GenestrongQ126-the-partners-named-but-never-written

NSD1 - the H3K36 dimethyltransferase whose catalytic activity a NUP98 fusion cannot do without, and which cells can otherwise lose almost for free

also called nuclear receptor binding SET domain protein 1 · KMT3B · ARA267 · androgen receptor-associated protein of 267 kDa · STO (PREVIOUS OFFICIAL SYMBOL, not an alias - for Sotos syndrome)

NSD1 is an enzyme that puts a chemical tag on the histone spools DNA is wound around, and losing one working copy of it causes Sotos syndrome, an overgrowth condition. In leukaemia it gets welded onto NUP98, and the resulting protein needs NSD1's enzyme activity to work - even though ordinary cells barely notice when NSD1 is deleted outright.

The same thing, for a scientist

**NSD1 is UniProt Q96L73 (NSD1_HUMAN, 2,696 aa), HGNC:14234 `nuclear receptor binding SET domain protein 1` at 5q35.3, primary Ensembl ENSG00000165671; `gene_exact:NSD1 AND organism_id:9606 AND reviewed:true` returned exactly one row, and its PREVIOUS official symbol is `STO`, which is a previous symbol and not an alias - rule 69 - so a literature search on NSD1 alone will miss the early Sotos-syndrome literature written under STO** [E1][E2]. It is the prototype of a three-member mammalian family (NSD1, NSD2/MMSET/WHSC1, NSD3/WHSC1L1) all of which are mutated in human disease [E4]. **Catalysis.** NSD1 is a SET-domain histone methyltransferase that **primarily dimethylates nucleosomal H3K36**; the 1.7 A structure of its catalytic domain shows a regulatory loop in a conformation that blocks free access of H3K36 to the bound SAM, with modelling suggesting the nucleosome stabilises the active conformation - an autoregulatory mechanism that also explains why NSD1 prefers nucleosomal over free-histone substrate [E3]. **Disease, wild type.** Haploinsufficiency of NSD1 is the major cause of **Sotos syndrome**: 1 nonsense, 3 frameshift and 20 submicroscopic deletion mutations among 42 sporadic cases, with the gene isolated from the 5q35 breakpoint of a translocation case [E5]. That is a **dosage** disease in a **developmental** setting, which is a different failure mode from the leukaemic one. **Disease, fused.** In AML the recurring t(5;11)(q35;p15.5) fuses NSD1 to NUP98. NUP98-NSD1 induces AML in vivo, sustains myeloid self-renewal, enforces HoxA7/A9/A10 and Meis1, binds genomic elements adjacent to HoxA7/A9, maintains H3K36 methylation and histone acetylation there, and prevents EZH2-mediated repression of Hox-A during differentiation. **Mutations in NSD1 that inactivate the H3K36 methyltransferase activity preclude Hox-A activation and myeloid progenitor immortalisation**, as does deleting the NUP98 FG repeats [E4]. So both halves are load-bearing and the 3' half is load-bearing *as an enzyme*. **And the number that makes the pair interesting.** Over 1,258 DepMap screens NSD1's mean Chronos gene effect is **-0.1231 with only 2.9% of lines below -0.5 and none below -1.0** - essentially dispensable - while NUP98 on the same call is -0.809 with 91.3% below -0.5 [E6]. **The fusion's indispensable catalytic half is the dispensable gene and its scaffold half is the near-pan-essential one.** **A brake upstream, from a preprint.** ZMYND11 co-immunoprecipitates KMT2A *and NSD1* from mouse brain and HEK293T, and the commonest ZRSID-associated missense, R600W, shows a **selective loss of the top interactors including KMT2A and NSD1 as well as MLL1-complex members ASH2L and RBBP5**; the paper then explicitly narrows to KMT2A and performs no functional experiment on NSD1 [E7]. Read carefully, this places a shared negative regulator upstream of two of this graph's fusion families - but it does so on one TMT-MS comparison, in brain and in HEK293T, with no orthogonal validation of the NSD1 arm and no haematopoietic system anywhere in the paper.

