and lethality in AML with MLL1 rearrangement or mutant NPM1, represses c-Myc, PU.1 and CDK4/6, and reduces burden and improves survival in patient-derived xenografts, synergising with decitabine, BET inhibitor, menin inhibitor or venetoclax.

   BRG1/BRM inhibitor targets AML stem cells and exerts superior preclinical efficacy combined with BET or menin inhibitor. · Fiskus W, Piel J, Collins M, Hentemann M, et al. · Blood 2024
   PMID 38437498DOI
   ▸What this evidence cannot show

   Preclinical; myeloid only; single laboratory for the combination panel. 'Selective' means selective among kinases/ATPases, not selective for leukaemia over normal HSPC - the paper reports attenuated AML-initiating potential but a normal-marrow comparison is not the primary readout.

   ▸Evidence that argues against this (1)

   • The same compound class is reported ineffective in KMT2A-R B-ALL lines, so the myeloid result may not generalise to the B-lineage infant disease. Abhinav Ayyadevara VSS et al. 2025
9. E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   In B-lineage cells, BRG1/SMARCA4 is recruited into nuclear subdiffractive clusters formed by the pioneer factor EBF1, and four tyrosine-to-alanine substitutions in EBF1's intrinsically disordered region reduce both condensate formation in vitro and subdiffractive clusters in vivo, with loss of BRG1 recruitment and of target-gene activation - so SMARCA4's subnuclear position here is conferred by a partner, not intrinsic.

   Regularly spaced tyrosines in EBF1 mediate BRG1 recruitment and formation of nuclear subdiffractive clusters. · Zolotarev N, Wang Y, Du M, Bayer M, Grosschedl A, Cisse I, Grosschedl R · Genes Dev 2024
   PMID 38233109DOI
   ▸What this evidence cannot show

   This is a condensate claim and this graph's own standards apply: neither 1,6-hexanediol sensitivity nor classical FRAP distinguishes liquid-liquid phase separation from binding to spatially clustered sites, and the paper's in vitro condensate assay plus in vivo cluster imaging does not separate the two models either. 'Subdiffractive cluster' is the honest description and is what is recorded here. The system is EBF1-driven B-cell differentiation, not KMT2A-rearranged leukaemia.

   ▸Evidence that argues against this (1)

   • An orthogonal and simpler account of where SMARCA4 sits does not need clusters at all: SWI/SNF subunits act as mitotic bookmarks, remaining associated with chromosomes through mitosis. That is a chromosome-scale localisation established by a different method, and it predicts that the functionally important SMARCA4 pool is defined by the cell-cycle phase rather than by a partner's disordered region. Zhu Z et al. 2023
10. E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In AML with MLL1 rearrangement or mutant NPM1, the oral BRG1/BRM (SMARCA4/SMARCA2) inhibitor FHD-286 induces differentiation and lethality, with perturbed chromatin accessibility and repression of c-Myc, PU.1 and CDK4/6, and in patient-derived xenografts it reduces AML burden, improves survival and attenuates the AML-initiating potential of stem-progenitor cells; it synergises with decitabine, BET inhibitors, menin inhibitors and venetoclax.

    BRG1/BRM inhibitor targets AML stem cells and exerts superior preclinical efficacy combined with BET or menin inhibitor. · Fiskus W, Piel J, Collins M, Hentemann M, Cuglievan B, Mill CP, Birdwell CE, Das K, Davis JA, Hou H, Jain A, Malovannaya A, Kadia TM, Daver N, Sasaki K, Takahashi K, Hammond D, Reville PK, Wang J, Loghavi S, Sen R, Ruan X, Su X, Flores LB, DiNardo CD, Bhalla KN · Blood 2024
    PMID 38437498DOI
    ▸What this evidence cannot show

    FHD-286 is a DUAL BRG1/BRM inhibitor, so nothing here attributes the differentiation or stem-progenitor effect to SMARCA4 rather than SMARCA2 - and the graph's symbol-collision registry records that the two share a 92-character protein-name prefix and both alternative names contain 'BRG1', so this is the exact pair where misattribution is easiest. AML only: MLL1-rearranged and NPM1-mutant, no B-lineage and no KMT2A::AFF1 arm, and no infant material. Read: abstract only.

