Oncofusion

start / graph / drug-decitabine

Drugmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q81-methylome-axis

Decitabine - degrades DNMT1, and in ALL cells it degrades KMT2A too: the hypomethylating agent with two mechanisms and a menin-inhibitor combination already in the clinic

also called decitabine · 5-aza-2'-deoxycytidine · 2'-deoxy-5-azacytidine · DAC · Dacogen · decitabine/cedazuridine · ASTX727 · Inqovi

Decitabine is an old, approved leukaemia drug that gets built into DNA and traps the enzyme that copies methyl marks, destroying it. A 2025 paper found it also destroys KMT2A protein itself in leukaemia cells - so the drug that strips methylation may be hitting the fusion's own parent gene product, and it works together with menin inhibitors.

The same thing, for a scientist

**Decitabine (CHEMBL1201129, first approval 2006) is a deoxycytidine analogue that is incorporated into DNA and triggers selective proteasomal degradation of DNMT1** - shown to be specific to DNMT1 among DNMT1/3A/3B, to occur in the nucleus, to require a functional ubiquitin-activating enzyme and the DNMT1 KEN box, BAH domain and NLS, and notably **not to require covalent enzyme-DNA bond formation**, since a catalytic-site cysteine-to-serine mutant is still degraded [E1]. **In 2025 a second mechanism was reported that is directly this project's business: in ALL cell lines and xenografts, KMT2A protein was diminished in every line that experienced DNMT1 degradation, only cells with reduced KMT2A protein showed biological effects, and the authors report the effect can act synergistically with menin inhibition** [E2]. The structural rationale given is the similarity between the DNA-binding domains of DNMT1 and KMT2A - which in this graph's vocabulary is the CXXC domain that both proteins carry and that `ana-cxxc-domain-family-census-human` places in a family of exactly twelve. In KMT2A-rearranged infant ALL specifically, decitabine was the agent in the founding hypermethylation study: MLL-r but **not** MLL-wild-type lines showed dose- and time-dependent cytotoxicity and re-expression of 4 of 5 silenced genes, with methylation-specific PCR confirming a relative decrease in promoter methylation [E3]. Single-agent decitabine also significantly prolonged survival in KMT2A-rearranged infant ALL xenografts [E4]. Clinically it is now in a menin-inhibitor combination with data: **SAVE, an all-oral revumenib + decitabine/cedazuridine + venetoclax phase I-II in relapsed/refractory AML, 42 patients, 40% KMT2Ar, composite CR 71%, CR/CRh 60%, median CR/CRh duration not reached in the KMT2Ar subgroup** [E5]. **The bite-back is unusually sharp for a drug this old.** SAVE's minimum age is 12 years, so the index population is not in it [E6]; the triplet contains venetoclax and decitabine as well as revumenib, so no attribution to the pairwise interaction is possible; menin-binding-site mutations still emerged in 13% [E5]; and decitabine's primary target DNMT1 is a **pan-essential gene that KMT2A-rearranged lines depend on LESS than other leukaemia lines** - Chronos -0.572 versus -0.675, difference +0.103, rank 16,949 of 17,728 [E7].

Go deeper
Two mechanisms, and the second one is new to this graph's therapeutic reasoning

Mechanism 1 - DNMT1 degradation. Ghoshal 2005 is the mechanistic anchor and it is more specific than the usual summary. Among DNMT1, DNMT3A and DNMT3B, only DNMT1 is degraded; neither wild-type nor catalytic-mutant DNMT3A/3B is sensitive. Degradation is proteasomal, nuclear, blocked by proteasome inhibitors, requires a functional E1, occurs even in the absence of DNA replication, and requires the DNMT1 KEN box, BAH domain and NLS. A catalytic-site C-to-S mutant that cannot form the covalent intermediate is still degraded, so covalent trapping is not the requirement it is usually described as. 5-fluorodeoxyuridine and 5-fluorocytidine do not do it 1.

