Oncofusion

start / graph / gene-dek

Proteinmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q101-writers-and-readers

DEK - a chromatin architectural protein that gatekeeps H3.3 and boosts HP1alpha, is a fusion partner in its own right, and is the substrate behind this graph's third CK2 route

also called DEK · protein DEK · DEK oncoprotein

DEK is an abundant nuclear protein that holds chromatin in shape: it helps heterochromatin stay compact and it controls which chaperones are allowed to deposit a particular histone variant. It is also half of a fusion protein that causes its own kind of leukaemia, and switching off a kinase that modifies it kills leukaemia stem cells while sparing normal ones.

The same thing, for a scientist

DEK (P35659, 375 aa, HGNC:2768) is named in twenty-eight entries of this graph and had none of its own, which matters because it does three different jobs here. **As chromatin architecture:** DEK is a Suppressor of Variegation - it binds HP1alpha directly and **markedly enhances HP1alpha's binding to H3K9me3**, and Dek loss in *Drosophila* gives a Su(var) phenotype with global heterochromatin reduction [E2]. **As a histone-variant gatekeeper:** DEK restricts access to non-nucleosomal H3.3 by dedicated chaperones, balancing HIRA-dependent and DAXX/ATRX-dependent H3.3 deposition; DEK depletion re-routes H3.3 destined for PML nuclear bodies onto chromatin, displaces PML bodies and ATRX from telomeres and produces a fragile-telomere phenotype [E3]. **As a leukaemia protein:** it is the 5-prime partner of DEK::NUP214 in t(6;9) AML, which this graph holds as `mechanism-dek-nup214-xpo1-dependent-transcriptional-activator`, and CK2-phosphorylated DEK sustains leukaemia stem cells through GABPA and PBX3 in MLL-AF9 and HOXA9/MEIS1 models and patient-derived xenografts [E4]. That last paper is **already cited four times here from the CK2 side** - so this entry deliberately does not restate the CK2 route and instead supplies the DEK-side biology those entries assume and do not contain. UniProt's entire FUNCTION comment for DEK is five words: *Involved in chromatin organization* [E1].

What it looks like
DEK375 aa
SAPSAP · residues 149–183DEK-CDEK-C · residues 319–375NLS NLS (nuclear localization signal) · residues 205–2210100200300
loading AF-P35659-F1…
AlphaFold AF-P35659-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
DEK as a Su(var): direct HP1alpha binding and enhanced H3K9me3 recognition

2 (Genes Dev 2011) identified DEK - described there as an abundant nuclear protein with a previously enigmatic in vivo function - as a Suppressor of Variegation crucial to global heterochromatin integrity:

  • DEK interacts directly with HP1alpha and markedly enhances its binding to H3K9me3.
  • Loss of Dek in Drosophila produces a Su(var) phenotype and a global reduction in heterochromatin.

Why that lands in this project. mark-h3k9me3 is an entry here and its argument is that H3K9me3 restrains the MLL-fusion programme - SETDB1 overexpression in mouse MLL-AF9 cells silences a gene set whose promoters are H3K9-hypomethylated in AML patient samples. That entry lists the HP1 readers CBX5/HP1alpha, CBX1 and CBX3 explicitly as readers unwritten in this graph. DEK is the amplifier of the first of those readers, and it is a leukaemia protein with a clinical-stage kinase inhibitor pointed at it. Nobody has put DEK and H3K9me3 together in a KMT2A-rearranged cell: DEK[tiab] AND heterochromatin[tiab] = 14, DEK[tiab] AND HP1[tiab] = 1 and that single record is a KSHV LANA interactome paper, not a DEK study 5.

And the honest limit on the genetics. The Su(var) phenotype and the global heterochromatin reduction are Drosophila; the HP1alpha binding and H3K9me3 enhancement are biochemical. There is no mammalian haematopoietic heterochromatin experiment in that paper.

DEK as an H3.3 gatekeeper, which plugs straight into a hypothesis this graph already holds

3 (Genome Res 2014) makes DEK a gatekeeper of chromatin that controls which chaperone gets to deposit histone variant H3.3:

  • DEK regulates the differential HIRA-dependent and DAXX/ATRX-dependent distribution of H3.3 across chromosomes, in somatic and embryonic stem cells.
  • By live-cell imaging, non-nucleosomal H3.3 normally destined for PML nuclear bodies is re-routed to chromatin after DEK depletion, giving HIRA-dependent widespread deposition plus incorporation into heterochromatin foci in a DAXX/ATRX-requiring process.
  • In ES cells, DEK loss displaces PML bodies and ATRX from telomeres, redistributes H3.3 from telomeres to chromosome arms and pericentric heterochromatin, and induces a fragile-telomere phenotype.

This graph already holds hyp-kmt2a-aff1-alters-h3-3-deposition-at-fusion-targets and exp-h3-3-turnover-at-fusion-targets-under-dot1l-inhibition, and it has entries for gene-hira, gene-daxx, gene-atrx and gene-pml. DEK is the upstream switch on exactly that machinery and was not in the hypothesis. It also connects the H3.3 question to the subnuclear-compartment question the PI keeps asking: DEK depletion moves H3.3 between compartments - out of PML bodies, onto chromatin, into pericentric foci - which is a localisation phenotype, not only a deposition one, and this graph holds localization-chromocentres-pericentromeric-kmt2a-r and localization-polycomb-bodies-kmt2a-r as compartment entries.

