start / graph / ana-no-hdac-inhibitor-result-in-this-disease-attributes-to-an-enzyme

AnalysisstrongQ39-acetylation-cycle

No HDAC-inhibitor result in this disease attributes to an HDAC enzyme - the only per-enzyme evidence is genetic, and it says the enzyme is not fusion-specific

also called the HDAC attribution problem · class-level versus enzyme-level HDAC claims · HDAC enzyme attribution audit

Every drug study of HDAC blockers in this leukaemia used a drug that blocks several of the eleven HDAC enzymes at once, so none of them can say which enzyme mattered. The one study that did answer the question used genetics instead of drugs - and its answer was that the enzyme it found, HDAC3, is needed just as much by two completely unrelated blood cancers.

The same thing, for a scientist

**CORRECTION, cycle 143: the HDAC6 clause of this entry is overturned by its own stated falsifier.** This entry declined HDAC6 because `(MLL[tiab] OR KMT2A[tiab]) AND HDAC6[tiab]` returns 0 records, and said it would change its mind given *"any HDAC6-selective agent tested in a KMT2A-rearranged model."* Querying drug names and cell-line names instead returns 15 records: PTG-0861 (HDAC6 IC50 5.92 nM, ~36x selective) is active against MV4-11 (KMT2A::AFF1) per PMID 32615502, and an artemisinin-derived selective HDAC6 inhibitor gives MV4-11 IC50 28 nM with a MOLM-13-Luc xenograft per PMID 41701111. See `gene-hdac6`. **The rest of this entry stands, and so does its central claim in weakened form** - the HDAC6 evidence is antiproliferative and not fusion-specific, and a dedicated B-ALL isoform study (PMID 25688158) points at HDAC1/2. **The SIRT6 and KAT6B clauses were re-tested the same way and both survived.** The beat was dispatched with the question: **does any HDAC-inhibitor result in KMT2A-rearranged leukaemia attribute to a specific HDAC?** This entry is the audit. **The answer is no from the pharmacology and yes-but from the genetics, and the two halves point in opposite directions.** Every compound in this disease's HDAC literature is multi-isoform by design: vorinostat, panobinostat, romidepsin, belinostat and chidamide all cover class I and/or class II including HDAC1, 2, 3 and 6 [E1]; class IIA-selective compounds (LMK-235, MC1568, TMP195) reach a **sub-class of four** and all three failed in vivo at maximum achievable dose in a KMT2A-rearranged ALL xenograft [E2]. **This is structurally identical to two failures this graph has already paid for**: CX-4945 is ATP-competitive and pan-catalytic-subunit, so every CK2 result is CK2-level rather than CSNK2A1- or CSNK2A2-level; and every BET result rests on a pan-BET inhibitor, so 'BRD3/4' has never been separated. **In all three families the tool cannot see the difference the claim requires.** The genetic exception is Matthews et al., who depleted individual Hdacs in three defined malignancies and found **Hdac3 required in all three** - MLL-AF9;Nras AML, PML-RARa APL and Emu-Myc lymphoma - with HDAC1 and HDAC2 needing **co-depletion** [E3]. So the only per-enzyme attribution available is simultaneously the evidence that the enzyme is not fusion-directed. The strongest single-enzyme claims in the beat are **not** from inhibitors at all: HDAC7 silenced by KMT2A::AFF1 and EZH2 [E4], and HDAC9 as a 211-patient expression signature [E5] - both expression findings, neither a drug result.

Go deeper

The audit table

Every HDAC claim in this disease found in this beat, scored on what it can attribute:

evidence	tool	isoforms engaged	attributes to an enzyme?
t(4;11) infant ALL cell death, connectivity mapping	unspecified HDACi	class-level	no
romidepsin in primary infant ALL lines, cytarabine potentiation	romidepsin	HDAC1, 2, 3	no
panobinostat xenograft efficacy, H2Bub depletion	panobinostat	pan	no
vorinostat in 50 infants (NCT02553460)	vorinostat + bortezomib	pan, plus a second drug class	no
chidamide + MI-3 synergy in MLL-r AML	chidamide	HDAC1, 2, 3, 10	no
chidamide + cladribine, HDAC2/c-Myc/RCC1	chidamide	HDAC1, 2, 3, 10	partly - by co-IP and ChIP, not by the compound
class IIA knockdown kills KMT2A-r ALL	shRNA HDAC4+5+7	three at once	no
LMK-235 / MC1568 / TMP195 in vivo	class IIA-selective	four (HDAC4,5,7,9)	no - and all failed
Hdac3 depletion in MLL-AF9;Nras	shRNA, per enzyme	one	YES - and it is required in two unrelated malignancies too
HDAC1+HDAC2 co-depletion	shRNA + RGFP233	two, deliberately	yes, to the PAIR
HDAC7 silenced by KMT2A::AFF1/EZH2	expression + rescue	one	yes, but not a drug result
HDAC9 signature in 211 patients	mRNA expression	one	yes, but not a drug result

Twelve rows. One drug result gets partway to an enzyme, and it does so by co-immunoprecipitation rather than by selectivity. The pattern is not a gap in the literature; it is a property of the chemistry.

