ExplainerstrongQ47-swi-snf

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

also called mSWI/SNF · BAF complex · GBAF · BRG1/BRM-associated factor

SWI/SNF is not one machine but three related ones, built from a shared set of parts plus a few parts unique to each. Almost every sentence written about "SWI/SNF in leukaemia" is really about one of the three, and this graph had been treating the name as a single undifferentiated thing.

The same thing, for a scientist

Mammalian SWI/SNF exists as three final-form assemblies that share a core (SMARCA4/SMARCA2 ATPase, SMARCC1/2, SMARCD1-3, ACTL6A, SS18) and are distinguished by mutually exclusive subunits: **cBAF** carries ARID1A/ARID1B plus DPF1/2/3 and SMARCB1/SMARCE1; **PBAF** carries ARID2, PBRM1, PHF10 and BRD7; **ncBAF** (originally GBAF) carries BICRA/GLTSCR1 or BICRAL and BRD9 and, critically, **lacks SMARCB1, SMARCE1 and the DPF/BAF45 subunits entirely** [E1][E2][E3]. The three have different chromatin addresses - ncBAF uniquely localises to CTCF sites and promoters - and different disease dependencies [E3]. This distinction is load-bearing for this beat: a lineage-split DepMap scan of 28 SWI/SNF genes run for this beat finds that the subunits with a KMT2A-rearranged dependency in **both** the myeloid and the B-lineage halves of the index arm are **BRD9, BICRA, SS18 and SMARCD1** - three of the four are ncBAF-defining or ncBAF-shared, and **no cBAF-defining or PBAF-defining subunit survives** [E4]. A claim about "SWI/SNF" that does not say which complex is therefore not yet a claim.

Go deeper

What is shared and what is not

Shared across all three: the ATPase (SMARCA4/BRG1 or SMARCA2/BRM, mutually exclusive), SMARCC1/SMARCC2 (BAF155/BAF170), a SMARCD/BAF60 paralogue, ACTL6A/BAF53A, actin and SS18 12.

cBAF: + ARID1A or ARID1B (BAF250A/B), + a DPF/BAF45 paralogue, + SMARCB1 (BAF47), + SMARCE1 (BAF57).
PBAF: + ARID2 (BAF200), + PBRM1 (BAF180), + PHF10 (BAF45A), + BRD7; also retains SMARCB1/SMARCE1.
ncBAF / GBAF: + BICRA (GLTSCR1) or BICRAL, + BRD9; no BAF45, no BAF47, no BAF57 2.

The ncBAF composition was defined by density-gradient separation and immunoprecipitation, and the complex has ATPase activity and bulk chromatin affinity comparable to BAF despite missing those core subunits 2. Mashtalir et al. then mapped the assembly pathway of all three by cross-linking mass spectrometry of endogenous complexes 1.

Why the distinction changes the answer in this disease

Run over the same 11 KMT2A-rearranged and 43 other-leukaemia lines, the genome-wide scan this project already holds puts SMARCA4 at rank 16,333 of 17,728 with a positive difference (+0.0791: marginally less required in KMT2A-rearranged lines than in other leukaemias), ARID1A at 12,593 (+0.0276) and SMARCB1 at 14,024 (+0.0415); against BRD9 at 39, SS18 at 224, SMARCD1 at 59 and BICRA at 517 5. Splitting the index arm by lineage, only SMARCD1, BRD9, BICRA and SS18 are depleted in both halves 4.

So the honest summary of the dependency data is: canonical BAF is required by leukaemia cells and is not KMT2A-selective; the ncBAF module is the part with a KMT2A-rearranged signature. Everything the older literature calls a "SWI/SNF requirement in MLL-rearranged leukaemia" was measured before ncBAF had a name (2018), so none of it was designed to separate the two.