The leukaemia half, stated without restating what is already written

Two things are already in this graph and are not repeated here.

  1. mechanism-dek-nup214-xpo1-dependent-transcriptional-activator carries the t(6;9)/DEK::NUP214 biology - CUT&RUN occupancy at FOXC1 and the HOXA/HOXB clusters, selective ex vivo response to selinexor and eltanexor among 57 primary AML samples, and the XPO1-dependence of the fusion's chromatin binding. That entry's own framing is the one this gene entry serves: DEK is itself a chromatin-binding nuclear factor, so this fusion had an alternative route to a promoter and still required XPO1. This entry supplies the DEK biology that sentence presumes.
  2. The CK2 route is written from the kinase side in gene-csnk2a1, exp-ck2-inhibitor-two-route-separation, mechanism-ck2-t2724-s2726-licenses-taspase1-cleavage and mechanism-set-pp2a-phosphatase-arm-kmt2a-r. PMID 42378254 is already cited four times here (rule 82 checked, 2026-09-22), and gene-csnk2a1's reading of it is correct and is preserved: DEK is not a KMT2A contact, so this is a third CK2 route into the same disease rather than a third route into the same protein.

What this entry adds from 4 that the CK2-side entries do not carry: the DEK-side mechanism and the dose-direction. DEK recruits the transcription factor GABPA to upregulate the cofactor PBX3, sustaining the leukaemogenic program, and that DEK-GABPA interaction strictly depends on the 4S phosphorylation which stabilises the complex conformation. DEK deletion enhances LSC chemosensitivity to azacitidine plus venetoclax, and DEK OVEREXPRESSION confers robust chemoresistance - a bidirectional dose-response that makes DEK level, not only DEK phosphorylation, a resistance variable. PBX3 is named in 47 entries here and exp-pbx-family-protein-panel-and-combinatorial-knockout-in-kmt2a-aff1 is a planned experiment; DEK is now an upstream handle on PBX3 that experiment does not include.

And the retrieval point. DEK[tiab] AND KMT2A[tiab] = 7 and every one of the seven is a cytogenetic panel, genomic-landscape survey or transcript-screening paper that lists t(6;9) alongside KMT2A rearrangements - none is a DEK mechanism paper. DEK[tiab] AND MLL[tiab] = 18, AND leukemia[tiab] = 139. So registry row 124 applies in a new shape: here both index-gene operands are noise, and the mechanism is reached through neither 5.

Localization, cell type, and the thinnest UniProt FUNCTION comment in this beat

Subcellular: Nucleus, two experimental citations (ECO:0000269 PubMed:17524367, 19695025), with an annotated nuclear localization signal at 205-221 1. Subnuclear: NO COMPARTMENT ASSIGNED in the database - which is a striking omission for a protein whose published phenotype is the redistribution of a histone variant between PML nuclear bodies, telomeres, chromosome arms and pericentric heterochromatin. Recorded as a database absence with the literature reading it contradicts: the sources place DEK's activity at PML bodies, telomeres and heterochromatin 3, while the record gives only Nucleus.

Domain architecture: a SAP domain at 149-183, a DEK-C domain at 319-375, an NLS at 205-221, and disorder over 1-61 and 184-325 - so roughly two-thirds of the protein is disordered around two small structured modules. Note for any breakpoint reasoning: gene-med19 reports that MED19/LCMR1 binds DEK's N-terminal region, which is the portion retained in DEK::NUP214, and its own counter-evidence correctly warns that retention of a region is not retention of a contact.

Cell type and state. UniProt: ubiquitous, expressed at relatively high levels 1. DepMap: not a dependency and in fact slightly positive - mean gene effect +0.1433 over 1,258 lines, 0.0% of lines depleted below -0.5, lymphoid tissue +0.1311 over 97 lines 6; in this project's own KMT2A scan both arms are positive (+0.0971 KMT2A-rearranged, +0.1320 other leukaemia) at rank 4,830 of 17,728, which ranks noise. That directly contradicts the LSC result, and the resolution is in the readout: 4 found DEK dependence in leukaemia-stem-cell-enriched populations in mouse models and PDXs, and a pooled dropout screen across 1,258 bulk cell lines measures neither stem-cell maintenance nor engraftment. The two measurements are not in conflict about DEK; they are about different cells.

Other complexes worth recording, because DEK is promiscuous and its non-chromatin memberships are how it appears in interactome data: an mRNA-splicing-dependent exon junction complex with RBM8A, RNPS1, SRRM1 and ALYREF; the B-WICH complex with SMARCA5/SNF2H, BAZ1B/WSTF, SF3B1, MYO1C, ERCC6, MYBBP1A and DDX21; and direct interactions with histones H2A, H2B, H3, H4, acetylated H4, non-phosphorylated DAXX and HDAC2 1.

