ARID1A (BAF250A) - required in the MLL-AF9 screen, present at the Myc enhancers, and flat in every selectivity measure

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.

• ARID1A is a tumour suppressor across epithelial cancers; any agent that destabilises cBAF in a child is an ARID1A-loss-mimicking intervention with a known oncogenic phenotype in other tissues.
• Because Arid1b and Arid2 deletion each PROMOTE MLL-AF9 leukaemogenesis, an intervention that degrades the complex non-selectively could accelerate the disease it is meant to treat - and no published experiment has looked for that with a pan-SWI/SNF agent.
• The paralogue pair means a selective ARID1A agent would be immediately buffered by ARID1B, whose loss is itself pro-leukaemic.

1. Does ARID1A loss phenocopy ARID1B loss in MLL-AF9 - i.e. does it also promote leukaemogenesis - or are the two paralogues genuinely opposite? Nobody has deleted them side by side in one model.
2. Is ARID1A's presence at the E1-E5 Myc enhancers required for their activity, or incidental to a complex whose ATPase is the active part?
3. In human KMT2A-rearranged B-ALL, are ARID1A mutations ever selected at relapse? The co-mention cohorts that generate the 35-record count have never been re-analysed for that.

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

   ARID1A is UniProt O14497, 2,285 aa, primary gene ARID1A, synonyms BAF250 BAF250A C1orf4 OSA1 SMARCF1; it is a cBAF-defining subunit mutually exclusive with ARID1B.

   UniProtKB O14497 (ARI1A_HUMAN) · UniProt Consortium · UniProtKB 2026
   ▸What this evidence cannot show

   gene_exact:ARID1A AND organism_id:9606 AND reviewed:true, run 2026-09-20; one reviewed row; gene_primary and length read back from the fetched record.

   ▸Evidence that argues against this (1)

   • The synonym SMARCF1 places ARID1A inside the SMARC* naming family, so an alias-matching lookup on SMARC symbols can pull ARID1A in beside subunits it is not equivalent to. Mashtalir N et al. 2018
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   BAF250a (ARID1A) was a requirement for MLL-AF9/NrasG12D RN2 proliferation in the 93-hairpin SWI/SNF shRNA screen, and BAF250a occupancy was detected at the E1-E5 distal Myc enhancer region in those cells.

   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

   Mouse line; occupancy shown by ChIP in a supplementary figure alongside BAF60a, establishing the presence of a SWI/SNF complex rather than an ARID1A-specific function.

   ▸Evidence that argues against this (1)

   • The same paper concludes that most direct MLL-AF9 target genes do NOT respond to Brg1 loss, so a subunit's presence at the Myc enhancers is not evidence it acts on the fusion's own programme. Xiu Y et al. 2018
3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

   ARID1A ranks 12,593 of 17,728 with a positive difference (+0.0276; KMT2A-r -0.4118, other leukaemia -0.4394, pan-essentiality -0.2337) and shows no depletion in either lineage half (myeloid z +0.04, B-lineage +0.19).

   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 against 43 comparator lines; an absence of selectivity, not an absence of dependency - the index mean of -0.41 is a real requirement shared with other leukaemias.

   ▸Evidence that argues against this (1)

   • ARID1B can substitute in the same slot, so a single-gene knockout may be buffered; both score near zero, which is consistent either with redundancy or with neither being limiting. Bluemn T et al. 2022
4. E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

   A mutant Arid1a allele leads to haematopoietic stem cell expansion through a non-cell-autonomous mechanism, so ARID1A perturbation in the haematopoietic system is not straightforwardly growth-limiting.

   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

   Carried from Shi et al.'s introduction, citing Krasteva et al. 2012, which this beat did NOT retrieve or read. Graded weak for that reason; fetch the primary before relying on it.

   ▸Evidence that argues against this (1)

   • A non-cell-autonomous stem-cell phenotype in a mouse says nothing about the requirement of ARID1A inside an established human leukaemic blast, which is the question here. Xiu Y et al. 2018
5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Conditional deletion of Arid1b in a mouse MLL-AF9 model promotes leukaemogenesis at both initiation and maintenance, while Arid2 has dual roles - enhancing leukaemogenesis when deleted at initiation yet required during maintenance; Arid2 and Arid1b regulate largely distinct gene sets and converge oppositely on Etv4/Etv5.