Mechanism 2 - KMT2A degradation. Brock 2025, in ALL cell lines and xenograft models: decitabine-induced demethylation raised DNMT1 and KMT2A expression, yet KMT2A protein fell in every line that degraded DNMT1, and only the lines with reduced KMT2A protein showed biological effects. KMT2A wild-type and KMT2A-rearranged cells arrested in G2 and G1 respectively, attributed to p27/p16. The paper's own stated rationale is structural similarity between the DNA-binding domains of DNMT1 and KMT2A, and it reports synergy with menin inhibition 2.

Why this matters here. DNMT1 and KMT2A are two of the twelve human CXXC proteins, and gene-dnmt1 already carries the finding that DNMT1's CXXC is the only heterologous CXXC that will run an MLL-AF9 leukaemia. If decitabine's second mechanism really runs through a shared DNA-binding module, then a hypomethylating agent is an accidental, low-potency CXXC-directed agent - which is a hypothesis the CXXC dossier should hold, and a claim that nothing yet proves.

Evidence in the index disease, ranked by how close it gets
systemresulthow close to infant KMT2A::AFF1
MLL-r infant ALL primary samples + cell lines 3global promoter hypermethylation; decitabine cytotoxic in MLL-r and not MLL-wild-type lines; 4 of 5 silenced genes re-expressed; MSP confirms methylation lossclosest - human, infant, MLL-r, but n = 5 primary samples for the methylome arm
infant KMT2A-r ALL xenografts 4single-agent decitabine significantly prolonged in vivo survivalhuman infant cells in mouse
adult ALL lines and xenografts 2KMT2A protein degradation; synergy with menin inhibitionadult ALL, mixed KMT2A status
R/R AML, age >=12 5triplet with revumenib and venetoclax, CRc 71%adult and adolescent AML, not the index disease

Rule 73 applied to the headline: the strongest clinical number for decitabine in a menin combination comes from relapsed or refractory AML in patients aged 12 and over, not from infant B-lineage KMT2A::AFF1, and the first clause of any claim built on it must say so.

The DepMap row that argues against the target

DNMT1 in the genome-wide KMT2A-selectivity scan: mean Chronos -0.5723 in the 11 KMT2A-rearranged lines, -0.6754 in the 43 other leukaemia lines, -0.5957 across all 1,208 lines. Difference +0.103, rank 16,949 of 17,728 (95.6th percentile of the anti-selective direction) 7.

Read carefully, three things are true at once and they are easy to conflate:

  1. DNMT1 is strongly essential in nearly everything (-0.60 across 1,208 lines). That is a therapeutic-index problem, not a selectivity argument.
  2. KMT2A-rearranged lines need it LESS than other leukaemia lines. Rule 84 demands the null: being in the anti-selective half is 50% by construction and this graph's base rate is 45.3%. The 95.6th percentile is well beyond that null, so the direction is a real signal rather than a coin flip.
  3. It does not follow that decitabine will not work. Decitabine kills by DNA damage and by re-expression of silenced genes, not by removing a dependency, and the MLL-r-selective cytotoxicity in 3 is measured directly and runs the other way. A CRISPR knockout of DNMT1 and a hypomethylating nucleoside are not the same perturbation, which is exactly the distinction rule 89's family of errors keeps collapsing.

The honest summary is: the genetic dependency data give a KMT2A-rearranged cell no special reason to need DNMT1, and the pharmacology gives it a reason to be sensitive to decitabine anyway. Those are compatible and the second is better evidence for the clinic.