Facts and figures
uniprotP35659
uniprot IDDEK_HUMAN
gene symbolDEK
length AA375
hgnc IDHGNC:2768
ensembl gene IDENSG00000124795
alphafold IDAF-P35659-F1
accession assertionP35659 asserted explicitly. gene_exact:DEK AND organism_id:9606 AND reviewed:true returns exactly ONE reviewed human record (2026-09-22), 375 aa, with NO gene synonyms at all - the only protein in this beat with an empty synonym list. Note that DEK is 375 aa and PRMT6 is also 375 aa, so a bare length readback distinguishes neither; they are unrelated proteins and this is coincidence, not a paralogue hazard, but it is a reason not to confirm either accession by length.
PDB count rcsb6
PDB noteSix entries by RCSB accession-index query on P35659 (2026-09-22), the full list; UniProt cross-references the same six. 1Q1V and 2JX3 are the historic small-module structures; 8KCY, 8KD1, 9L1X and 9L22 are recent. No structure with NUP214, with a nucleosome, or of the DEK::NUP214 fusion junction.
domain architecture noteTwo small structured modules - a SAP domain at 149-183 and a DEK-C domain at 319-375 - separated and flanked by extensive disorder (1-61 and 184-325), with an NLS at 205-221. Roughly two-thirds disordered. THE PHOSPHO-SITES THAT MATTER FOR LEUKAEMIA SIT IN THE DISORDER: Ser301/303/306/307, the CK2 4S cluster whose phosphorylation is strictly required for the DEK-GABPA interaction, lie inside the 184-325 disordered region and immediately N-terminal to DEK-C. For breakpoint reasoning: gene-med19 maps a MED19/LCMR1 contact to DEK's N-TERMINAL region, which is the portion retained in DEK::NUP214.
enzyme classNOT AN ENZYME. DEK is a chromatin architectural protein and histone-variant gatekeeper; it binds DNA and histones and modulates other proteins' activities (HP1alpha reading, HIRA and DAXX/ATRX chaperone routing) without catalysing anything itself.
localizationNUCLEUS. UniProt P35659 fetched 2026-09-22 carries one location value, Nucleus, with TWO experimental citations (ECO:0000269 PubMed:17524367, 19695025), plus an annotated nuclear localization signal at 205-221. No cytoplasmic value and no localisation note. Note that DEK also has a documented extracellular/secreted existence in the autoimmunity literature (DEK[tiab] AND 'juvenile idiopathic arthritis'[tiab] = 4, 2026-09-22) which the UniProt record does not reflect; that literature was NOT read for this entry and is named here only so the nuclear-only record is not mistaken for a complete account.
localization evidence codeECO:0000269 with two citations - better than the single-citation records elsewhere in this beat. Neither paper was read for this entry, so rule 91 applies: the codes say curators saw nuclear experiments, not that either is the founding or best DEK localisation study.
subnuclear localizationNO COMPARTMENT ASSIGNED IN THE DATABASE, AND THE LITERATURE CONTRADICTS THAT ABSENCE - which is why this field is a finding rather than a gap. UniProt gives only `Nucleus`. The primary literature places DEK's activity squarely in named subnuclear compartments: it is titled a PML-ASSOCIATED protein, its depletion re-routes non-nucleosomal H3.3 that is normally destined for PML NUCLEAR BODIES onto chromatin, and DEK loss in embryonic stem cells DISPLACES PML BODIES AND ATRX FROM TELOMERES and redistributes H3.3 into PERICENTRIC HETEROCHROMATIN (PMID 25049225); separately DEK is required for global heterochromatin integrity and enhances HP1alpha binding to H3K9me3 (PMID 21460035). So the defensible subnuclear statement is: PML nuclear bodies, telomeres and heterochromatin, from the literature, with NO corresponding database annotation. REOPENING CONDITION: DEK immunofluorescence in a t(4;11) line scored against PML bodies and chromocentres, both of which this graph already holds as compartment entries.
localization shown for fusionNO for KMT2A fusions - nobody has looked at DEK in any KMT2A-rearranged cell. For DEK's OWN fusion the situation is better but still not a localisation result: DEK::NUP214 occupancy was mapped by CUT&RUN at the FOXC1 promoter and the HOXA/HOXB clusters and shown to fall under XPO1 inhibition (this graph, mechanism-dek-nup214-xpo1-dependent-transcriptional-activator), which is a genomic address rather than a subnuclear one.
epigenetic readoutARCHITECTURAL MODULATOR - neither writer, eraser nor reader, and emphatically not not-applicable. DEK has no catalytic domain and no annotated reader module (SAP at 149-183 and DEK-C at 319-375, plus disorder), yet it acts on two specific marks and one histone variant. (1) IT AMPLIFIES A READER OF A MARK: DEK binds HP1alpha directly and MARKEDLY ENHANCES HP1alpha's binding to H3K9me3, and Dek loss gives a Su(var) phenotype with globally reduced heterochromatin (PMID 21460035) - so the mark is H3K9me3, the reader is HP1alpha/CBX5, and DEK is the potentiator. mark-h3k9me3 in this graph lists CBX5/HP1alpha among 'readers unwritten in this graph'. (2) IT GATEKEEPS A HISTONE VARIANT: DEK restricts access to non-nucleosomal H3.3 by dedicated chaperones, balancing HIRA-dependent against DAXX/ATRX-dependent deposition, and its depletion re-routes H3.3 from PML bodies onto chromatin and into pericentric heterochromatin with loss of telomeric ATRX (PMID 25049225). (3) IT BINDS HISTONES DIRECTLY, including acetylated H4, and sits in the B-WICH remodelling complex with SMARCA5 (UniProt). BOUNDED, queries and translations read back 2026-09-22: DEK[tiab] AND histone[tiab] = 38, AND heterochromatin[tiab] = 14, AND nucleosome[tiab] = 11, AND chaperone[tiab] = 6, AND H3.3[tiab] = 5, AND HP1[tiab] = 1 (off topic - a KSHV LANA interactome paper), AND KMT2A[tiab] = 7 (all seven cytogenetic panels), AND MLL[tiab] = 18. NO H3K9me3, HP1alpha or H3.3 measurement involving DEK exists in any KMT2A-rearranged cell.