   Differential roles of BAF and PBAF subunits, Arid1b and Arid2, in MLL-AF9 leukemogenesis. · Bluemn T, Schmitz J, Zheng Y, Burns R, et al. · Leukemia 2022
   PMID 35022500DOI
   ▸What this evidence cannot show

   Mouse conditional knockouts; ARID1A itself was not part of this study, so the contrast with ARID1A is across papers and models rather than within one experiment. Disease-stage dependence is the paper's own headline and means any single-timepoint result can be misleading.

   ▸Evidence that argues against this (1)

   • If Arid1b deletion promotes leukaemogenesis and ARID1A is required, the ARID slot's two occupants oppose each other, and a pan-cBAF intervention would do both things at once with an unpredictable net effect. Shi J et al. 2013
6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

   `ARID1A AND (MLL OR KMT2A)` returns 35 PubMed records, dominated by sequencing cohorts that name both genes in a mutation table rather than studying the relationship.

   PubMed E-utilities esearch record counts: `ARID1A AND (MLL OR KMT2A)` = 35; `ARID1A AND leukemia` = 56; alias `BAF250A AND leukemia` = 16; complex form `SWI/SNF AND (MLL OR KMT2A)` = 27 run per rule 64; operand controls large; run 2026-09-20 · NCBI E-utilities, run for this beat on 2026-09-20 · PubMed 2026
   ▸What this evidence cannot show

   Bounded absence about STUDY DESIGN, not about record count. The 35 were judged from titles in the esearch listing, not fetched; three titles in the adjacent SMARCA4 list were fetched in full and all three were noise (non-small-cell lung cancer, NRG1 variants, Burkitt lymphoma), which is what motivates the judgement. A co-mention cohort could still contain an ARID1A-by-KMT2A interaction analysis in a supplementary table.

   ▸Evidence that argues against this (1)

   • The functional ARID1A results in this disease (PMID 24285714) use BAF nomenclature only and are not reachable from the ARID1A symbol, so a symbol-keyed count both over-counts noise and under-counts signal. Shi J et al. 2013
7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

   ARID1A/ARID1B define cBAF and ARID2 defines PBAF; the two are in different complexes rather than being interchangeable.

   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 29374058DOI
   ▸What this evidence cannot show

   Biochemical separation in prostate cancer lines; the partition is treated as general and has not been re-derived in haematopoietic cells.

   ▸Evidence that argues against this (1)

   • The third complex, ncBAF, contains no ARID protein at all, so 'the ARID slot' is a property of two of the three assemblies only. Michel BC et al. 2018
8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Loss of Baf200 (ARID2) in a mouse MLL-AF9 leukaemogenesis model accelerates tumour burden and shortens host survival, while Baf200 is required for normal fetal and adult HSC function.

   The chromatin remodeling subunit Baf200 promotes normal hematopoiesis and inhibits leukemogenesis. · Liu L, Wan X, Zhou P, Zhou X, et al. · J Hematol Oncol 2018
   PMID 29482581DOI
   ▸What this evidence cannot show

   Mouse conditional knockouts (Tie2-Cre, Vav-iCre, Mx1-Cre) with fetal liver and bone marrow transplantation; the leukaemia arm is a single MLL-AF9 model and the readout is burden and survival, not a defined molecular mechanism.

   ▸Evidence that argues against this (1)

   • Bluemn et al. report Arid2 as REQUIRED during MLL-AF9 maintenance while enhancing leukaemogenesis when deleted at initiation, so the direction of the Arid2 effect depends on when it is removed and the two papers are only reconcilable through disease stage. Bluemn T et al. 2022
9. E_q95_arid1a_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   ARID1A's compartment is well sourced and its determinant is mapped: UniProt O14497 carries 'Nucleus' on ECO:0000255 PROSITE-ProRule PRU00355 plus two ECO:0000269 citations, and a nuclear localization signal at residues 1368-1387; the 2016 citation is a dedicated localisation paper that identified a previously undescribed BIPARTITE-LIKE NLS contributing to ARID1A nuclear import, confirmed functionally with GFP constructs fused to wild-type or mutant NLS sequences, and showed that cyto-nuclear-localised bipartite-NLS-mutant ARID1A is MORE STABLE than nuclear wild-type ARID1A.