    ▸Evidence that argues against this (1)

    • A genome-scale CRISPR viability screen puts SMARCA4 in the ANTI-selective half for this disease: 92.1st percentile of 17,728 genes, mean Chronos -0.447 in 11 KMT2A-rearranged lines against -0.526 in 43 other leukaemia lines, difference +0.079. On fitness, SMARCA4 is if anything slightly less required in KMT2A-rearranged cells than in other leukaemias. The differentiation result and the fitness result are different readouts and need not agree, but the fitness half must be stated. Oncofusion knowledge graph, analyses/depmap-kmt2a-selective-scan-full; DepMap Chronos via Open Targets; 11 KMT2A-rearranged versus 43 other leukaemia lines, 17,728 genes 2026
11. E11moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In human MLL-rearranged leukaemia, SMARCA4 together with the SWI/SNF subunits SMARCD2/BAF60B and DPF2/BAF45D is selectively required for leukaemic cell expansion and self-renewal in vitro and in leukaemia, and all three are required for the expression of haematopoietic-stem-cell-associated genes - though, contrary to earlier mouse work, not for the expression of c-MYC-regulated genes.

    SWI/SNF Subunits SMARCA4, SMARCD2 and DPF2 Collaborate in MLL-Rearranged Leukaemia Maintenance. · Cruickshank VA, Sroczynska P, Sankar A, Miyagi S, Rundsten CF, Johansen JV, Helin K · PLoS One 2015
    PMID 26571505DOI
    ▸What this evidence cannot show

    Human cells with subunit knockdown and expression profiling; MLL-rearranged but the specific fusion partners are not resolvable from the abstract, and no KMT2A::AFF1 or infant arm. The paper explicitly contradicts the earlier murine claim that SWI/SNF sustains MLL-rearranged leukaemia through c-MYC, which is a reminder that species disagreement exists inside this literature. Read: abstract only.

    ▸Evidence that argues against this (1)

    • The later pharmacological study in MLL1-rearranged AML reports repression of c-Myc as a central consequence of BRG1/BRM inhibition, which is the opposite of the 'not required for the expression of c-MYC-regulated genes' conclusion drawn from human knockdown. The disagreement is unresolved and neither result can be quoted without the other. Fiskus W et al. 2024
12. E12established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    SMARCA4 IS A REMODELLER AND A READER, NOT A WRITER OR ERASER: UniProt P51532 (primary gene SMARCA4, 1,647 aa, fetched 2026-09-21) annotates a QLQ domain at 171-206, an HSA domain at 460-532, a helicase ATP-binding domain at 766-931 containing the DEGH box at 881-884, a helicase C-terminal domain at 1084-1246, and a bromodomain at 1453-1564, with Pfam cross-references including PF00439 (bromodomain) and no methyltransferase, demethylase, acetyltransferase or deacetylase family.

    UniProtKB reviewed entry P51532 (SMCA4_HUMAN): sequence length and domain, region and motif feature table with Pfam cross-references, fetched live from the UniProt REST API · UniProt Consortium · UniProtKB 2026
    ▸What this evidence cannot show

    A domain-architecture statement fetched live, not an experiment. An annotated bromodomain establishes the presence of an acetyl-lysine reader fold; it does NOT establish that SMARCA4's bromodomain reads a specific acetyl mark in cells, and the affinity of the BRG1 bromodomain for acetylated histone peptides is reported in the wider literature as weak - nothing read here measures it. The primary gene name was read back from the fetched record per this graph's accession-resolution rule, and 1,647 aa discriminates SMARCA4 from SMARCA2 at 1,590 aa, which rule 69 part 3 flags as a 3.5% difference with protein names that both contain 'BRG1'.