epigenetic mark namedH3K9me3 - not written by DEK but READ BY HP1alpha, whose binding DEK enhances. Histone variant H3.3 - not modified by DEK but routed by it, between HIRA-dependent chromatin-wide deposition and DAXX/ATRX-dependent heterochromatic deposition. Acetylated histone H4 is a direct binding partner per UniProt with no functional consequence recorded.
cell typeUbiquitous and abundant per UniProt. The leukaemia-relevant cell type is LEUKAEMIA STEM CELL: DEK phosphorylation was identified by phosphoproteomic comparison of normal HSPCs against LSC-ENRICHED populations, DEK deficiency impaired LSC maintenance in MLL-AF9 and HOXA9/MEIS1 mouse models and in patient-derived xenografts, and CX-4945 depleted LSCs while SPARING NORMAL HSPCs (PMID 42378254). Other measured systems: mouse embryonic stem cells and somatic cell lines (H3.3 routing), Drosophila (heterochromatin genetics), and 57 primary AML samples in the t(6;9) drug screen this graph already holds. DepMap: 1,258 lines mean +0.1433; lymphoid tissue n = 97 mean +0.1311. NO measurement in an infant t(4;11) sample and none in B-lineage.
cell stateSTEMNESS and CHEMORESISTANCE, and the dose-direction is bidirectional, which is unusual and useful: DEK DELETION enhances leukaemia-stem-cell chemosensitivity to azacitidine plus venetoclax, while DEK OVEREXPRESSION confers robust chemoresistance (PMID 42378254) - so DEK level, not only DEK phosphorylation, is a resistance variable, and this graph holds cell-state-relapse-initiating-persister and cell-state-apoptotic-priming-bcl2 as the states that would be affected. A second state is the CELL-CYCLE-independent chromatin-maintenance role in ES cells. No differentiation, quiescence or niche state has been examined for DEK in a KMT2A-rearranged cell.
shown for fusionPARTIAL, named rather than graded. IN AN MLL FUSION: DEK deficiency impaired LSC maintenance and AML progression in MLL-AF9-driven mouse leukaemia, and CX-4945 selectively depleted LSCs - main-text results, not a supplementary panel (PMID 42378254). IN ITS OWN FUSION: DEK::NUP214 occupancy at FOXC1 and HOXA/HOXB by CUT&RUN with XPO1-dependence, in t(6;9) AML (already written here as mechanism-dek-nup214-xpo1-dependent-transcriptional-activator). NOTHING in KMT2A::AFF1, nothing in B-lineage, nothing in an infant sample, and no physical contact with KMT2A or any KMT2A fusion reported anywhere - gene-csnk2a1's statement that DEK is not a KMT2A contact stands.
absence classliterature-absence-bounded
absence class noteMIGRATED 2026-09-23, mechanically and not by judgement: the class string itself begins 'literature'. The field previously held the whole verdict as free text and that string is preserved here verbatim because it says more than the token can: 'literature - bounded with counts, resolved id lists, querytranslations and controls; plus one DATABASE absence that the literature contradicts (see subnuclear_localization).'
domains
SAP149.0SAP: 149NLS (nuclear localization…205.0NLS (nuclear localization signal): 205DEK-C319.0DEK-C: 319
PDB IDS
[
  "1Q1V",
  "2JX3",
  "8KCY",
  "8KD1",
  "9L1X",
  "9L22"
]
domains — the numbers behind the chart
[
  {
    "name": "SAP",
    "start": 149,
    "end": 183,
    "pfam": null
  },
  {
    "name": "DEK-C",
    "start": 319,
    "end": 375,
    "pfam": null
  },
  {
    "name": "NLS (nuclear localization signal)",
    "start": 205,
    "end": 221,
    "pfam": null
  }
]
complex membership
{
  "exon_junction_complex": "DEK + RBM8A + RNPS1 + SRRM1 + ALYREF/THOC4, mRNA-splicing-dependent",
  "b_wich": "SMARCA5/SNF2H + BAZ1B/WSTF + SF3B1 + DEK + MYO1C + ERCC6 + MYBBP1A + DDX21",
  "direct_partners": "histones H2A, H2B, H3, H4 and acetylated H4; non-phosphorylated DAXX; HDAC2; HP1alpha (direct, PMID 21460035); GABPA (phospho-4S-dependent, PMID 42378254); MED19/LCMR1 via DEK's N-terminal region (PMID in gene-med19 E9)",
  "note": "No DEK-containing complex has been reported to include KMT2A, a KMT2A fusion or menin. gene-csnk2a1's reading is preserved: DEK is not a KMT2A contact."
}
fusion partner role
{
  "fusion": "DEK::NUP214, t(6;9)(p22.3;q34.1) - a WHO-recognised AML entity of younger adults",
  "dek_portion_retained": "the N-terminal region, which is where gene-med19 maps the MED19/LCMR1 contact; retention of a region is not retention of a contact, as that entry's own counter-evidence warns",
  "already_written_here": "mechanism-dek-nup214-xpo1-dependent-transcriptional-activator - CUT&RUN at FOXC1 and HOXA/HOXB, selinexor and eltanexor response among 57 primary AML samples, XPO1-dependent chromatin binding. NOT restated in this entry.",
  "this_entry_adds": "the DEK-side biology that entry presumes when it says 'DEK is itself a chromatin-binding nuclear factor' - HP1alpha amplification on H3K9me3, H3.3 chaperone gatekeeping, the PML-body and telomere phenotypes, and the domain map showing that the CK2 4S cluster lies in disorder just N-terminal to DEK-C."
}
depmap
{
  "source": "DepMap CRISPR (Chronos) via Open Targets Platform GraphQL, fetched 2026-09-22",
  "release": "NOT STATED BY THE API - unverified (rule 83)",
  "is_essential": false,
  "n_screens": 1258,
  "mean_gene_effect": 0.1433,
  "fraction_lines_depleted_below_minus_0_5": 0,
  "lymphoid_tissue": {
    "n_lines": 97,
    "mean_gene_effect": 0.1311
  }
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": 0.0971,
  "chronos_mean_other_leukaemia": 0.132,
  "difference": -0.0349,
  "pan_essentiality_all_lines": 0.1415,
  "selectivity_rank": 4830,
  "rank_denominator": 17728,
  "percentile": 27.245,
  "direction": "NOT A DEPENDENCY IN EITHER ARM - both means are POSITIVE. The rank ranks a 0.035-unit difference between two positive numbers and carries no information. AND IT DOES NOT MEASURE THE PUBLISHED PHENOTYPE: the DEK result is leukaemia-stem-cell maintenance, engraftment and chemoresistance in mouse models and PDXs, which a pooled bulk-line dropout screen cannot see in either direction.",