Name list, because an absence about this drug is bounded by it

Rules 70-72: the name list is the instrument. Measured 2026-09-21 on PubMed:

stringrecords (tiab)
decitabine3,057
"5-aza-2'-deoxycytidine"4,224 - MORE than the INN
DAC3,497, of which 2,909 (83%) mention neither decitabine nor azacitidine
Dacogen46
decitabine AND azacitidine647 (21% of decitabine papers)

Two traps follow and neither is in data/symbol-collisions.json today. DAC is overwhelmingly not this drug - the off-topic hits sampled are a dual-atom catalyst, digital arterial calcification, direct air capture, double anterior chamber and double arterial cannulation. And the chemical name reaches more records than the INN, so a decitabine query that omits 5-aza-2'-deoxycytidine is missing the larger half of the literature. Rows for both are proposed in this beat's memo; the registry file was not edited.

Facts and figures
chembl IDCHEMBL1201129
targetgene-dnmt1
secondary target proposedgene-kmt2a - protein reduced in every decitabine-responsive ALL line that degraded DNMT1 (PMID 39754404); mechanism proposed on DNA-binding-domain similarity, NOT demonstrated as direct engagement
modalitysmall molecule (deoxycytidine nucleoside analogue, incorporated into DNA)
statusApproved (ChEMBL max_phase 4, first approval 2006); oral fixed-dose decitabine/cedazuridine (ASTX727) is a separate later approval and is the form used in SAVE
mechanism summaryDNA incorporation -> selective proteasomal degradation of DNMT1 (not DNMT3A/3B), nuclear, replication-independent, KEN-box/BAH/NLS-dependent, and NOT requiring covalent enzyme-DNA bond formation. A second reported mechanism in ALL cells is loss of KMT2A protein in the same cells that degrade DNMT1.
shown for fusionCell lines and primary samples: MLL-rearranged infant ALL lines are selectively sensitive (PMID 20215641); infant KMT2A-r ALL xenografts show a survival benefit (PMID 36380143); adult ALL lines show KMT2A protein loss (PMID 39754404). Clinically, only in a triplet in R/R AML aged >=12 (SAVE). NO trial of decitabine in infant KMT2A::AFF1 ALL.
measured on2026-09-21, PubMed esearch, querytranslation inspected
retracted citation flaggedPMID 15899874 is RETRACTED (RetractionIn PMID 35412342); demoted to weak and replaced as support by PMID 41460614.
trials
[
  "NCT05360160"
]
key ae
[
  "myelosuppression: in SAVE grade >=3 febrile neutropenia 36%, lung infection 21%, thrombocytopenia 21% (triplet, not attributable to decitabine alone)",
  "differentiation syndrome 10% in SAVE (5% grade 3) - a menin-inhibitor effect in that combination"
]
resistance
[
  "emergent menin-binding-site mutations in 13% of SAVE patients despite the triplet",
  "ABC-transporter upregulation proposed as the mechanism of hypomethylating-agent antagonism with methotrexate in infant KMT2A-r ALL samples"
]
name list for absence bounding
[
  "decitabine (3,057 tiab)",
  "5-aza-2'-deoxycytidine (4,224 tiab - larger than the INN)",
  "2'-deoxy-5-azacytidine",
  "DAC (3,497 tiab, 83% off-target)",
  "Dacogen (46 tiab)",
  "ASTX727",
  "Inqovi",
  "cedazuridine (as the oral combination partner)"
]
How this could bite us back
  • DNMT1 is essential in nearly every cell line screened (Chronos -0.60 across 1,208), so decitabine's therapeutic index is bought by pharmacokinetics and schedule, not by target selectivity - and in a neonate, marrow and gut toxicity are the limiting reality rather than an abstraction.
  • Hypomethylating agents were ANTAGONISTIC with methotrexate in a subset of infant KMT2A-r ALL samples, with ABC-transporter upregulation as a candidate mechanism. Methotrexate is backbone therapy in Interfant, so a decitabine arm added to standard infant therapy could be actively harmful at the specific point where it overlaps.
  • Decitabine is a DNA-hypomethylating agent given to infants whose entire future is ahead of them. Global demethylation reactivates transposable elements and imprinted loci; this graph already holds hyp-transposable-element-reactivation-kmt2a-r, and no long-term follow-up of methylome-directed therapy in infancy exists at all.
  • If decitabine really lowers KMT2A protein, it lowers WILD-TYPE KMT2A as well - the allele the patient needs for normal haematopoiesis and, in germline haploinsufficiency, for neurodevelopment. A drug that degrades KMT2A in an infant is not obviously the same proposition as one that blocks a fusion interface.
  • The best combination evidence (SAVE) is a triplet in patients aged 12 and over with 40% KMT2A-rearranged AML. Reading it as support for decitabine plus a menin inhibitor in infant B-lineage ALL crosses age, lineage, disease and line of therapy in one step.
Still open
  1. Is the KMT2A species that decitabine lowers the full-length protein, the taspase1 N320/C180 fragments, or the fusion? The 2025 abstract does not say and this graph's whole argument turns on which.
  2. Is the reported decitabine-menin-inhibitor synergy attributable to KMT2A protein loss, or to hypomethylation re-expressing genes that menin inhibition also de-represses? No experiment separates them, and the SAVE triplet cannot.
  3. Does decitabine change the unmethylated-CpG landscape enough to move fusion occupancy? Bio-CAP-seq plus fusion ChIP before and after decitabine in SEM is a two-week experiment that nobody has run.
  4. Why did azacitidine priming fail to improve survival in infants when decitabine is selectively cytotoxic to MLL-r lines and prolongs xenograft survival? Schedule, drug choice, or the preclinical model?
  5. Decitabine and azacitidine are not interchangeable - azacitidine is a ribonucleoside with substantial RNA incorporation. Has anyone compared them head-to-head in an infant KMT2A-rearranged system beyond the additivity screen in PMID 36380143?
Every source behind this entry (8)
  1. E1weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    5-aza-deoxycytidine (decitabine) induces selective proteasomal degradation of DNMT1 but not DNMT3A or DNMT3B; the degradation is nuclear, requires a functional ubiquitin-activating enzyme and the DNMT1 KEN box, BAH domain and nuclear localization signal, occurs even without DNA replication, and does NOT require covalent enzyme-DNA bond formation, since a catalytic-site cysteine-to-serine mutant is still degraded.