    ▸Evidence that argues against this (1)

    • The reader-domain observation must not become the mechanism. The functional account of SWI/SNF in this disease is ATP-dependent accessibility, not acetyl-mark reading: PU.1 binds most of its targets SWI/SNF-independently and recruits SWI/SNF to open chromatin for RUNX1, LMO2 and MEIS1, and the clinical-stage molecules in this class - FHD-286, BRM014 - are ATPase inhibitors, not bromodomain binders. So the bromodomain is present and is not where the pharmacology acts, and an entry that led with it would misdirect a target-selection reader. Chambers C et al. 2023
13. E13strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    SMARCA4's EPIGENETIC CURRENCY IS ACCESSIBILITY FOR A TRANSCRIPTION-FACTOR NETWORK THAT INCLUDES MEIS1: in AML, PU.1/SPI1 binds most of its targets in a SWI/SNF-independent manner and recruits SWI/SNF to promote accessibility for other AML core regulatory factors including RUNX1, LMO2 and MEIS1; SWI/SNF inhibition reduces DNA accessibility and the binding of those factors at PU.1 sites and redistributes PU.1 to promoters, induces leukaemic differentiation and reduces leukaemic stem cell burden in an immunocompetent mouse model and in primary human AML, and also impairs PU.1-dependent B-cell and monocyte populations with leukopenia.

    SWI/SNF Blockade Disrupts PU.1-Directed Enhancer Programs in Normal Hematopoietic Cells and Acute Myeloid Leukemia. · Chambers C, Cermakova K, Chan YS, et al. · Cancer Res 2023
    PMID 36662812DOI
    ▸What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche. MYELOID: an immunocompetent AML mouse model and primary human AML samples, with the non-tumour arm in normal haematopoietic cells - NOT a KMT2A-rearranged system and NOT infant B-lineage, so rule 73 applies and MEIS1's appearance here is as an AML core regulatory factor, not as a KMT2A-fusion target. The readout is DNA accessibility and factor binding, NOT a histone mark: no H3K27ac, H3K4me1 or H3K79me2 measurement is reported in the abstract, so nothing about covalent marks follows. 'SWI/SNF inhibition' is ATPase inhibition, which hits SMARCA2 as well as SMARCA4, so the effects are not attributable to SMARCA4 alone - the same paralogue problem this entry already records for FHD-286. Not cited anywhere in this corpus before 2026-09-21.

    ▸Evidence that argues against this (1)

    • The B-lineage direction is the one that matters for this project and it runs the other way. KMT2A-rearranged B-ALL cell lines are RESISTANT to the BRG1 inhibitors BRM014 and FHD-286 while Ph-like B-ALL lines are markedly sensitive, with elevated BRG1 expression in the sensitive group - already recorded on this entry as E5. So an accessibility mechanism demonstrated in AML, in which SWI/SNF opens chromatin for a network including MEIS1, does not predict sensitivity in the index disease's own lineage, and the off-tumour cost in that setting is specifically depletion of PU.1-dependent B cells. Abhinav Ayyadevara VSS et al. 2025
14. E14strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.analysis

    BAF IS NOT SIGNIFICANT IN THIS PROJECT'S OWN COMPLEX ANALYSIS AT A REALISTIC BACKGROUND, AND SMARCA4 IS ABSENT FROM THE INDEX FUSION'S HITS: BAF/SWI-SNF scores 6 of 17 members found against 0.82 expected by chance (p = 9.44e-05, q_BH = 1.89e-04) against a 20,431-protein proteome denominator, but q rises to 0.0756 against a nuclear background of 5,754 and to 0.1859 against the 4,600 proteins recovered before filtering; per bait, the MLL-AF4 (index fusion) BAF hits are SMARCB1, SMARCC1 and SMARCC2 with only 1 of 3 at p < 0.05, SMARCA4 is found in exactly one of the seven baits (MLL-CBP), and the only pan-fusion BAF member is SMARCC1.