Go deeper

The ZMYND11 claim, read from the full text rather than restated

The dispatch brief for this beat said Q123 had found that "ZMYND11's R600W selectively strips both KMT2A and NSD1, so a ZMYND11 brake sits upstream of two of this graph's fusion families", flagged it as a preprint and told me to verify before building on it. I fetched PMC12636335 and read the relevant section. Q123's reading of the preprint is accurate, and the sentence is:

"Deletion of the MYND domain (ZMYND11-DeltaMYND) led to a drastic reduction in the majority of proteins interacting with ZMYND11, whereas the ZMYND11-R600W mutant showed a more selective loss of the top ZMYND11 interactors, including KMT2A and NSD1 as well as MLL1 complex members ASH2L and RBBP5 (Fig 5B, Fig S8F)." 7

Four qualifications that the one-sentence version loses, and that matter before anyone builds on it.

The NSD1 arm is TMT-MS only. The authors write "Due to the high level of enrichment of KMT2A in ZMYND11 immunoprecipitates ... we subsequently focused on KMT2A". The validation western (Fig 5C) is KMT2A. There is no orthogonal confirmation of the ZMYND11-NSD1 interaction and no functional NSD1 experiment in the paper.
The system is neurons and HEK293T. No haematopoietic cell, no leukaemia model, and no NUP98::NSD1 anywhere in the paper. The disease framing throughout is ZMYND11-related syndromic intellectual disability.
The 293T interactome is explicitly caveated by the authors as including "a wider range of transcriptional regulators and chromatin-associated proteins that may reflect lower-affinity interactions observed due to ZMYND11 overexpression".
It is a preprint, bioRxiv 2025, not peer reviewed 7.

One detail worth carrying forward that neither the brief nor Q123's entry records. ZMYND11's PWWP domain reads H3K36me3, and the authors show that the PWWP mutations F291A+W294A, which prevent H3K36me3 binding, do not affect ZMYND11 binding to KMT2A - so the KMT2A contact is H3K36me3-independent 7. NSD1 makes H3K36me2, not me3. A tidy story in which ZMYND11 reads NSD1's own mark and thereby restrains it is therefore not what this paper shows, and the co-IP is more likely a MYND-domain protein-protein contact than a mark-reading one. The graph should not build the mark-reading version.

Two diseases, two dose directions, and why that is the interesting part

	Sotos syndrome	NUP98::NSD1 AML
lesion	haploinsufficiency - deletion, nonsense, frameshift 5	gain of a chimera; the wild-type allele is untouched 4
what NSD1's catalysis does	lost, on one allele	required: catalytic-dead NSD1 abolishes Hox-A activation and immortalisation 4
tissue	developmental overgrowth, brain	myeloid progenitor
direction	less NSD1 activity	NSD1 activity redirected to a new address

The two are not opposite readings of one axis, and this entry should not be quoted as saying they are. Sotos is a dosage phenotype of the whole protein in development; the leukaemic lesion delivers NSD1's catalytic domain to HoxA under NUP98's control. What they share is that NSD1's enzymatic output matters in both, which is the argument for treating the SET domain rather than the protein as the object of interest.

And the DepMap number cuts against a naive reading of both. 2.9% of 1,258 lines fall below -0.5 6. Per rule 84 the null for "is this gene in the anti-selective half" is 50% by construction and is not reported here; what is reported is an absolute, and 2.9% is far outside anything a base rate explains. A gene that is nearly free to delete in a cancer cell line panel is nonetheless a dosage-sensitive developmental gene and an obligate catalytic half of a leukaemic fusion. Cell-line proliferation is simply not the assay those two facts live in - which is a caution that generalises to every DepMap figure in this corpus.