Facts and figures

why the older literature cannot settle it	ncBAF was named in 2018 (PMIDs 29374058, 30397315, 30343899). Every MLL-rearranged SWI/SNF paper before that date - including the anchor PMID 26571505 (2015) and PMID 24285714 (2013) - knocked down subunits that are shared between complexes (SMARCA4, SMARCD2, SMARCC1) or are cBAF-specific, and none could attribute a phenotype to a complex.

complexes

[
  {
    "name": "cBAF (canonical BAF)",
    "defining_subunits": [
      "ARID1A",
      "ARID1B",
      "DPF1",
      "DPF2",
      "DPF3"
    ],
    "also_contains": [
      "SMARCB1",
      "SMARCE1"
    ],
    "shared_core": [
      "SMARCA4/SMARCA2",
      "SMARCC1",
      "SMARCC2",
      "SMARCD1-3",
      "ACTL6A",
      "SS18",
      "ACTB"
    ]
  },
  {
    "name": "PBAF (polybromo-BAF)",
    "defining_subunits": [
      "ARID2",
      "PBRM1",
      "PHF10",
      "BRD7"
    ],
    "also_contains": [
      "SMARCB1",
      "SMARCE1"
    ],
    "shared_core": [
      "SMARCA4/SMARCA2",
      "SMARCC1",
      "SMARCC2",
      "SMARCD1-3",
      "ACTL6A",
      "ACTB"
    ]
  },
  {
    "name": "ncBAF (non-canonical BAF, originally GBAF)",
    "defining_subunits": [
      "BICRA",
      "BICRAL",
      "BRD9"
    ],
    "absent_subunits": [
      "SMARCB1",
      "SMARCE1",
      "DPF1/2/3 (BAF45)"
    ],
    "shared_core": [
      "SMARCA4/SMARCA2",
      "SMARCC1",
      "SMARCD1",
      "ACTL6A",
      "SS18"
    ],
    "chromatin_address": "CTCF sites and promoters"
  }
]

kmt2a selectivity by module

{
  "cBAF_defining": "no selectivity: ARID1A rank 12593 (+0.0276), ARID1B 11527 (+0.0185), DPF2 1321 (-0.0927), DPF1 4166, DPF3 9884",
  "cBAF_core": "no selectivity: SMARCA4 16333 (+0.0791), SMARCB1 14024 (+0.0415), SMARCC1 8543 (-0.0046), SMARCE1 7543 (-0.0123), ACTL6A 17410 (+0.1454)",
  "PBAF_defining": "no selectivity: ARID2 8209 (-0.0072), PBRM1 2137, PHF10 5718, BRD7 1949",
  "ncBAF_defining_or_shared": "BRD9 rank 39 (-0.3896), SS18 224 (-0.2124), BICRA 517 (-0.1440), SMARCD1 59 (-0.3574)"
}

How this could bite us back

Because ncBAF and cBAF share the ATPase, a BRG1/BRM inhibitor hits all three complexes at once. Any therapeutic rationale built on ncBAF selectivity is not delivered by the drug class that is currently in the clinic for this disease.
Conversely, a BRD9 degrader spares cBAF and PBAF - so if the KMT2A-rearranged dependency is really the shared ATPase, a BRD9-only agent will underperform its DepMap rank.

Still open

Which of the three complexes is present on KMT2A-fusion-bound chromatin in a KMT2A-rearranged blast? No complex-resolved ChIP or CUT&RUN in this disease has been published as far as this beat could find.
Is the BRD9/BICRA/SS18 dependency a genuine ncBAF signal or three separate reflections of one thing the lines need (for example MYC)?
Does the ncBAF signature hold in infant KMT2A-rearranged B-ALL, which is absent from DepMap entirely?