    PMID 15899874DOI
    What this evidence cannot show

    *** THIS CITATION IS RETRACTED. *** PMID 15899874 (Mol Cell Biol 2005) carries PublicationType 'Retracted Publication'; RetractionIn Mol Cell Biol 2022;42(5):e0054621, PMID 35412342. Found at the coordinator 2026-09-21 by running scripts/check-retracted-sources.py, which the beat did not run. Retained only as the historical source of the KEN-box/BAH/NLS dissection; it must not be cited as support. The surviving claim - that hypomethylating agents deplete DNMT1 - rests on PMID 41460614 (Peng 2025, Cell Oncol), which is independent and not retracted. Cell-line biochemistry with ectopic mutant constructs, 2005, no leukaemia model and nothing KMT2A-related. The APC/Cdh1 arm of the mechanism rests on overexpression and siRNA in the same system. Read from abstract only - the figures were not opened.

    Evidence that argues against this (2)
    • A retracted paper cannot support a mechanistic claim, and the detailed domain dissection drawn from it is exactly the content a retraction calls into question. The claim survives in narrower form on an independent source. Peng XM et al. 2025
    • The standard textbook account of decitabine has covalent trapping of DNMT1 at the incorporated azanucleoside as the proximal cause of degradation; this paper's catalytic-mutant result contradicts that as a requirement, so the mechanism most often cited for this drug is at best incomplete. Brock L et al. 2025
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In adult acute lymphoblastic leukaemia cell lines and xenograft models, decitabine diminished KMT2A protein in every line that experienced DNMT1 degradation, only cells with reduced KMT2A protein showed biological effects (G2 arrest in KMT2A wild-type and G1 arrest in KMT2A-rearranged cells, attributed to p27/p16), and the authors report that this mode of action can act synergistically with menin inhibition.