  "provenance": "Joined 2026-09-22 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), 11 KMT2A-rearranged and 43 other leukaemia lines; the scan records NO DepMap release (rule 83)."
}
reached for by
[
  "mechanism-dek-nup214-xpo1-dependent-transcriptional-activator - the t(6;9) fusion's XPO1-dependent chromatin binding, 19 mentions",
  "gene-med19 - MED19/LCMR1 binds DEK's N-terminal region, the portion retained in DEK::NUP214, 18 mentions",
  "gene-nup214 - the 3-prime partner, 12 mentions",
  "gene-csnk2a1, gene-csnk2a2, gene-csnk2b and exp-ck2-inhibitor-two-route-separation - the CK2-DEK-4S route, already written from the kinase side",
  "hyp-xpo1-is-a-shared-address-for-hoxa-driven-leukaemia, gene-foxc1, mechanism-nes-crm1-chromatin-addressing, drug-eltanexor - the export axis"
]
How this could bite us back
  • The intervention is not DEK-directed, it is CK2-directed, and this graph has already written why that is a problem: a CX-4945 phenotype in KMT2A-rearranged cells has at least THREE candidate mediators - taspase1 licensing, the SET/PP2A arm and DEK-4S - so attributing a response to DEK without a DEK phospho-readout is uncontrolled, and CK2 is a pleiotropic kinase with hundreds of substrates.
  • The heterochromatin direction may invert the therapeutic argument: if DEK amplifies HP1alpha reading of H3K9me3, and H3K9 methylation restrains the MLL-fusion programme as mark-h3k9me3 argues, then removing DEK weakens a restraint at the same time as it removes an LSC dependency. Net direction unknown, and nobody has measured both in one system.
  • Lineage and regimen transfer: every DEK leukaemia-stem-cell result is myeloid (MLL-AF9, HOXA9/MEIS1, AML PDXs) and the chemosensitisation was to azacitidine plus venetoclax, an adult AML backbone. Neither the lineage nor the regimen is this project's, and infant t(4;11) therapy is a completely different set of drugs.
  • Essential-function and normal-tissue risk: DEK is ubiquitous, abundant, and its loss in embryonic stem cells produces a FRAGILE TELOMERE phenotype and telomere dysfunction. In an infant - a patient with a lifetime of haematopoiesis ahead - inducing telomere fragility in stem cells is a late-effects risk that no leukaemia experiment measures, and the LSC-sparing-of-normal-HSPCs claim rests on the authors' own comparison.
  • The dose-direction is a resistance mechanism ready-made: DEK OVEREXPRESSION confers robust chemoresistance. Any pressure that selects for higher DEK - and DEK is amplified or overexpressed in several cancers - produces resistance to the venetoclax-containing regimens this graph is otherwise interested in.
  • A screen will not find this: DEK's DepMap mean is POSITIVE (+0.1433) in 1,258 lines and 0.0% of lines are depleted, so no pooled viability screen in this project or anywhere else will nominate DEK, and any target-ranking analysis here that leans on DepMap will rank it dead last while the mouse and PDX data say otherwise.
Still open
  1. Does DEK depletion change H3.3 distribution in a KMT2A-rearranged cell, and is the change global or at fusion targets? This graph's H3.3 hypothesis predicts a local change at fusion targets and the DEK mechanism predicts a genome-wide redistribution between compartments; the same CUT&RUN would separate them, and the existing planned experiment is scoped so that it would not.
  2. Is HP1alpha recruitment to H3K9me3 at fusion-restrained loci DEK-dependent? mark-h3k9me3 argues that H3K9 methylation restrains the MLL-fusion programme and lists CBX5/HP1alpha as a reader this graph has never written; DEK is the documented amplifier of that reader and has a clinical-stage kinase inhibitor pointed at it.
  3. Do DEK's heterochromatin function and its leukaemia-stem-cell function point in opposite directions? Weakening heterochromatin should RELEASE the restraint mark-h3k9me3 describes, while depleting DEK is reported to kill LSCs. Either the two functions are separable - plausibly by the 4S phosphorylation - or one of the directions is wrong.
  4. Where is DEK in a t(4;11) blast, and is it in PML bodies? The database says only Nucleus while the literature calls DEK a PML-associated protein and shows DEK loss moving PML bodies off telomeres. One immunofluorescence experiment would fill a subnuclear field that no database currently supports.
  5. Is DEK-4S phosphorylation present in KMT2A::AFF1 B-lineage blasts? Every DEK LSC experiment is myeloid - MLL-AF9, HOXA9/MEIS1, AML PDXs - and this graph's index disease is infant B-lineage. A phospho-western on infant t(4;11) material is a one-day experiment that would tell us whether the CK2-DEK axis exists in the right lineage at all.
  6. Does DEK level predict venetoclax or azacitidine response? DEK deletion sensitises and DEK overexpression confers resistance, and this graph holds cell-state-apoptotic-priming-bcl2, drug-venetoclax and an entire relapse-persister layer with no DEK measurement in any of it.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    DEK (P35659, 375 aa) carries a SAP domain at 149-183, a DEK-C domain at 319-375 and a nuclear localization signal at 205-221 with disorder over 1-61 and 184-325; it binds DNA and histones H2A, H2B, H3, H4 and acetylated H4 as well as non-phosphorylated DAXX and HDAC2; it is a component of an mRNA-splicing-dependent exon junction complex with RBM8A, RNPS1, SRRM1 and ALYREF and of the B-WICH complex with SMARCA5, BAZ1B, SF3B1, MYO1C, ERCC6, MYBBP1A and DDX21; it is located in the Nucleus with two experimental citations; it is ubiquitous and expressed at relatively high levels; and UniProt's entire FUNCTION comment for it is the five words 'Involved in chromatin organization'.

