Decitabine's target is a pan-essential gene, and KMT2A-rearranged lines depend on it LESS than other leukaemia lines - rank 16,949 of 17,728
also called DNMT1 therapeutic index · the anti-selective hypomethylating target
The enzyme decitabine destroys is needed by almost every cell that has been tested, so the drug's safety comes from dosing and timing rather than from hitting only cancer. And in the cell lines available, KMT2A-rearranged leukaemias need it slightly less than other leukaemias do, not more.
The same thing, for a scientist
**Two facts about DNMT1 from the same genome-wide scan, and they point in different unhelpful directions** [E1]. **(i) Essential-function toxicity.** Mean Chronos gene effect across **1,208 lines is -0.5957** - DNMT1 is strongly essential in the great majority of cells screened. A hypomethylating agent's therapeutic index is therefore bought by pharmacokinetics, schedule and the replication dependence of the lesion, not by target selectivity. In a neonate, the practical expression of that is marrow and gut. **(ii) Anti-selectivity.** KMT2A-rearranged lines depend on DNMT1 *less* than other leukaemia lines: **-0.5723 versus -0.6754, difference +0.1031, rank 16,949 of 17,728, 95.6th percentile.** **Rule 84 applies and is cleared:** membership of the anti-selective half is ~50% by construction and this graph's base rate is 45.3%, so a rank in that half would carry no information - but the 95.6th percentile is far beyond it, and the direction is a real signal. **What it does NOT license is the conclusion that decitabine will fail.** Directly measured pharmacology runs the other way: decitabine is cytotoxic in MLL-rearranged and not MLL-wild-type cell lines, re-expresses silenced genes, and prolongs survival in infant KMT2A-rearranged ALL xenografts [E2][E3]. **A CRISPR knockout and a hypomethylating nucleoside are different perturbations** - the drug kills by DNA damage from the trapped adduct and by gene re-expression, not by removing a dependency. **The risk this entry records is therefore specific: nothing in the genetic-dependency data gives a KMT2A-rearranged cell a special reason to need DNMT1, so any therapeutic argument of the form "DNMT1 is a KMT2A-rearranged dependency" is unsupported and mildly contradicted.** The clinical argument for a hypomethylating agent in this disease has to rest on the hypermethylation phenotype and on the drug's pharmacology, and **the one prospective test of that argument in infants moved the biomarker and not the survival curve** [E4].
Three statements that are easy to run together and must not be
| statement | true? | what it supports |
|---|---|---|
| DNMT1 is essential in most cells | yes, -0.60 across 1,208 lines | a therapeutic-index warning |
| KMT2A-rearranged lines need DNMT1 less | yes, +0.103, 95.6th percentile | an argument against calling DNMT1 a subtype dependency |
| Therefore decitabine will not work in KMT2A-rearranged leukaemia | no | nothing - and the direct pharmacology contradicts it |
The third is the error this entry exists to prevent, and it is the same family as rule 89's: a measurement that is near the claim is not the claim. A knockout scan measures whether a cell can live without a gene. A nucleoside analogue measures what happens when you put a covalent trap in replicating DNA. The second can be selectively lethal in cells the first says are not dependent, and in this case it demonstrably is.
The honest joint reading: the genetics say do not sell DNMT1 as a KMT2A-rearranged target; the pharmacology says decitabine may still be a useful drug here; and the one clinical test in the index population says the benefit did not appear.