    PMID 39754404DOI
    What this evidence cannot show

    Read from abstract only in this beat; the synergy claim, the line panel, the combination index and whether the degraded species is full-length KMT2A, the taspase1 fragments or the fusion protein were NOT verified against the figures - and that last distinction is the one this project cares about most. Stated as ADULT ALL; the abstract's framing is a proposal to extend decitabine from AML to ALL. 'Degradation' in the title is inferred from reduced protein against raised transcript, which is consistent with degradation but is not a direct turnover measurement in the abstract. This entry's confidence is moderate for that reason.

    Evidence that argues against this (1)
    • Decitabine's target DNMT1 is essential across essentially all cell lines, and KMT2A-rearranged lines depend on it LESS than other leukaemia lines, so an anti-leukaemic effect ascribed to a KMT2A-directed mechanism has to be separated from generic DNMT1-loss toxicity - which the abstract does not do. Oncofusion knowledge-graph project 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In human MLL-rearranged infant ALL, promoter CpG-island hypermethylation is global relative to both normal marrow and other childhood ALLs (hypermethylated:hypomethylated CpG ratios 1.7 and 2.9, P < .001, HELP assay, n = 5 per group), and decitabine produced dose- and time-dependent cytotoxicity in MLL-r but NOT MLL-wild-type cell lines with re-expression of 4 of 5 silenced genes and a methylation-specific-PCR-confirmed decrease in promoter methylation.

    PMID 20215641DOI
    What this evidence cannot show

    n = 5 MLL-r infant samples, 5 other childhood ALL, 5 normals - very small for a genome-wide methylation comparison, and HELP is an enrichment assay rather than bisulfite. The MLL-r-selective cytotoxicity is in cell lines, and cell-line MLL status confounds with lineage and with many other genotype differences. Read from abstract only.

    Evidence that argues against this (1)
    • A clinical test of exactly this rationale - hypomethylating priming in newly diagnosed infant KMT2A-rearranged ALL - was safe and pharmacodynamically active and did NOT improve event-free survival, so the preclinical selectivity did not translate. Guest EM et al. 2024
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In infant KMT2A-rearranged ALL xenografts, single-agent decitabine (and single-agent azacitidine) significantly prolonged in vivo survival; in vitro, hypomethylating agents were additive with most conventional chemotherapeutics but ANTAGONISTIC with methotrexate in a subset of samples, with ABC-transporter upregulation identified as a candidate mechanism.

    PMID 36380143DOI
    What this evidence cannot show

    Xenograft survival, not cure; the headline combination the paper advances is azacitidine plus venetoclax rather than decitabine. Read from abstract only; the number of xenografts and the survival statistics were not extracted. The methotrexate antagonism is reported 'in a subset of samples' without a stated proportion.

    Evidence that argues against this (1)
    • Methotrexate is a backbone agent in every infant ALL protocol including Interfant, so an antagonism between hypomethylating agents and methotrexate is a direct argument against the combination that a survival benefit in xenografts would otherwise motivate. Cheung LC et al. 2023
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In the SAVE phase I-II study of all-oral revumenib plus decitabine/cedazuridine plus venetoclax in 42 patients aged 12 and over with relapsed or refractory AML (40% KMT2A-rearranged, median two prior lines, 52% prior venetoclax), the composite complete remission rate was 71% and CR/CRh 60% with MRD negativity in 80% of those; median CR/CRh duration was not reached in the KMT2A-rearranged subgroup; differentiation syndrome occurred in 10%; and emergent menin-binding-site mutations occurred in 13%.

    PMID 42272166DOI
    What this evidence cannot show

    Single-arm, n = 42, three active drugs - no arm isolates the decitabine contribution and no pairwise revumenib+decitabine data exist. Median age 40, minimum eligible age 12 years, so the infant index population is excluded by protocol. Adult and adolescent AML, not infant B-ALL. An Erratum to this article exists (PMID 42492014) and was not retrieved in this beat; the numbers above are from the article record and should be re-checked against the erratum before use.