Every source behind this entry (5)

E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Mammalian SWI/SNF exists as three final-form assemblies - cBAF, PBAF and ncBAF - whose modular organisation and assembly pathway were mapped by cross-linking mass spectrometry of endogenous complexes.
Modular Organization and Assembly of SWI/SNF Family Chromatin Remodeling Complexes. · Mashtalir N, D'Avino AR, Michel BC, Luo J, et al. · Cell 2018
PMID 30343899 DOI
What this evidence cannot show
Biochemistry in mammalian and Drosophila cell lines, not in leukaemia blasts; the assembly pathway is inferred from subunit-deletion stability rather than watched in real time.
Evidence that argues against this (1)
- The three-complex partition is a description of steady-state purified assemblies. Whether an intact cell contains only these three, or a continuum with these as the abundant species, is not settled by affinity purification; sub-stoichiometric and transitional species would be invisible. Alpsoy A et al. 2018
E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
ncBAF (GBAF) contains BICRA/GLTSCR1 or BICRAL plus BRD9 and lacks BAF45, BAF47 and BAF57, while retaining BAF155, a BAF60, SS18, BAF53a and BRG1/BRM.
Glioma tumor suppressor candidate region gene 1 (GLTSCR1) and its paralog GLTSCR1-like form SWI/SNF chromatin remodeling subcomplexes. · Alpsoy A, Dykhuizen EC · J Biol Chem 2018
PMID 29374058 DOI
What this evidence cannot show
Defined in prostate cancer lines (LNCaP, PC3) by density gradient and IP; composition in haematopoietic cells is assumed rather than shown here.
Evidence that argues against this (1)
- The same authors report GLTSCR1 is dispensable for LNCaP viability while required in PC3, so ncBAF composition does not predict ncBAF dependency; cell type decides. Xiu Y et al. 2018
E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
ncBAF complexes uniquely localise to CTCF sites and promoters and are synthetic-lethal targets in synovial sarcoma and malignant rhabdoid tumour, both of which have perturbed cBAF.
A non-canonical SWI/SNF complex is a synthetic lethal target in cancers driven by BAF complex perturbation. · Michel BC, D'Avino AR, Cassel SH, Mashtalir N, et al. · Nat Cell Biol 2018
PMID 30397315 DOI
What this evidence cannot show
The synthetic-lethal relationship is demonstrated in cBAF-perturbed cancers (SMARCB1 loss or SS18-SSX), which KMT2A-rearranged leukaemia is not; transferring it to this disease is an analogy, not a result.
Evidence that argues against this (1)
- In synovial sarcoma the BRD9 dependency is specifically tied to BRD9 being a component of SS18-SSX-containing BAF complexes; no equivalent physical association has been shown for any KMT2A fusion. Brien GL et al. 2018
E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In a lineage-split DepMap panel of 28 SWI/SNF genes, only SMARCD1, BRD9, BICRA and SS18 are depleted in both the myeloid and B-lineage halves of the KMT2A-rearranged arm; no cBAF-defining or PBAF-defining subunit is.
SWI/SNF panel lineage-split DepMap dependency scan, 28 genes · Oncofusion knowledge graph, analyses/swi-snf-lineage-split/run.py · DepMap Chronos gene effect via Open Targets GraphQL 2026
What this evidence cannot show
Eight myeloid and five B-lineage index lines, none infant-derived, so one line moves a mean. Split is by lineage label, not expression signature. z is in SDs of the other-leukaemia arm and is not an effect size; report the raw difference alongside. SMARCD2 is a myeloid-identity artefact by the same test (myeloid z -1.61, B-lineage -0.04).
Evidence that argues against this (1)
- A dependency screen measures what a cell line needs, not what a complex does; BRD9 and BICRA are near-zero pan-essential (+0.04, +0.09), which is what gives them a large z from a modest effect. The genome-wide ranking of SMARCA4 is also consistent with SMARCA4 simply being required everywhere, which the same table shows (pan-essentiality -0.36). Oncofusion knowledge graph et al. 2026
E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Genome-wide, SMARCA4 ranks 16,333/17,728 with difference +0.0791 (KMT2A-r mean -0.4469, other leukaemia -0.5260), while BRD9 ranks 39 (-0.3896), SMARCD1 59 (-0.3574), SS18 224 (-0.2124) and BICRA 517 (-0.1440).
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 KMT2A-rearranged against 43 other leukaemia lines, Cellosaurus fusion annotation may be incomplete; DepMap release version not stated by the API. A positive difference means less required in the index arm, not 'not required' - SMARCA4's KMT2A-r mean is -0.45, a real dependency with no selectivity.
Evidence that argues against this (1)
- Pharmacological BRG1/BRM inhibition does kill MLL1-rearranged AML cells and prolongs xenograft survival, which a genetic SMARCA4 knockout ranking does not predict - because a dual ATPase inhibitor removes the SMARCA2 paralogue at the same time. Fiskus W et al. 2024