Facts and figures

uniprot	Q96L73
length AA	2696
gene primary	NSD1
hgnc	HGNC:14234
hgnc name	nuclear receptor binding SET domain protein 1
ensembl	ENSG00000165671
cytoband	5q35.3
accession resolution	gene_exact:NSD1 AND organism_id:9606 AND reviewed:true returned exactly one row, Q96L73, primary gene name NSD1; HGNC:14234 independently carries Q96L73 and ENSG00000165671. Both fetched live 2026-09-23. Rule 62 does not fire. Rule 69: STO is a PREVIOUS OFFICIAL SYMBOL and is the string the earliest Sotos literature was indexed under.
catalysis	SET-domain histone methyltransferase, EC 2.1.1.357, primarily DIMETHYLATES nucleosomal H3K36. Autoregulatory loop occludes the substrate channel in the free catalytic domain; nucleosome binding is modelled to stabilise the active conformation (PMID 21196496).
epigenetic readout	H3K36me2
mark written	H3K36me2 (nucleosomal, dimethyl-specific)
subcellular localization	nucleus
subnuclear localization	NOT ESTABLISHED BY AN EXPERIMENTAL DATABASE ANNOTATION. UniProt Q96L73's entire location block is 'Nucleus. Chromosome {ECO:0000305}.' - the nuclear call carries no evidence code and the chromosome call is ECO:0000305 curator inference, not experiment (rule 91 item 4). The only experimental chromatin placement this beat found is of the NUP98-NSD1 FUSION at genomic elements adjacent to HoxA7 and HoxA9 (PMID 17589499), which is not the wild-type protein.
localization shown for fusion	NUP98::NSD1, at Hox-A regulatory elements (PMID 17589499). Nothing for wild-type NSD1 in a haematopoietic cell was found in this beat.
relationship to existing entries	fusion-nup98-nsd1 holds the fusion, its epidemiology and its clinical behaviour; epi-nup98-nsd1-flt3-itd-confound holds the FLT3-ITD co-occurrence problem. This entry holds the GENE: accession and the STO previous-symbol trap, the H3K36me2 catalytic mechanism and its autoregulation, the Sotos haploinsufficiency phenotype, the dispensability/indispensability contrast against NUP98, and the verified-from-full-text state of the ZMYND11 preprint claim. It does not restate the fusion's clinical content.
relationship to the dispatch brief	The brief's premise about the ZMYND11 preprint was CHECKED AGAINST THE FULL TEXT AND HELD. What did not hold is the strength the one-sentence version implies: the NSD1 arm is a single unvalidated mass-spectrometry comparison in non-haematopoietic systems, and the authors' own next sentence narrows the paper to KMT2A.

alias symbol

[
  "ARA267",
  "FLJ22263",
  "KMT3B"
]

prev symbol

[
  "STO"
]

family

[
  "NSD1",
  "NSD2 / MMSET / WHSC1",
  "NSD3 / WHSC1L1"
]

cell types where function was measured

[
  "mouse myeloid progenitors / myeloid stem cells, for the fusion (PMID 17589499)",
  "human developmental tissue, inferred from the Sotos haploinsufficiency phenotype (PMID 11896389)",
  "mouse cortical neurons and HEK293T, for the ZMYND11 co-IP (PMID 41279818, preprint)"
]

cell states

[
  "differentiation: NUP98-NSD1 maintains H3K36 methylation and acetylation at Hox-A and prevents EZH2-mediated repression DURING DIFFERENTIATION, so the fusion's effect is state-conditional rather than constitutive (PMID 17589499)"
]

essentiality

{
  "depmap_source": "Open Targets GraphQL target.depMapEssentiality, fetched 2026-09-23",
  "depmap_release": null,
  "open_targets_dataVersion": "year 26, month 06, iteration null",
  "n_screens": 1258,
  "chronos_mean": -0.1231,
  "chronos_median": -0.1265,
  "frac_below_minus_0_5": 0.0286,
  "frac_below_minus_1_0": 0,
  "leukaemia_strata": {
    "Acute Myeloid Leukemia": {
      "n": 35,
      "mean": -0.1216
    },
    "B-Cell Acute Lymphoblastic Leukemia": {
      "n": 13,
      "mean": -0.2199
    },
    "T-Lymphoblastic Leukemia/Lymphoma": {
      "n": 8,
      "mean": -0.0384
    }
  },
  "no_disease_model_in_the_panel": "No NUP98::NSD1-rearranged line is known to this beat to be in DepMap; ana-depmap-nup98-men1-power already records the NUP98 line-availability problem. The figure bounds NSD1 knockout in unrelated lines and says nothing about the disease.",
  "human_haploinsufficiency": "Sotos syndrome - dosage-sensitive in development despite near-zero cell-line dependency (PMID 11896389)"
}