    Evidence that argues against this (1)
    • Menin-binding-site mutations still emerged in 13% of patients on the triplet, so adding a hypomethylating agent and a BCL2 inhibitor did not prevent the class's characteristic on-target resistance mechanism. Issa GC et al. 2026
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.trial_registry

    The SAVE protocol NCT05360160 is a phase I/II study of SNDX-5613 with decitabine/cedazuridine (ASTX727) and venetoclax in AML, recruiting since 2022-10-14 with an estimated enrolment of 43 and a MINIMUM ELIGIBLE AGE OF 12 YEARS, with no upper age limit and no whyStopped field set.

    NCT05360160
    What this evidence cannot show

    Read from the raw ClinicalTrials.gov API v2 on 2026-09-21 (not the MCP tool), fields limited to identification, status, design, interventions and eligibility ages. Registry status RECRUITING with estimated completion 2026-12-01; registry records lag reality and the enrolment figure shown is the estimate, not the 42 reported in the publication.

    Evidence that argues against this (1)
    • A minimum age of 12 means the entire index disease of this project - KMT2A-rearranged ALL in infants under one year - is outside this protocol, so the best evidence for a menin-inhibitor plus hypomethylating-agent combination comes from a population that does not include the patients the project is about. Guest EM et al. 2024
  7. E7moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    DNMT1, decitabine's primary target, is strongly essential across cell lines generally (mean Chronos -0.5957 across 1,208 lines) but KMT2A-rearranged lines depend on it LESS than other leukaemia lines: -0.5723 versus -0.6754, difference +0.1031, rank 16,949 of 17,728 (95.6th percentile).

    Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap CRISPR Chronos via the Open Targets Platform GraphQL API) · Oncofusion knowledge-graph project · analyses/depmap-kmt2a-selective-scan-full/results.json, generated 2026-09-19T10:12:06Z; Open Targets dataVersion year 26 month 06 2026
    What this evidence cannot show

    Rule 84: membership of the anti-selective half is ~50% by construction and this graph's base rate is 45.3%, so the informative part is the 95.6th percentile, not the sign. The 11 KMT2A-rearranged lines are myeloid-enriched and none is infant-derived, so lineage is confounded with fusion status. A CRISPR knockout and a hypomethylating nucleoside are different perturbations and this row does not predict decitabine sensitivity. Open Targets dataVersion 26.06; DepMap's own release identifier is not returned by that endpoint.

    Evidence that argues against this (1)
    • Directly measured pharmacology runs the other way: decitabine was cytotoxic in MLL-rearranged but not MLL-wild-type cell lines, which a lower genetic dependency on DNMT1 would not predict. Schafer E et al. 2010
  8. E8established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Decitabine is ChEMBL1201129, a small molecule with maximum phase 4 and a first approval recorded in 2006, carrying the synonyms 2'-deoxy-5-azacytidine, DAC and Dacogen.

    ChEMBL molecule record CHEMBL1201129 (DECITABINE) · ChEMBL, EMBL-EBI · ChEMBL database 2026
    What this evidence cannot show

    Fetched 2026-09-21. ChEMBL's first_approval is a single global year and does not distinguish indication, formulation or jurisdiction; the oral decitabine/cedazuridine fixed-dose combination (ASTX727/Inqovi) is a later and separate approval and is the form used in SAVE. No paediatric labelling is asserted here and none was checked against a regulatory registry in this beat.

    Evidence that argues against this (1)
    • Rule 75: a drug's availability can be regulatory rather than bibliographic, and this entry did NOT query an approval registry - so nothing here bounds whether decitabine is labelled for use in children or infants in any jurisdiction. ChEMBL et al. 2026
Connected to