Connected to

← part ofARID1B (BAF250b) - the cBAF paralogue whose deletion PROMOTES MLL-AF9 leukaemogenesis, which makes it the one gene in this beat whose published KMT2A-fusion result points the wrong way for a drug target
one of the two mutually exclusive ARID1 subunits that DEFINE canonical BAF
← part ofSMARCC1 / BAF155 (Q92922) - the SWI/SNF core scaffold, in the pan-MLL-fusion core, and the missing gene entry in this graph's SWI/SNF beat
core scaffold shared by cBAF, PBAF and ncBAF, so the row cannot attribute a phenotype to any one complex
related to →BRD9 — dispensable in the average cell, rank 39 of 17,728 in KMT2A-rearranged lines, and with an on-target proof most epigenetic targets never get
BRD9 is the ncBAF-defining bromodomain subunit; entry already in the graph
part of →Plain-English: chromatin and epigenetics — the sticky notes on the DNA, and which ones matter here
ATP-dependent remodelling is one arm of chromatin regulation
related to →KMT2A-rearranged acute myeloid leukaemia in children and infants
module-level dependency profile measured on KMT2A-rearranged lines
← part ofSMARCA4 (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
shared ATPase of all three mSWI/SNF assemblies
← part ofSMARCA2 (BRM) - the paralogue nobody measured, and the reason a DepMap rank cannot predict what a BRG1/BRM inhibitor does
alternative shared ATPase of all three assemblies
← part ofSMARCD2 (BAF60B) - named in one KMT2A paper in the whole of PubMed, and the dependency it carries is myeloid identity, not the rearrangement
BAF60 slot, shared across all three assemblies
← part ofDPF2 (BAF45D) - a cBAF-specific reader required in MLL-AF9 cells, with one PubMed record in this disease and no selectivity signal
BAF45 slot; cBAF only, absent from ncBAF
← part ofSMARCD1 (BAF60A) - rank 59 of 17,728, survives the lineage split, and had never been named in this graph or studied in this disease
BAF60 slot, shared across all three assemblies
← part ofSMARCB1 (BAF47/INI1/SNF5) - zero entries, zero mentions and, on the evidence, correctly so: it is lost in primary AML rather than required
cBAF and PBAF core; absent from ncBAF
← part ofARID1A (BAF250A) - required in the MLL-AF9 screen, present at the Myc enhancers, and flat in every selectivity measure
cBAF-defining ARID slot
← part ofARID2 (BAF200) - the PBAF subunit whose loss ACCELERATES MLL-AF9 leukaemia, and which is nonetheless required to maintain it
PBAF-defining ARID protein
← part ofBICRA (GLTSCR1) - the ncBAF scaffold, zero PubMed records in KMT2A leukaemia, and a lineage-independent dependency in KMT2A-rearranged lines
ncBAF-defining scaffold
← part ofSS18 - a SWI/SNF subunit that is itself a fusion partner in another cancer, and a lineage-independent dependency in KMT2A-rearranged lines
shared between cBAF and ncBAF
← supportsSWI/SNF split by sub-complex and by lineage: every KMT2A-selective subunit is in the ncBAF module and none is in canonical BAF
the module-level result the complex distinction is needed to state
← depends onThe SWI/SNF dependency in KMT2A-rearranged leukaemia belongs to ncBAF, not to canonical BAF - and the drug in the clinic is aimed at the wrong module
the complex distinction the hypothesis is stated in