How this could bite us back

The obvious therapeutic reading - inhibit NSD1's SET domain, because catalytic-dead NSD1 abolishes transformation - carries a developmental liability that the leukaemia literature does not price. Losing ONE allele of NSD1 causes Sotos syndrome in a human. A systemic catalytic inhibitor in a child is a pharmacological approximation of that state in every tissue, and the phenotype includes overgrowth and learning difficulty rather than an acute, monitorable toxicity.
The NSD family shares a catalytic fold, and selectivity against NSD2 and NSD3 is not established for anything this beat found. An NSD1-directed agent that also hits NSD2 walks into multiple myeloma biology; one that hits NSD3 walks into a different set of amplified cancers. The bite-back is that the selectivity work has to precede the efficacy claim, not follow it.
The dependency evidence is structurally unavailable, not merely absent. There is no NUP98::NSD1 line in DepMap, so the only quantitative dependency figure anyone can quote for NSD1 is -0.1231 over lines that do not have the disease. An argument of the form 'DepMap says NSD1 is dispensable, so the fusion is not NSD1-dependent' is a category error and should be refused whenever it appears.
If the ZMYND11 brake is real, its therapeutic implication may run the wrong way. ZMYND11 loss-of-function is a neurodevelopmental disorder, so any strategy that mimics ZMYND11 to restrain NSD1 or KMT2A is restoring a brake whose absence is tolerated developmentally but whose presence is required for normal neuronal gene programmes - and the one drug the preprint shows rescuing ZMYND11 loss is revumenib, which points at KMT2A, not at NSD1.

Still open

Is the ZMYND11-NSD1 interaction real, and does it exist in a haematopoietic cell? The preprint's NSD1 arm is one TMT-MS comparison in brain and HEK293T with no validation blot and no functional test. A FLAG-ZMYND11 IP blotted for endogenous NSD1 in a myeloid line, with R600W as the published negative control, is a one-week experiment - and exp-revumenib-and-menin-eviction-in-zmynd11-mbtd1-aml already proposes an arm that would answer it as a by-product.
Does ZMYND11 restrain NUP98::NSD1, and can it? The fusion keeps NSD1's C-terminal catalytic region and discards NSD1's N-terminus. Whether the ZMYND11-contacting surface of NSD1 is inside or outside the retained portion was NOT determined in this beat - the preprint maps the contact on the ZMYND11 side only - and until it is, 'a ZMYND11 brake sits upstream of two fusion families' is an inference about a protein region nobody has located.
Where is wild-type NSD1 on chromatin in a haematopoietic cell? UniProt has no experimental localisation for it at all, and the only experimental chromatin placement in this beat's reading is of the fusion at Hox-A. If wild-type NSD1 does not normally occupy Hox-A, the fusion is a relocation; if it does, the fusion is an amplification. Those are different mechanisms and the graph currently assumes neither.
Do Sotos-syndrome patients have a measurable leukaemia excess, and if so, is it NUP98::NSD1? NSD1 haploinsufficiency and NUP98::NSD1 are the two human NSD1 lesions and they sit at opposite ends of the dosage axis. This beat did NOT search the Sotos cancer-risk literature and states no view; the question is named because it is cheap and would either connect or cleanly separate the two.
Is NSD1's dimethyl product specificity separable pharmacologically from NSD2's? The family shares the catalytic fold, NSD2 has advanced inhibitors, and any NSD1-directed agent for NUP98::NSD1 inherits that selectivity problem. No NSD1-selective chemical matter was found in this beat's reading and none is asserted.

Every source behind this entry (7)