    What this evidence cannot show

    This is the thinnest functional record of any protein in this beat and the thinness is itself worth recording: a five-word FUNCTION comment for a protein with a Genes Dev paper on heterochromatin integrity, a Genome Res paper on histone-variant chaperone routing and a 2026 Blood paper on leukaemia stem cells. The SUBUNIT block carries no inline citations for most of its statements. No subnuclear location value appears at all despite the published PML-body and telomere phenotypes. The accession is asserted explicitly; note in passing that DEK (375 aa) and PRMT6 (375 aa) are the same length, which is coincidence rather than a paralogue hazard but does mean a bare length check distinguishes neither from the other.

    Evidence that argues against this (1)
    • The record's own SUBUNIT block argues against reading DEK as a dedicated chromatin factor: it places DEK in an mRNA-splicing-dependent EXON JUNCTION COMPLEX and in the B-WICH complex alongside SF3B1 and DDX21, both RNA-processing contexts, and DEK binds essentially every core histone plus DNA non-specifically. A protein that is abundant, disordered over two-thirds of its length and binds all four histones is exactly the kind of interactor that appears in affinity-purification data for almost any nuclear bait, so any interactome-derived DEK association should be treated as low-prior unless it comes with a direct functional readout. UniProt Consortium 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    DEK is a Suppressor of Variegation crucial to global heterochromatin integrity: it interacts directly with Heterochromatin Protein 1 alpha and MARKEDLY ENHANCES HP1alpha binding to trimethylated H3K9, and loss of Dek in Drosophila produces a Su(var) phenotype with a global reduction in heterochromatin - making DEK a factor in the balance between heterochromatin and euchromatin in vivo.