    Complex-level enrichment over the MLL-fusion AP-MS interactome, recomputed with three alternative backgrounds: BAF/SWI-SNF 6 of 17 found, 0.82 expected, p = 9.44e-05, q_BH = 1.89e-04, q at nuclear background (5,754) = 0.0756, q at recovered background (4,600) = 0.1859; per-bait decomposition over seven MLL-fusion baits · Oncofusion knowledge graph, analyses/mll-fusion-interactome-complexes/run.py, over the AP-MS network of PMID 29777171 (network size 960) · Oncofusion knowledge graph (internal analysis) 2026
    ▸What this evidence cannot show

    This is a re-read of a stored analysis in this repository, not a fresh computation, and it inherits every caveat that analysis records - chief among them its own background_caveat: 'the honest denominator is the 4,600 proteins recovered before filtering; that list is not in the supplementary table. Using the proteome inflates significance for any complex of abundant nuclear proteins. Read the RANKING, and treat absolute p-values as an upper bound.' The underlying data are one AP-MS study (PMID 29777171) with a 960-protein network, so absence of SMARCA4 from the MLL-AF4 column is a negative in one experiment and not a demonstration of non-interaction; rule 93 applies, since a bait-specific detection failure is a common outcome in affinity proteomics. The 17-member BAF definition is the analysis's own subunit list and a different definition would move the numbers.

    ▸Evidence that argues against this (1)

    • The functional evidence is stronger than the proteomic evidence and points the other way, which is why this item bounds a claim about MODULE MEMBERSHIP and not about requirement. In human MLL-rearranged leukaemia, SMARCA4 together with the SWI/SNF subunits SMARCD2/BAF60B and DPF2/BAF45D is selectively required for leukaemic cell survival, and SMARCA4 depletion in human THP-1 (KMT2A::MLLT3) and MV4-11 (KMT2A::AFF1) cells causes loss of expansion, self-renewal and the undifferentiated state with human-cDNA rescue as the on-target control - already this entry's E2 and E11. A protein can be required by a leukaemia without being a component of its fusion complex, and conflating the two is the error this item exists to prevent in the opposite direction as well. Cruickshank VA et al. 2015

• binds →BENC - the MYC blood enhancer cluster 1.7 Mb downstream, required for MLL-AF9 leukaemia maintenance, and named in zero entries of this graph until now

  Brg1 occupies the cluster and is required there for TF occupancy and looping to the Myc promoter

• ← related toIn inv(16), disrupting the fusion interface swaps BRG1 for RING1B at three MYC enhancers - the cleanest version anywhere of the experiment nobody has run on menin

  BRG1 is the displaced factor, leaving 4-6 h after RING1B arrives

• ← related toThe MYC contradiction can be advanced but not closed: the enhancer is independently validated, the SWI/SNF route is not, and the two papers never measured the same thing

  residual question: does SWI/SNF hold the element in HUMAN KMT2A-rearranged cells

• related to →The case that the KMT2A-fusion condensate is an artefact — what 1,6-hexanediol, overexpressed tags and in-vitro droplets actually prove

  SMARCA4's only subnuclear record is an EBF1-driven subdiffractive cluster, and this entry sets what such a claim can and cannot support

• related to →CSNK2A1 (casein kinase 2 alpha) - the one kinase that turns up three separate times in KMT2A biology

  CK2-dependent phosphorylation of BRG1 occurs during mitosis, linking this tranche's kinase to this tranche's remodeller in the mitotic compartment

• regulates →The differentiation block — what it actually is, operationally, in KMT2A::AFF1 leukaemia

  BRG1/BRM inhibition induces differentiation in MLL1-rearranged and NPM1-mutant AML with repression of c-Myc, PU.1 and CDK4/6; DUAL inhibitor, so BAF-ATPase-level rather than SMARCA4-specific

• regulates →The stem-like self-renewal state — a limited stem-cell module grafted onto a committed cell

  SMARCA4 knockdown reduces leukaemic self-renewal and HSC-associated gene expression in human MLL-rearranged leukaemia; FHD-286 attenuates AML-initiating potential of stem-progenitor cells in MLL1r PDX

• ← related toSMARCA2 (BRM) - the paralogue nobody measured, and the reason a DepMap rank cannot predict what a BRG1/BRM inhibitor does

  Leukaemic BAF is assembled around BRG/SMARCA4 (PMID 24478402), so SMARCA2's nuclear-matrix pool may be near-empty in the cells of interest - the paralogue question gates any localisation experiment here (Q91 2026-09-22)

• related to →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here

  REMODELLER AND READER, NOT WRITER OR ERASER: P51532 carries QLQ, HSA, helicase ATP-binding (DEGH box 881-884), helicase C-terminal and a BROMODOMAIN at 1453-1564 (PF00439) - the only acetyl-lysine reader module in this tranche - and no methyltransferase, demethylase, acetyltransferase or deacetylase family. Its currency is ATP-dependent accessibility, not covalent modification.