E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
NSD1 is UniProt Q96L73 (NSD1_HUMAN), a 2,696-residue histone-lysine N-methyltransferase annotated as H3 lysine-36 specific (EC 2.1.1.357), with gene-name synonyms ARA267 and KMT3B, and a subcellular-location block reading only 'Nucleus. Chromosome {ECO:0000305}.'
UniProtKB entry Q96L73 (NSD1_HUMAN): reviewed record, length, gene names and subcellular-location comments · UniProt Consortium · UniProtKB/Swiss-Prot 2026
What this evidence cannot show
Fetched live 2026-09-23 via gene_exact:NSD1 AND organism_id:9606 AND reviewed:true, which returned exactly ONE row. RULE 91, ITEM 4 APPLIES: the entire localisation annotation is 'Nucleus' with NO evidence code at all, plus 'Chromosome' at ECO:0000305, which is CURATOR INFERENCE, not experiment. For a chromatin methyltransferase that is a striking gap, and it means this graph cannot cite UniProt for an experimental NSD1 localisation. The subnuclear dimension for NSD1 is therefore recorded here as UNMEASURED IN THE DATABASE rather than as a nuclear call. The one place this beat found NSD1 physically placed on chromatin by experiment is at the Hox-A locus as part of the NUP98 fusion (E4), which is the fusion, not the wild-type protein.
Evidence that argues against this (1)
- An absent evidence code is a statement about UniProt's curation, not about the literature, and this entry's 'subnuclear localisation unmeasured' framing overstates it. NSD1 localisation experiments may well exist and simply not be curated - which is this project's own `measured-not-curated` absence class, closable by reading rather than by wet work. The honest position is that this beat did not search for an NSD1 localisation paper, and the entry should not be quoted as evidence that none exists.
E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
HGNC:14234 assigns the symbol NSD1 the name 'nuclear receptor binding SET domain protein 1' at 5q35.3, with primary Ensembl gene id ENSG00000165671, UniProt Q96L73, alias symbols ARA267, FLJ22263 and KMT3B, and PREVIOUS SYMBOL STO.
HGNC record for NSD1 (HGNC:14234) · HUGO Gene Nomenclature Committee · HGNC REST 2026
What this evidence cannot show
Fetched live 2026-09-23; exactly one document returned. RULE 69: STO is listed in `prev_symbol`, not in `alias_symbol`. A previous official symbol is the string the older literature was indexed under, and this project has undercounted by tenfold before by searching only the current symbol. Anyone bounding an absence about NSD1 in the pre-2003 literature must run STO as well - and must expect heavy contamination, because 'STO' is also a common abbreviation and a mouse fibroblast line. That combination - a real retired symbol that is also a contaminated token - is a candidate row for data/symbol-collisions.json, where a grep returns no NSD1 or STO row today.
Evidence that argues against this (1)
- The rule-69 instruction to search the previous symbol has a real cost here that the instruction does not price. 'STO' is a three-letter token that matches an enormous amount of unrelated text - it is a common abbreviation and also a widely used mouse feeder-cell line name - so unioning it into a query trades a small recall gain for a large precision loss. Rule 88's own scope note says short tokens need word-boundary matching; for STO even that will not be enough, and a beat that runs it naively will report an inflated count.
E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
NSD1 is a SET-domain histone methyltransferase that primarily dimethylates nucleosomal histone H3 lysine 36; the 1.7 A structure of its catalytic domain shows a regulatory loop in a conformation that prevents free access of H3K36 to the bound S-adenosyl-L-methionine, and molecular dynamics plus computational docking indicate the loop can adopt an active conformation that the nucleosome may stabilise - an autoregulatory mechanism underlying its nucleosomal substrate selectivity and dimethyl product specificity.
The structure of NSD1 reveals an autoregulatory mechanism underlying histone H3K36 methylation · Qiao Q, Li Y, Chen Z, Wang M, Reinberg D, Xu RM · J Biol Chem 2011
PMID 21196496 DOI
What this evidence cannot show
Abstract read this session; full text not read. The structure is of the ISOLATED catalytic domain, not of NSD1 on a nucleosome, and the active conformation and the nucleosome's stabilising role are established by simulation and docking rather than by a nucleosome-bound structure. The PDB id was NOT fetched in this beat and is therefore deliberately not stated - per this project's standing rule, an identifier written without being fetched is exactly the error that keeps recurring. Anyone adding a structural figure must resolve it first.
Evidence that argues against this (1)
- A 2011 structure of an ISOLATED catalytic domain, with the active conformation reached by simulation and docking rather than observed, is the weakest form of a mechanistic claim, and the NSD family has since been the subject of nucleosome-bound structural work that this beat did NOT pull. The autoregulatory-loop model may have been superseded, refined or contradicted, and nothing in this entry would reveal it. Anyone building on the autoinhibition mechanism must check the post-2011 structural literature first.
E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In AML the recurring t(5;11)(q35;p15.5) fuses NSD1 to NUP98. NUP98-NSD1 induces AML in vivo, sustains self-renewal of myeloid stem cells in vitro, enforces HoxA7/A9/A10 and Meis1 expression, binds genomic elements adjacent to HoxA7 and HoxA9, maintains H3K36 methylation and histone acetylation there, and prevents EZH2-mediated repression of the Hox-A locus during differentiation; deletion of the NUP98 FG-repeat domain, OR mutations in NSD1 that inactivate the H3K36 methyltransferase activity, OR mutations that prevent binding of NUP98-NSD1 to the Hox-A locus, each preclude both Hox-A gene activation and myeloid progenitor immortalisation. NSD1 is the prototype of a mammalian family (NSD1, NSD2/MMSET/WHSC1, NSD3/WHSC1L1) essential in development and mutated in AML, overgrowth syndromes, multiple myeloma and lung cancer.
NUP98-NSD1 links H3K36 methylation to Hox-A gene activation and leukaemogenesis · Wang GG, Cai L, Pasillas MP, Kamps MP · Nat Cell Biol 2007
PMID 17589499 DOI
What this evidence cannot show