   UniProtKB O14497 (ARI1A_HUMAN), SUBCELLULAR LOCATION comments plus the NLS motif at 1368-1387 with evidence codes and sources, fetched live 2026-09-22 · UniProt Consortium · Databases 2026
   ▸What this evidence cannot show

   The GFP-NLS-fusion work is construct-based and the stability comparison is between a mutant and wild type rather than between compartments of the endogenous protein. The other experimental citation, PubMed 11318604, is the cDNA cloning and characterisation of mammalian orthologues of the Drosophila osa gene with direct physical interaction with the Brahma complex - a legitimate but incidental source for a compartment.

   ▸Evidence that argues against this (1)

   • THE STABILITY RESULT IS THE INTERESTING PART AND IT ARGUES THAT COMPARTMENT AND DOSE ARE COUPLED FOR THIS TUMOUR SUPPRESSOR, WHICH MATTERS FOR READING PATIENT MISSENSE ALLELES. If mislocalised ARID1A is MORE stable than nuclear ARID1A, then total protein level is a poor proxy for functional ARID1A, and an immunohistochemistry score that counts protein without resolving compartment can read a functionally null tumour as ARID1A-proficient. That is directly relevant to this graph's sarcoma material, where ARID1A mutations were called by sequencing in both VIM::KMT2A cases and the variant-level detail sits in a supplementary table nobody in this project has opened - so neither the allele class nor whether any of them is an NLS-region missense is currently known here. Massoth LR et al. 2020
10. E_q95_arid1a_loc2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    ARID1A has a 2026-reported SUBNUCLEAR ARCHITECTURAL function that is explicitly SWI/SNF-INDEPENDENT: ARID1A safeguards HETEROCHROMATIN ARCHITECTURE outside the SWI/SNF complex by interacting with TRIM28 to preserve heterochromatin rigidity; loss of ARID1A protein displaces SETDB1 from the TRIM28-containing heterochromatin complex, reversing H3K9me3-mediated repression at endogenous retroelement regions and triggering viral mimicry with type I interferon responses, and disrupting the ARID1A-TRIM28 interaction with synthetic peptides reproduces the phenotype in ARID1A wild-type tumours.

    ARID1A Mediates SWI/SNF-Independent Maintenance of Heterochromatin Architecture to Restrain Viral Mimicry and Immunogenicity in Colon Cancer. · Li Q, Zhang Z, Wang Y, Peng X, Fang Y, Zhang Y, Chen L, Huang T, Yang Z, Li C, Li L, Mills GB, Shen X, Wang H, Shen J · Cancer Res 2026
    PMID 41661654DOI
    ▸What this evidence cannot show

    Colorectal cancer, published 2026, and a single-group result that has not been independently replicated; the mechanism is inferred from displacement and peptide-disruption experiments rather than from imaging heterochromatin directly. It is nonetheless the only statement in the literature that gives ARID1A a nuclear ADDRESS as opposed to a complex membership, and the direction - a BAF subunit acting away from BAF - is the opposite of the assumption a chromatin graph would make.

    ▸Evidence that argues against this (1)

    • AND THIS IS WHY THIS ENTRY MUST NOT IMPLY THAT BAF IS AN ESTABLISHED MODULE IN THIS DISEASE - THE PROJECT'S OWN ANALYSIS SAYS IT IS NOT, AND ARID1A IS NOT EVEN IN IT. In this project's complex-enrichment analysis over the 960-protein MLL-fusion interactome, BAF/SWI-SNF is found at 6 of 17 members with p = 9.4e-05 against the proteome but q = 0.076 against a nuclear background and 0.186 against the 4,600 proteins actually recovered - not significant at a realistic background, and the analysis's own caveat says to read the ranking rather than absolute p-values. The six members recovered are PBRM1, SMARCA4, SMARCB1, SMARCC1, SMARCC2 and SMARCE1. ARID1A IS NOT ONE OF THEM. So ARID1A reaches this graph through a co-mutation in sarcoma and an MLL-AF9 screen, not through the fusion interactome, and a SWI/SNF-independent heterochromatin function is a better fit to that evidential position than a BAF-subunit framing would be. Oncofusion knowledge graph project 2026
11. E_q95_arid1a_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    ARID1A's only connection to a KMT2A rearrangement in a patient is a CO-MUTATION in sarcoma, not a localisation or an interaction: both VIM::KMT2A spindle-to-round-cell sarcomas in a 14,680-sarcoma genomic profiling cohort carried concurrent CTNNB1, SMARCB1 and ARID1A mutations, with tumour mutational burdens of 0.8 and 1.6 mutations per megabase - so two of the three co-mutated genes are BAF/SWI-SNF components in a tumour whose fusion retains a catalytically intact KMT2A SET domain.