| kind | essential-function |
|---|---|
| target | gene-dnmt1 |
| severity | moderate - a therapeutic-index constraint rather than a disqualifier; the class is approved and in routine use, and the specific risk here is to the argument rather than to the patient |
| mitigation | Do not argue for a hypomethylating agent in KMT2A-rearranged leukaemia on the grounds of a subtype dependency on DNMT1; there is none, and the scan runs the other way. Argue from the hypermethylation phenotype and the direct MLL-r-selective cytotoxicity instead, and state the AALL15P1 null alongside. |
| what this does not show | It does not predict decitabine failure. A CRISPR knockout and a covalent nucleoside adduct are different perturbations, and the direct pharmacology shows MLL-r-selective cytotoxicity that the dependency data do not anticipate. |
affects
[ "drug-decitabine", "drug-azacitidine" ]
depmap
{
"gene": "DNMT1",
"mean_kmt2a_rearranged": -0.5723,
"mean_other_leukemia": -0.6754,
"difference": 0.1031,
"pan_essentiality_all_lines": -0.5957,
"n_all_lines": 1208,
"rank_of_17728": 16949,
"percentile": 95.606,
"null_base_rate_anti_selective_half": "45.3% of this graph's gene entries, ~50% by construction",
"release_note": "Open Targets dataVersion 26.06, scan generated 2026-09-19"
}- DNMT1 is essential in nearly every cell line screened. In an infant the practical expression of that is marrow and gut, over a schedule repeated across four chemotherapy courses, in a patient with a normal life expectancy of eighty years if cured. The drug is approved and used, and that does not make the target selective.
- If decitabine also lowers KMT2A protein, it lowers the wild-type allele - which is required for normal haematopoiesis and, in germline haploinsufficiency, for neurodevelopment. Two essential-function exposures, one drug, and neither has long-term paediatric follow-up.
- The strongest argument for a hypomethylating agent in this disease has already been tested prospectively in infants and did not improve survival. Any new proposal has to explain what it is doing differently, and 'a different drug in the same class' is a hypothesis rather than an explanation.
- Which 11 cell lines are the KMT2A-rearranged arm of this scan, and is the DNMT1 difference driven by one or two of them? Rule 90 records that the line identities are held in a separate analysis and this row was not checked at line level.
- Is DNMT1 dependency different in B-lineage versus myeloid KMT2A-rearranged lines? The scan's arm is myeloid-enriched and the index disease is B-lineage.
- What is the actual mechanism of decitabine's MLL-r-selective cytotoxicity, if not a DNMT1 dependency? The candidates are the hypermethylation burden itself, the reported KMT2A protein loss, and differential nucleoside transport - and nothing distinguishes them.
- In an infant, what is the marrow recovery time after hypomethylating exposure relative to an older child? AALL15P1 reports 6.5% dose-limiting toxicity but the abstract does not give count recovery kinetics.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In the genome-wide KMT2A-selectivity scan, DNMT1 has a mean Chronos gene effect of -0.5957 across 1,208 cell lines, -0.5723 in the 11 KMT2A-rearranged lines and -0.6754 in the 43 other leukaemia lines, giving a difference of +0.1031 and rank 16,949 of 17,728 (95.6th percentile) in the anti-selective direction.
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 2026What this evidence cannot show
Rule 84: membership of the anti-selective half is ~50% by construction with a graph base rate of 45.3%, so only the 95.6th percentile carries information, not the sign. The 11 KMT2A-rearranged lines are myeloid-enriched and none is infant-derived, so fusion status is confounded with lineage. Rule 90: the scan persists gene-level statistics but the resolved cell-line identities are recorded in a separate analysis, so a line-level check of this row was not performed here. Open Targets dataVersion 26.06; DepMap's own release identifier is not returned by that endpoint. The difference of 0.103 is small in absolute terms against a pan-essential baseline of -0.60.
Evidence that argues against this (1)
- A CRISPR knockout removes the protein; decitabine traps it on DNA and destroys it while leaving a covalent adduct behind. The two perturbations differ in the lesion they create, and a gene that is not a selective dependency can still be a selective drug target - which is what the direct cytotoxicity data show here. Schafer E et al. 2010
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Decitabine produced dose- and time-dependent cytotoxicity in MLL-rearranged but NOT MLL-wild-type cell lines, with re-expression of 4 of 5 silenced genes and methylation-specific-PCR-confirmed loss of promoter methylation - a selectivity the genetic dependency data do not predict.