Abstract read this session; full text not read. Mouse myeloid progenitor system with retroviral expression, so expression level is non-physiological. The separation-of-function logic is the valuable part and it is a triple: FG deletion, catalytic-dead NSD1, and locus-binding-deficient NSD1 each abolish the phenotype - but all three are loss-of-phenotype arms in one model, and a loss of immortalisation is compatible with a construct that is simply destabilised. This paper is already cited in this corpus for the FG-repeat arm (protein-nup98-domains E6); it is cited here for the NSD1 CATALYTIC arm, which is the half a gene-nsd1 entry needs and which that entry does not carry.
Evidence that argues against this (1)
- Three loss-of-phenotype arms in one retroviral mouse model is weaker than it reads. A catalytically dead SET mutant can be destabilised or mislocalised, so 'catalysis is required' is confounded with 'the protein must be intact'. And the human disease is not the model: this corpus's own epi-nup98-nsd1-flt3-itd-confound records FLT3-ITD co-occurring with NUP98::NSD1 in 74-91% of cases, so a single-lesion mouse may be modelling a genotype that barely occurs in patients.
E5established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
NSD1 was isolated from the 5q35 breakpoint in an individual with Sotos syndrome carrying a chromosomal translocation; 1 nonsense, 3 frameshift and 20 submicroscopic deletion mutations of NSD1 were identified among 42 individuals with sporadic Sotos syndrome, indicating that haploinsufficiency of NSD1 is the major cause of Sotos syndrome.
Haploinsufficiency of NSD1 causes Sotos syndrome · Kurotaki N, Imaizumi K, Harada N, Masuno M, et al. · Nat Genet 2002
PMID 11896389 DOI
What this evidence cannot show
Abstract read this session; full text not read, and the author list is the four names in the fetched efetch record followed by 'et al.' rather than completed from recall. n = 42 sporadic cases in a single 2002 cohort, Japanese-ascertained, and the deletion proportion in that cohort (20/24 mutation-positive) is known to be population-dependent - later European series report a much lower microdeletion fraction. This beat did NOT pull those later series, so the 'major cause' conclusion is quoted as the paper's and the deletion/point-mutation ratio should not be quoted from it as a general figure.
Evidence that argues against this (1)
- A single 2002 cohort of 42 sporadic cases, ascertained in one population, cannot fix the mutation-class distribution of a syndrome, and the striking 20-of-24 microdeletion fraction is the part most likely to be population-specific. This beat did NOT retrieve the later series and therefore states no alternative figures - but the 'major cause' conclusion should be read as established while the deletion-versus-point-mutation ratio should not be quoted from this paper as a general property of Sotos syndrome.
E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
NSD1 is close to dispensable in DepMap: over 1,258 screens the mean Chronos gene effect is -0.1231, the median -0.1265, with 2.9% of lines below -0.5 and 0.0% below -1.0. NUP98, its 5' fusion partner, is -0.8090 with 91.3% below -0.5 on the same call.
Open Targets Platform GraphQL target.depMapEssentiality, genes resolved through the HGNC primary ensembl_gene_id (NSD1 = ENSG00000165671) · Open Targets Platform; Broad Institute DepMap · Open Targets Platform API 2026
What this evidence cannot show
Fetched live 2026-09-23; persisted with the per-gene numbers and leukaemia strata in analyses/q126-the-partners-named-but-never-written/depmap-essentiality.json (rule 90). RULE 83: meta.dataVersion is {year 26, month 06, iteration null}, the OPEN TARGETS platform release, NOT a DepMap release name; no DepMap release is recorded. Leukaemia strata for NSD1, reported as counts and not contrasts: AML n=35 mean -0.1216, B-ALL n=13 mean -0.2199, T-ALL n=8 mean -0.0384. CRITICAL LIMITATION: DepMap contains no NUP98::NSD1-rearranged line to this beat's knowledge - ana-depmap-nup98-men1-power in this corpus already records the NUP98 line-availability problem - so this figure says what NSD1 knockout does to unrelated cancer lines and says NOTHING about NSD1 dependency in the disease.
Evidence that argues against this (1)
- A near-zero dependency score is exactly what a proliferation screen returns for a gene whose two established disease roles are a developmental dosage phenotype and an obligate catalytic contribution to a chimera that no screened line carries. Reading -0.1231 as 'NSD1 does not matter' would be the error; the catalytic-dead NSD1 arm of PMID 17589499 abolishes myeloid immortalisation in the one system where the fusion exists. Wang GG et al. 2007
E7weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.preprint
In a 2025 bioRxiv preprint, ZMYND11 was identified as interacting with the H3K36 dimethyltransferase NSD1 as well as with KMT2A in mouse brain and in HEK293T; deletion of the MYND domain drastically reduced the majority of ZMYND11 interactions, whereas the ZRSID-associated R600W substitution showed a more selective loss of the top ZMYND11 interactors including KMT2A and NSD1 as well as MLL1-complex members ASH2L and RBBP5. The authors then state that because of the high enrichment of KMT2A they subsequently focused on KMT2A. Separately, PWWP-domain mutations F291A+W294A that prevent ZMYND11 binding to H3K36me3 did NOT affect ZMYND11 binding to KMT2A.
ZMYND11 Restrains KMT2A to Enable a Neuronal Developmental Program · Greben AW, Wu XS, Robb JE, Vogel IR, et al. · bioRxiv 2025
PMID 41279818 DOI
What this evidence cannot show
FULL TEXT READ THIS SESSION via Europe PMC fullTextXML for PMC12636335, and the sentences quoted in the section above are verbatim from it. Confidence is WEAK, not strong, and the reasons are specific: (1) it is a preprint, not peer reviewed; (2) the NSD1 arm rests on ONE TMT-MS comparison with NO orthogonal validation - the confirmatory western in Fig 5C is KMT2A, and the authors explicitly narrow to KMT2A; (3) no functional NSD1 experiment appears anywhere in the paper; (4) the systems are mouse cortical neurons and HEK293T, with no haematopoietic cell, no leukaemia model and no NUP98::NSD1; (5) the authors themselves attribute part of the 293T interactome to overexpression-driven low-affinity binding. The author list is the four names in the fetched efetch record plus 'et al.'; the DOI and PMCID were fetched, not recalled.
Evidence that argues against this (1)
- The attractive version of this - ZMYND11 reads H3K36me3 and thereby restrains the enzyme that writes H3K36 methylation - is NOT what the paper shows and should not be written into this graph. NSD1 makes H3K36me2, while ZMYND11's PWWP reads H3K36me3 (an SETD2 product), and the same paper shows the KMT2A contact survives PWWP mutations that abolish H3K36me3 binding. The ZMYND11-NSD1 co-IP is therefore most simply a MYND-domain protein-protein contact, and the mark-reading narrative is a coincidence of vocabulary. Greben AW et al. 2025