    Pan-sarcoma genomic analysis of KMT2A rearrangements reveals distinct subtypes defined by YAP1-KMT2A-YAP1 and VIM-KMT2A fusions. · Massoth LR, Hung YP, Nardi V, Nielsen GP, Hasserjian RP, Louissaint A Jr, Fisch AS, Deshpande V, Zukerberg LR, Lennerz JK, Selig M, Glomski K, Patel PJ, Williams KJ, Sokol ES, Alexander BM, Vergilio JA, Ross JS, Pavlick DC, Chebib I, Williams EA · Mod Pathol 2020
    PMID 32461620DOI
    ▸What this evidence cannot show

    n = 2, in a commercial profiling cohort enriched for advanced disease, so neither the frequency nor the co-occurrence pattern is a population estimate; the variant-level detail is in a supplementary table that has not been opened in this project, so the specific alleles, their hotspot status and their zygosity are unknown here. The fusion partner is VIM::KMT2A in adult soft-tissue sarcoma, not KMT2A::AFF1 in infant B-ALL.

    ▸Evidence that argues against this (1)

    • THE CO-MUTATION CUTS AGAINST TREATING THE FUSION AS THE SOLE DRIVER AND THAT IS A DIFFERENCE FROM THE INDEX DISEASE, NOT A SUPPORT FOR IT. In infant KMT2A::AFF1 the fusion is frequently the only recurrent lesion; here three pathogenic mutations accompany it at a TMB under 2 per megabase, which is a striking combination and argues that the sarcoma arm needs cooperating chromatin lesions where the leukaemia arm does not. Pushed the other way, this weakens rather than strengthens any attempt to import ARID1A biology into the leukaemia: the one clinical setting where ARID1A and a KMT2A rearrangement co-occur is the setting where the fusion is demonstrably insufficient. At n = 2 it is a pattern to test and not a finding, which is what the sarcoma entry itself says. Li Q et al. 2026
12. E_q95_arid1a_loc4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Bounded 2026-09-22: no ARID1A compartment has been measured in any KMT2A-fusion context, and the five apparent hits are all off topic. The gene clause over ARID1A, BAF250, BAF250A, SMARCF1 and 'p270' returns 2,593 records and intersects the nine-term KMT2A/MLL fusion clause at 33; with a localisation clause the gene clause returns 193 and with the fusion clause added 5 - title-checked to 0 on topic, being sequencing, immunophenotyping and methylation surveys where MLL appears as a mutated gene name; with a subnuclear clause it returns 8 and with the fusion clause added 0. The NLS sub-query returns 10 and the sarcoma sub-query 59.

    NCBI E-utilities esearch db=pubmed; every count stored with the querytranslation read back, in data.localization_bounded_absence · NCBI E-utilities · Databases 2026
    ▸What this evidence cannot show

    Titles and abstracts only; supporting figures not searched (rule 80). The alias set was expanded deliberately per brief rule 69 and the retired symbol SMARCF1 does reach at least one paper a current-symbol search would miss (PubMed 11159203, expression of a truncated SWI1 form in neuroblastoma) - so alias expansion helped here, in contrast to CBX8 in this same tranche where it inflated the query 117-fold. Per-gene hypothesis, not a rule.

    ▸Evidence that argues against this (1)

    • THE SUBNUCLEAR HIT LIST IS WHERE THE CONTAMINATION SHOWS, AND IT CHANGES HOW THE COUNT SHOULD BE READ. Of the 8 subnuclear records, the on-topic one is the 2026 heterochromatin-architecture paper; the rest include an LncSNHG5 lactylation study matching through 'small nucleolar RNA host gene', clinicopathological surveys, a case report and a 2001 neuroblastoma paper. So the informative subnuclear literature for a 2,285-residue tumour suppressor mutated across human cancer is ONE paper, published this year, from one group, in colorectal cancer. That is a thinner base than a gene of this prominence suggests, and it is the reason this entry's subnuclear field is written as a single-source claim rather than as an established address. Li Q et al. 2026

