Oncofusion

start / graph / gene-dpf2

Proteinmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q47-swi-snf

DPF2 (BAF45D) - a cBAF-specific reader required in MLL-AF9 cells, with one PubMed record in this disease and no selectivity signal

also called BAF45D · REQ · UBID4 · requiem · REQU_HUMAN

DPF2 is a small reader subunit that sits only in the canonical flavour of the SWI/SNF machine. Removing it stops MLL-rearranged leukaemia cells growing, but the same is true of leukaemia cells in general, and only one paper in PubMed has ever looked at it in this disease.

The same thing, for a scientist

DPF2 (Q92785, 391 aa) is a double-PHD-finger subunit of the BAF45 family, present in **cBAF only** - ncBAF lacks BAF45 entirely [E1][E2]. It is the third subunit of the anchor paper: shRNA depletion in mouse MLL-AF9 cells and human THP-1 costs expansion, self-renewal and the undifferentiated state, rescued by human cDNA, with extended survival in vivo [E3]. BAF45d was independently a hit in the 93-hairpin SWI/SNF screen in RN2 cells [E4]. In DepMap it ranks 1,321/17,728 with a difference of only -0.0927 and fails the lineage split in both halves (myeloid z -0.36, B-lineage -0.43) [E5][E6]. **`DPF2 AND (MLL OR KMT2A)` returns exactly one record, the anchor; `BAF45D AND leukemia` also returns exactly one, the same paper [E7].** This is the thinnest primary literature of any gene in this beat and the entry is written so that the thinness is on the record rather than inferred from silence.

What it looks like
loading AF-Q92785-F1…
AlphaFold AF-Q92785-F1 ↗
colourshape

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.

very high (90+)confident (70–90)low (50–70)very low (<50)
I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
Why the cBAF membership matters

DPF2 is one of the subunits that defines canonical BAF. ncBAF contains no BAF45 paralogue at all 2. So a DPF2 requirement is evidence for a cBAF requirement, which is the one module this beat's dependency scan finds no KMT2A selectivity for 5.

That gives a clean internal test. If cBAF is the relevant complex, DPF2 should score; it does not (rank 1,321, difference -0.0927, no lineage-split survival). If ncBAF is, DPF2 should not score and BRD9/BICRA should; they do. The anchor paper's DPF2 phenotype and the modern dependency profile therefore disagree, and the disagreement is informative rather than a contradiction to be papered over: shRNA in one mouse line and one human line, against a CRISPR panel of 54 lines.

The identifier trap in this gene

UniProt's entry name is REQU_HUMAN and the primary protein name is "Zinc finger protein ubi-d4". Neither string contains "DPF2". Reading the protein name back - the standard defence under rule 65 - returns something that does not look like the gene you asked for, and a reader could reasonably conclude the lookup failed. The identifiers are: primary gene DPF2, synonyms BAF45D REQ UBID4, accession Q92785, length 391 1.

The alias REQ is also a three-letter word-like token and is unusable as a literature search term; BAF45D is the alias that works and it returns exactly one leukaemia record 7.

Facts and figures
uniprotQ92785
gene symbolDPF2
protein nameZinc finger protein ubi-d4 (BRG1-associated factor 45D, BAF45D)
length AA391
alphafold IDAF-Q92785-F1
accession resolutionUniProt REST gene_exact:DPF2 AND organism_id:9606 AND reviewed:true, run 2026-09-20; one reviewed row; gene_primary DPF2, length 391 from the same record.
identifier trapentry name REQU_HUMAN and primary protein name 'Zinc finger protein ubi-d4' contain no DPF2 string; reading the protein name back looks like a failed lookup.
complexcBAF only (BAF45 slot); absent from ncBAF
classificationpan-leukaemia dependency required in MLL-AF9 models; no KMT2A selectivity; NOT a demonstrated interactor of a KMT2A fusion
paralogues
[
  "DPF1",
  "DPF3",
  "PHF10 (BAF45A, PBAF)"
]
depmap
{
  "rank": 1321,
  "of": 17728,
  "difference": -0.0927,
  "mean_kmt2a_r": -0.6189,
  "mean_other_leukaemia": -0.5262,
  "pan_essentiality_all_lines": -0.4049,
  "lineage_split_z_myeloid": -0.36,
  "lineage_split_z_b_lineage": -0.43,
  "verdict": "weak or ambiguous; not lineage-independent"
}
literature size
{
  "query": "DPF2 AND (MLL OR KMT2A)",
  "records": 1,
  "alias_query": "BAF45D AND leukemia",
  "alias_records": 1,
  "date": "2026-09-20",
  "tool": "PubMed E-utilities esearch"
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.6189,
  "chronos_mean_other_leukaemia": -0.5262,
  "difference": -0.0927,
  "pan_essentiality_all_lines": -0.4049,
  "selectivity_rank": 1321,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • There is no DPF2-directed agent, so this is a biology entry, not a target entry; any intervention runs through the shared ATPase.
  • The BAF45 slot has three paralogues, so a DPF2-selective agent would face immediate paralogue substitution.
  • A pan-leukaemia dependency with a pan-essentiality of -0.40 offers no therapeutic index over other leukaemias, and none over normal progenitors has been measured.
Still open
  1. Does DPF2's double PHD finger read a histone mark that is altered in KMT2A-rearranged chromatin? Its reader function has not been examined in this disease at all.
  2. Do DPF1 or DPF3 substitute for DPF2 in leukaemic cBAF? All three score near zero in DepMap, which could mean redundancy or could mean only DPF2 is expressed.
  3. Is the anchor paper's DPF2 phenotype a cBAF phenotype or a complex-destabilisation phenotype? A DPF2 point mutant that assembles but cannot read would separate them and has never been made in a leukaemia line.
Every source behind this entry (7)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    DPF2 is UniProt Q92785, 391 aa, primary gene DPF2, synonyms BAF45D REQ UBID4; its entry name is REQU_HUMAN and its primary protein name is 'Zinc finger protein ubi-d4'.