    The DEK oncoprotein is a Su(var) that is essential to heterochromatin integrity. · Kappes F, Waldmann T, Mathew V, Yu J, Zhang L, Khodadoust MS, Chinnaiyan AM, Luger K, Erhardt S, Schneider R, Markovitz DM · Genes and Development 2011
    PMID 21460035DOI
    What this evidence cannot show

    The in vivo genetics are DROSOPHILA - the Su(var) phenotype and the global heterochromatin reduction are fly results, and the mammalian arm is the biochemistry (direct HP1alpha binding, enhanced H3K9me3 recognition). There is no haematopoietic system, no leukaemia and no KMT2A fusion in this paper. It is cited once already in this graph, inside gene-med19's counter-evidence on the MED19-DEK contact; this entry carries it for DEK's own chromatin function. Abstract read in full 2026-09-22; full text not read, so the affinity change described as 'markedly enhances' is not quantified here.

    Evidence that argues against this (1)
    • If DEK amplifies HP1alpha reading of H3K9me3, the therapeutic direction collides with this graph's own H3K9me3 position, and it collides the wrong way for a DEK-directed agent. mark-h3k9me3 argues that H3K9 methylation RESTRAINS the MLL-fusion programme, with SETDB1 overexpression silencing a gene set that is H3K9-hypomethylated in AML patient samples. Removing or dephosphorylating DEK would then be expected to WEAKEN heterochromatin and weaken HP1alpha recruitment - releasing the restraint - which is the opposite of what the CK2-DEK therapeutic argument predicts. Either the heterochromatin function is separable from the LSC function, or one of the two directions is wrong; no experiment addresses this. Oncofusion Knowledge Graph 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    DEK regulates the differential HIRA-dependent and DAXX/ATRX-dependent distribution of histone variant H3.3 in somatic and embryonic stem cells: after DEK depletion, non-nucleosomal H3.3 normally destined for PML nuclear bodies is re-routed to chromatin, giving HIRA-dependent widespread deposition and DAXX/ATRX-requiring incorporation into heterochromatin foci; in embryonic stem cells DEK loss displaces PML bodies and ATRX from telomeres, redistributes H3.3 from telomeres to chromosome arms and pericentric heterochromatin, and induces a fragile-telomere phenotype - so DEK acts as a gatekeeper restricting broad access to H3.3 by dedicated chaperones.

    The PML-associated protein DEK regulates the balance of H3.3 loading on chromatin and is important for telomere integrity. · Ivanauskiene K, Delbarre E, McGhie JD, Kuntziger T, Wong LH, Collas P · Genome Research 2014
    PMID 25049225DOI
    What this evidence cannot show

    The systems are somatic cell lines and mouse embryonic stem cells - not haematopoietic, not leukaemic, no fusion. 'Gatekeeper' is the authors' own proposed model rather than a measured mechanism. The telomere phenotype is an ES-cell result and ES cells have unusual H3.3 and telomere biology, so it should not be carried into a leukaemic blast without checking. Abstract read in full 2026-09-22; full text not read, so effect sizes and the identity of the imaged loci are not characterised here.

    Evidence that argues against this (1)
    • This graph's H3.3 hypothesis is about the fusion changing H3.3 deposition at its own target genes, and a DEK mechanism would predict something different and broader - a genome-wide REDISTRIBUTION of H3.3 between compartments rather than a local change at fusion targets. Those two predictions are separable by the same experiment, and the existing planned experiment (exp-h3-3-turnover-at-fusion-targets-under-dot1l-inhibition) is scoped to fusion targets only, so it would register a global redistribution as a normalisation artefact rather than as a result. A DEK arm would have to be added, and no H3.3 measurement of any kind in a KMT2A-rearranged cell currently exists. Oncofusion Knowledge Graph 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    CK2-phosphorylated DEK sustains leukaemia stem cells: phosphoproteomic comparison of normal HSPCs with LSC-enriched populations identified DEK phosphorylation, DEK deficiency impaired LSC maintenance and AML progression in MLL-AF9-driven and HOXA9/MEIS1-driven mouse models and in patient-derived xenografts, DEK recruits the transcription factor GABPA to upregulate the cofactor PBX3 in a manner strictly dependent on DEK phosphorylation at Ser301/303/306/307 by casein kinase 2, DEK DELETION enhances LSC chemosensitivity to azacitidine plus venetoclax while DEK OVEREXPRESSION confers robust chemoresistance, and blocking the 4S sites by mutagenesis or with the clinical-stage CK2 inhibitor CX-4945 selectively depletes LSCs while sparing normal HSPCs and synergises with venetoclax.