• related to →MEIS1 — rate-limiting for MLL leukemia stem cells, a surviving co-dependency, and embryonic-lethal when removed

  ACCESSIBILITY, NOT A MARK: PU.1 recruits SWI/SNF to open chromatin for the AML core regulatory factors RUNX1, LMO2 and MEIS1, and SWI/SNF inhibition reduces accessibility and their binding at PU.1 sites. AML, not the index fusion, and no histone mark was measured.

• related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)

  NEGATIVE, AND MEASURED IN THIS REPOSITORY: in the MLL-fusion AP-MS complex analysis the MLL-AF4 bait's BAF hits are SMARCB1, SMARCC1 and SMARCC2 - SMARCA4 is absent, appearing in only one of seven baits (MLL-CBP) - and BAF as a complex reaches q = 0.076 at a nuclear background and 0.186 at the recovered background. BAF is not an established module of this fusion.

• part of →"SWI/SNF" is three different machines: cBAF, PBAF and ncBAF - and only one of them looks KMT2A-selective

  shared ATPase of all three mSWI/SNF assemblies

• depends on →KMT2A-rearranged acute myeloid leukaemia in children and infants

  maintenance dependency in MLL-AF9 and KMT2A-rearranged myeloid lines (shRNA, cDNA rescue)

• related to →MEN1 / menin — the adaptor the fusion cannot work without

  only curated interaction rows are null-effect BRET from one mutation screen; NOT a menin partner

• supports →Ten of menin's 33 IntAct partners come from one BRET screen for MUTATION-created interactions — and five are gains caused by a named allele

  SMARCA4 is one of the five null-effect rows that entry declines

• related to →The menin-inhibitor-resistant state has its own address: SMARCA4, BRD4 and CREBBP, with no MEN1 mutation anywhere

  SMARCA4 is a CRISPR-screen co-dependency of the menin-inhibitor-resistant state

• related to →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)

  requirement shown in THP-1 and NOMO-1 (KMT2A::MLLT3) and mouse MLL-AF9

• ← co-occurs withSMARCD2 (BAF60B) - named in one KMT2A paper in the whole of PubMed, and the dependency it carries is myeloid identity, not the rearrangement

  co-required with SMARCA4 and DPF2 in the anchor paper

• ← co-occurs withDPF2 (BAF45D) - a cBAF-specific reader required in MLL-AF9 cells, with one PubMed record in this disease and no selectivity signal

  third subunit of the anchor paper's trio

• ← co-occurs withARID1A (BAF250A) - required in the MLL-AF9 screen, present at the Myc enhancers, and flat in every selectivity measure

  both occupy the E1-E5 distal Myc enhancers in MLL-AF9 RN2 cells

• ← depends onHow BAF maintains KMT2A-rearranged leukaemia: it holds the blast undifferentiated, and it does so in other leukaemias too

  the ATPase whose loss produces the phenotype

• ← related toIs MYC the node? Two papers, the same MLL-AF9 model, opposite answers - and the second one failed to rescue

  the contradiction is about the same ATPase in both papers

• ← inhibitsFHD-286 - the dual BRG1/BRM inhibitor that carries every SWI/SNF therapeutic claim in this disease, with a minimum enrolment age of 16

  ATPase inhibitor; primary target

• ← refutesSWI/SNF split by sub-complex and by lineage: every KMT2A-selective subunit is in the ncBAF module and none is in canonical BAF

  no SMARCA4 selectivity in either lineage half