• ← supportsARID1B is not dispensable - it is paralogue-buffered. In ARID1A-mutant lines its CRISPR dependency goes from -0.0724 to -0.3709 (p=0.0002) while ARID1A's own falls away (+0.1925, p=0.044) and ARID2/SMARCA4 stay flat. Cycle 80's 'dispensable everywhere' framing was wrong and the caveat that entry carried was right

  instrument check: ARID1A's own dependency falls -0.2354 -> -0.0429 (+0.1925, p=0.044) in the lines Cellosaurus calls ARID1A-mutant, which is what validates the split

• ← related toIn St. Jude paediatric AML, 2 of the 6 ARID1A-mutant cases are KMT2A-rearranged and BOTH ARID1A mutations are TRUNCATING frameshifts in KMT2A::MLLT10 cases - so the exact combination cycle 86 showed the ARID1B dependency is confined to DOES occur in patients, and cycle 88's empty DepMap intersection is a cell-line sampling artefact rather than a fact about the disease. The paediatric ALL cohort returned zero ARID1A mutations, which is panel coverage and not biology

  6 of 887 paediatric AML cases carry an ARID1A mutation; the paediatric ALL cohort reported zero cohort-wide, a panel-coverage artefact rather than biology

• ← related toThe two TARGET B-ALL studies in cBioPortal are 868 of the same patients assayed twice, so joining them DOES answer the question I had just written off as unanswerable: all four ARID1A-mutant B-ALL patients are definitively KMT2A-negative, three of them with truncating calls. The answer is zero of three - and at a 2.7% KMT2A-positive rate an empty overlap has probability 0.92, so it is an upper bound and not a finding

  four ARID1A-mutant patients in 911 B-ALL patients, three with truncating calls, all four KMT2A-negative

• ← related toARID1B (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

  mutually exclusive cBAF paralogues; ARID1B deficiency is synthetically lethal with ARID1A mutation and >80% of their regulated loci are distinct

• ← co-occurs withVIM::KMT2A - the one KMT2A rearrangement that keeps the SET domain and still throws away the menin-binding exon

  concurrent ARID1A mutation in both cases of the pan-sarcoma screen (n=2)

• related to →Chromocentres and pericentromeric satellite heterochromatin — no KMT2A evidence, but the AT-hook argument says there should be

  SINGLE-SOURCE, 2026, and SWI/SNF-INDEPENDENT: ARID1A interacts with TRIM28 to preserve heterochromatin rigidity outside the SWI/SNF complex, and its loss displaces SETDB1, reversing H3K9me3 repression at endogenous retroelements and triggering viral mimicry (PMID 41661654, colorectal cancer). One group, one disease - recorded as single-source, not established. TRIM28 is not TRIM27 (Q95 2026-09-22)

• related to →SMARCC1 / 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

  ARID1A IS NOT IN THIS PROJECT'S BAF ENRICHMENT AND SMARCC1 IS. The MLL-fusion interactome complex analysis recovers PBRM1, SMARCA4, SMARCB1, SMARCC1, SMARCC2 and SMARCE1 - 6 of 17, q = 0.076 against a nuclear background and 0.186 against the recovered proteome, i.e. not significant at a realistic background. ARID1A reaches this graph through an MLL-AF9 screen and a sarcoma co-mutation instead (Q95 2026-09-22)

• co-occurs with →VIM::KMT2A - the one KMT2A rearrangement that keeps the SET domain and still throws away the menin-binding exon

  The only clinical setting where ARID1A and a KMT2A rearrangement co-occur: both VIM::KMT2A sarcomas in a 14,680-tumour cohort carried concurrent CTNNB1, SMARCB1 and ARID1A mutations at TMB 0.8 and 1.6 per Mb (PMID 32461620, n = 2). It argues the fusion is INSUFFICIENT in that arm, which weakens rather than supports importing ARID1A biology into infant KMT2A::AFF1 (Q95 2026-09-22)

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

  cBAF-defining ARID slot

• co-occurs with →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

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

• related to →KMT2A-rearranged acute myeloid leukaemia in children and infants

  required in the MLL-AF9 shRNA screen but flat in every selectivity measure

• ← related toARID2 (BAF200) - the PBAF subunit whose loss ACCELERATES MLL-AF9 leukaemia, and which is nonetheless required to maintain it

  different complexes, opposite reported directions in the same MLL-AF9 model