    UniProtKB Q92785 (REQU_HUMAN) · UniProt Consortium · UniProtKB 2026
    What this evidence cannot show

    gene_exact:DPF2 AND organism_id:9606 AND reviewed:true, run 2026-09-20; one reviewed row; gene_primary read back as DPF2 and length 391 from the same fetched record.

    Evidence that argues against this (1)
    • DPF1 and DPF3 are the other BAF45 paralogues and occupy the same slot, so an accession check confirms the protein but not that the phenotype belongs to DPF2 rather than to the slot. Mashtalir N et al. 2018
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    ncBAF (GBAF) contains no BAF45 subunit, so DPF2 membership is specific to canonical BAF.

    PMID 29374058DOI
    What this evidence cannot show

    Composition established by density gradient and IP in prostate cancer lines; not confirmed in haematopoietic cells.

    Evidence that argues against this (1)
    • Absence in a purified steady-state assembly does not exclude sub-stoichiometric or transient association, which affinity purification would not retain. Mashtalir N et al. 2018
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    shRNA depletion of Dpf2 reduces expansion, clonal expansion and self-renewal of mouse MLL-AF9 cells and of human THP-1, with apoptosis, c-Kit loss, Mac1 gain and extended survival in vivo; human DPF2 cDNA rescues.

    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 only, myeloid models only; DPF2 transcriptional changes overlap those of SMARCA4 and SMARCD2 (320 up / 209 down shared, ~4-6% of changing genes), which is a modest overlap being described as cooperative function.

    Evidence that argues against this (1)
    • The same paper shows Jurkat and T-ALL1 require SMARCA4, and reports the three subunits' gene-expression programmes as overlapping rather than fusion-specific - nothing in it isolates a KMT2A-dependent function. Xiu Y et al. 2018
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    BAF45d was identified as a requirement for MLL-AF9/NrasG12D RN2 proliferation in a 93-hairpin SWI/SNF shRNA screen.

    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

    One mouse line; hit criterion is >3-fold GFP+ depletion over 10 days for multiple validated hairpins.

    Evidence that argues against this (1)
    • The screen's hits span cBAF-specific (BAF250a, BAF45d) and shared (BAF155, BAF60a/b, BAF53a) subunits, so it identifies 'an intact complex is needed', not which complex. Michel BC et al. 2018
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    DPF2 ranks 1,321 of 17,728 with difference -0.0927 (KMT2A-r -0.6189, other leukaemia -0.5262, pan-essentiality -0.4049): a real dependency across leukaemia with almost no KMT2A selectivity.

    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; the raw difference of -0.0927 against a pan-essentiality of -0.40 is close to the floor-compression regime this project has flagged, where a small index-comparator difference is uninterpretable rather than small.

    Evidence that argues against this (1)
    • DPF1 and DPF3 are also present in some cells and could buffer a DPF2 knockout; the panel-wide DPF1 and DPF3 scores are near zero, which is consistent either with redundancy or with neither being expressed. Oncofusion knowledge graph et al. 2026
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    DPF2 fails the lineage split in both halves of the KMT2A-rearranged arm (myeloid z -0.36, B-lineage z -0.43), while being strongly depleted in the two myeloid non-KMT2A reference lines (OCI-AML3 -2.51, U-937 -2.20).

    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

    z is measured in SDs of the other-leukaemia arm; a reference line being more depleted than the index arm is the clearest possible statement that the dependency is not about the rearrangement. Eight myeloid and five B-lineage index lines only.

    Evidence that argues against this (1)
    • U-937 carries PICALM::MLLT10, which shares the elongation biology of KMT2A fusions, so its strong depletion is compatible with a fusion-class dependency that the KMT2A lines happen to under-read. Oncofusion knowledge graph et al. 2026
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    DPF2 has exactly one PubMed record in this disease under either its current symbol or its BAF alias: `DPF2 AND (MLL OR KMT2A)` = 1 and `BAF45D AND leukemia` = 1, both PMID 26571505.

    What this evidence cannot show

    Bounded absence. Bounded to PubMed on 2026-09-20 over DPF2 and BAF45D. The alias `REQ` was deliberately not used: it is a three-letter token that inflates rather than recalls, the failure mode this project measured at 57-fold for NELFE's alias RD. So the true count could be higher by papers that use only 'requiem'.

    Evidence that argues against this (1)
    • The 93-hairpin screen in PMID 24285714 contains a BAF45d result and is invisible to a DPF2 query because it uses only the BAF nomenclature in its figures - a live instance of the rule-64 failure mode inside this very beat. Shi J et al. 2013
Connected to