    Phosphorylated DEK sustains leukemia stem cells by enabling PBX3-driven transcriptional reprogramming. · Lei Y, Lai Y, Li Y, Wang Y, Fan H, Gong Q, Wu F, Yan Q, Zeng H, Deng J, Hou Y, Chen Z · Blood 2026
    PMID 42378254DOI
    What this evidence cannot show

    Rule 82 applies and was checked: this PMID is ALREADY CITED FOUR TIMES in this graph (exp-ck2-inhibitor-two-route-separation, gene-csnk2a1, mechanism-ck2-t2724-s2726-licenses-taspase1-cleavage, mechanism-set-pp2a-phosphatase-arm-kmt2a-r), all from the CK2 side, and this entry carries only the DEK-side content those do not - the GABPA-to-PBX3 mechanism and the bidirectional chemoresistance result. The systems are MYELOID: MLL-AF9 mouse, HOXA9/MEIS1 mouse and AML PDXs. There is no MLL-AF4, no B-lineage and no infant arm, and the chemotherapy backbone tested (azacitidine plus venetoclax) is an adult AML regimen, not an infant ALL one. Abstract-level evidence: the abstract was read in full, the full text was not, so the LSC-enrichment strategy, the PDX genotypes and the sparing of normal HSPCs are taken as the authors state them.

    Evidence that argues against this (1)
    • DEK is not a KMT2A contact, so this is a third CK2 route INTO THE SAME DISEASE rather than a third route into the same protein - which is gene-csnk2a1's own reading and is preserved here. The practical consequence, already written as an experiment in this graph, is that a CX-4945 phenotype in KMT2A-rearranged cells now has at least three candidate mediators (taspase1 licensing, the SET/PP2A arm, and DEK-4S), so any CX-4945 result attributed to one of them without a DEK phospho-readout is uncontrolled. That makes this evidence item a confounder for the CK2 axis as much as a finding about DEK. Oncofusion Knowledge Graph 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    DEK's mechanism literature is reachable through neither of this project's index-gene operands: DEK[tiab] = 600, AND leukemia[tiab] = 139, AND NUP214[tiab] = 70, AND histone[tiab] = 38, AND chaperone[tiab] = 6, AND heterochromatin[tiab] = 14, AND H3.3[tiab] = 5, AND nucleosome[tiab] = 11, AND HP1[tiab] = 1 (and that record is a KSHV LANA interactome paper), AND MLL[tiab] = 18 and AND KMT2A[tiab] = 7 - where all seven KMT2A records are cytogenetic panels, genomic-landscape surveys or fusion-transcript screening papers that list t(6;9) beside KMT2A rearrangements, and none is a DEK mechanism paper.

    What this evidence cannot show

    Titles and abstracts only; rule 80 applies. `DEK` is a short all-caps token and the counts should be read with that in mind - DEK[tiab] = 600 is small enough to suggest PubMed is not matching it as ordinary text, but no false-positive audit of those 600 was done here, so the denominator is unaudited. Registry row 124 was applied and does not rescue anything: AND MLL returns 18 and AND KMT2A returns 7, and neither set contains the heterochromatin, H3.3 or LSC papers this entry is built on. The seven KMT2A ids were resolved to titles rather than counted, per the rule that a control count is a list nobody opens.

    Evidence that argues against this (1)
    • The token deserves a collision check that this beat did not complete, and the risk runs towards over-counting. DEK is a three-letter all-caps string that is also a Dutch and Afrikaans word, an abbreviation used in several non-biological contexts, and part of hyphenated fusion names (DEK-NUP214, DEK-CAN) that a word-boundary regex treats differently from the bare symbol - which is measurable inside this corpus: a strict regex excluding hyphen-adjacent matches finds DEK in 23 entries, while a plain word-boundary regex finds 28. Five entries name DEK only as part of a hyphenated fusion name. A PubMed count on the bare symbol is therefore an upper bound of unknown looseness, and the specific mechanism counts (HP1 = 1, H3.3 = 5, chaperone = 6) are the ones to trust because their second operand does the work. Oncofusion Knowledge Graph 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    DEK is not a DepMap dependency and its mean gene effect is POSITIVE: +0.1433 across 1,258 screened cell lines with 0.0% of lines depleted below -0.5, and a lymphoid-tissue mean of +0.1311 over 97 lines; in this project's own KMT2A scan both arms are positive, +0.0971 in KMT2A-rearranged lines against +0.1320 in other leukaemia lines.

    What this evidence cannot show

    Rule 83: the API does not state the DepMap release; 1,258 is DEK's own screen count. The rank of 4,830 of 17,728 in this project's scan ranks a 0.035-unit difference between two POSITIVE means and is noise. Most importantly this figure does not measure what the leukaemia result is about: a pooled dropout screen across bulk cancer cell lines measures proliferation in culture, while the DEK result is about leukaemia-stem-cell maintenance, engraftment and chemoresistance in mouse models and PDXs. The two are not in conflict and should not be quoted against each other without that sentence attached.

    Evidence that argues against this (1)
    • The genetic and pharmacological experiments say DEK matters and they say it in the cell population that a screen cannot see. DEK deficiency impaired LSC maintenance and AML progression in MLL-AF9 and HOXA9/MEIS1 models and in PDXs; DEK overexpression conferred chemoresistance; CX-4945 depleted LSCs while sparing normal HSPCs. A +0.14 mean across 1,258 bulk lines is a statement about bulk proliferation, and this graph has already recorded the general form of this error - a coverage or dependency metric that measures a proxy rather than the thing it is named after. Lei Y et al. 2026
Connected to