Connected to

part of →NUP98::NSD1, t(5;11)(q35;p15.5)
contributes the 3' half; its H3K36 methyltransferase activity is required - catalytic-dead NSD1 abolishes Hox-A activation and immortalisation
← fuses withNUP98 - the mobile nucleoporin that leaves the pore to put H3K4me3 on haematopoietic promoters, and the shared 5' donor of seven fusions in this graph
t(5;11)(q35;p15.5); the near-pan-essential scaffold half fused to the near-dispensable catalytic half
← bindsZMYND11 (BS69) - an H3.3K36me3 reader that restrains elongation, is itself the 5' partner of a chromatin-hijacking AML fusion driving HOXA/MEIS1/MYB, and whose promoter is the third-strongest MLL1-N bin in RS4;11
TMT-MS co-IP in mouse brain and HEK293T; R600W selectively strips NSD1 with KMT2A. PREPRINT, no orthogonal validation of the NSD1 arm, no functional NSD1 experiment
related to →NUP98::NSD1 without FLT3-ITD loses its prognostic significance — the confound that undermines the fusion's own reputation
the gene entry behind that fusion's epidemiology
related to →SETD2 (Q9BYW2) - the H3K36 trimethyltransferase that binds all seven MLL fusions, and the only one of them with a published KMT2A::AFF1 dependency test
NSD1 writes H3K36me2, SETD2 writes H3K36me3 - the distinction is what defeats the tidy 'ZMYND11 reads NSD1's own mark' story