Promoter hypermethylation in MLL-r infant acute lymphoblastic leukemia: biology and therapeutic targeting. · Schafer E, Irizarry R, Negi S, et al. · Blood 2010What this evidence cannot show
Cell lines; MLL status in a cell-line panel is confounded with lineage and with every other genotype difference between the lines, so 'MLL-r selective' here is an association across a small panel rather than an isogenic comparison. The methylome arm is n = 5 per group by HELP. Read from abstract only.
Evidence that argues against this (1)
- A prospective trial of hypomethylating priming in exactly the index population demethylated 87% of treated infants and produced no survival benefit, so the cell-line selectivity did not translate to the clinic. Guest EM et al. 2024
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Single-agent decitabine and single-agent azacitidine each significantly prolonged in vivo survival in infant KMT2A-rearranged ALL xenografts, and hypomethylating agents were antagonistic with methotrexate in a subset of samples via candidate ABC-transporter upregulation.
Preclinical efficacy of azacitidine and venetoclax for infant KMT2A-rearranged acute lymphoblastic leukemia reveals a new therapeutic strategy. · Cheung LC, Aya-Bonilla C, Cruickshank MN, et al. · Leukemia 2023What this evidence cannot show
Xenograft survival prolongation, not cure; the number of models and the survival statistics were not extracted, and 'a subset of samples' for the methotrexate antagonism is not quantified in the abstract. Read from abstract only. Xenografts lack an immune system and a normal marrow niche, both of which bear on a drug whose target is pan-essential.
Evidence that argues against this (1)
- A xenograft cannot report the toxicity half of a pan-essential target: the recipient mouse's own marrow is not the readout of interest and the model has no normal human haematopoiesis to damage, so survival prolongation in a xenograft systematically flatters a drug against an essential gene. Oncofusion knowledge-graph project 2026
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The clinical consequence is on record: in infants with newly diagnosed KMT2A-rearranged ALL, azacitidine priming was safe (dose-limiting toxicity in 2 of 31 evaluable, 6.5%), demethylated 87% of tested peripheral blood mononuclear cell samples by whole-genome bisulfite sequencing, and did not affect event-free survival.
Azacitidine as epigenetic priming for chemotherapy is safe and well-tolerated in infants with newly diagnosed KMT2A-rearranged acute lymphoblastic leukemia: Children's Oncology Group trial AALL15P1. · Guest EM, Kairalla JA, Devidas M, et al. · Haematologica 2024What this evidence cannot show
A pilot for tolerability and biologic activity, not powered for efficacy; the survival comparison is historical rather than randomised. WGBS was on peripheral blood mononuclear cells, not sorted blasts. The trial used azacitidine, and azacitidine is a ribonucleoside whose biology differs from decitabine's - so this is a weaker test of decitabine than it appears. Read from abstract only.
Evidence that argues against this (1)
- Because hypomethylating agents are antagonistic with methotrexate and methotrexate is in the AALL15P1 backbone, the trial cannot distinguish 'no effect' from 'a real effect cancelled by an antagonism it was administered alongside'. Cheung LC et al. 2023
Essential-function risk on the drug's primary target, plus the finding that KMT2A-rearranged lines depend on DNMT1 less, not more.
Same target, same essential-function constraint, and the one prospective test in infants moved the biomarker and not the survival curve.
Chronos -0.5957 across 1,208 lines; -0.5723 in KMT2A-rearranged vs -0.6754 in other leukaemia lines; rank 16,949 of 17,728.
Where the therapeutic index actually bites: marrow and gut, across four courses, in a patient with a normal life expectancy of eighty years if cured.
The clinical expression of the risk-benefit: safe, pharmacodynamically active, no survival effect.
- ← confers resistance toDecitabine - 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
The essential-function and anti-selectivity risk attached to this drug's target.