The third instance of a pattern this graph should now name

This project has now hit the same structure three times in three unrelated protein families:

family	tool	why it cannot discriminate	what was wrongly attributable
CK2	CX-4945, TBB, DMAT	ATP-competitive, pan-catalytic-subunit	CSNK2A1 versus CSNK2A2 - and the symbols alias each other
BET	I-BET151, JQ1, I-BET762	conserved acetyl-lysine pocket; selectivity achieved is BD1/BD2, across paralogues	BRD2 versus BRD3 versus BRD4
HDAC	vorinostat, panobinostat, romidepsin, chidamide	zinc-binding pharmacophore conserved across the class	HDAC1 versus 2 versus 3 versus 6

The generalisation worth carrying: when a target family shares a conserved catalytic pocket or cofactor site, the inhibitor class will be family-selective and not member-selective, and every member-level claim in that literature is naming rather than measurement. The tell is easy to check and rarely checked - ask what the compound's mechanism of binding is, and whether the members share that site.

And the alias layer makes it worse in this beat specifically. gene_exact:HDAC7 in UniProt returns HDAC9 first, because HDAC9 carries HDAC7 and HDAC7B as synonyms and its own protein name reads 'Histone deacetylase 7B' 6. So in the one family where the pharmacology cannot separate the members, the nomenclature cannot reliably separate them either - and HDAC7 and HDAC9 are precisely the two class IIa enzymes with real findings in this disease, pointing in opposite therapeutic directions.

Bounded rejection: what was searched, read and not used

Recorded so a future cycle can tell a considered dismissal from an unasked question.

Enumerated and dropped as targets:

HDAC6. (MLL[tiab] OR KMT2A[tiab]) AND HDAC6[tiab] returns 0 records, query translation read, no [UID] rewrite 7. HDAC6 is the tubulin deacetylase and a major pharmacological focus elsewhere. What would change my mind: any HDAC6-selective agent tested in a KMT2A-rearranged model, or a complex-level result naming the aggresome pathway.
SIRT6. Returns 1 record (PMID 30448059), which is a DPY30 glucose-homeostasis paper and off-beat on title check 7. gene-sirt1 already exists in this graph and SIRT6 has no independent standing here. What would change my mind: an NAD+-dependent deacetylase result at a fusion target locus.
KAT6B. Not written. This graph's own per-gene analysis places KAT6B at DepMap rank 8,339 against KAT6A's 106 - a paralogue split of over 8,000 places - so KAT6B is the member that carries no signal. What would change my mind: a lymphoid or infant result, since the paralogue analysis is myeloid-weighted.

Papers that look on-beat by title and were judged not usable: PMID 32429384 (EGCG altering histone acetylation - human endothelial cells, not leukaemia); 28778402 (epigenetics in MDS - a review of a different disease); 25016802 (MLL in myeloma, Japanese-language review); 15774615 (MLL partial tandem duplication, a different lesion this graph treats separately); 12384433 (the HDAC inhibitor AN-9, 2002, selective toxicity to acute leukaemia but never pursued clinically and no KMT2A data).

Laboratories publishing repeatedly in this area whose work is now cited here rather than absent: the Princess Maxima / Stam group (class IIA, panobinostat/H2Bub, connectivity mapping), the Telethon Kids / Kees group (romidepsin), the Parra group (HDAC7), the Johnstone group (functional-genetic HDAC dissection).

Facts and figures

status	run
question	Does any HDAC-inhibitor result in KMT2A-rearranged leukaemia attribute to a specific HDAC enzyme?
answer	No from the pharmacology. The only per-enzyme attribution is genetic (Hdac3 shRNA), and that same experiment shows the requirement is shared with an unrelated APL and an unrelated B-lymphoma, i.e. it is not fusion-specific.
result summary	Twelve HDAC claims in this disease were scored on attributability. One drug result reaches an enzyme (HDAC2, via co-IP and ChIP rather than via compound selectivity). One genetic study reaches an enzyme and shows it is not fusion-specific. The two strongest single-enzyme claims (HDAC7, HDAC9) are expression findings, not drug results.
generalisation	When a target family shares a conserved catalytic pocket or cofactor site, the inhibitor class will be family-selective and not member-selective, and every member-level claim in that literature is naming rather than measurement.
alias layer	gene_exact:HDAC7 returns HDAC9 FIRST; HDAC9's protein name reads 'Histone deacetylase 7B'. In the one family where the pharmacology cannot separate members, the nomenclature cannot reliably separate them either - and HDAC7 and HDAC9 point in opposite therapeutic directions.

compounds and their coverage

{
  "vorinostat": "pan (class I and II, incl. HDAC1,2,3,6)",
  "panobinostat": "pan (class I and II)",
  "romidepsin": "class I (HDAC1,2,3)",
  "chidamide": "class I benzamide (HDAC1,2,3,10)",
  "belinostat": "pan",
  "LMK-235": "HDAC4/5",
  "MC1568": "class IIA",
  "TMP195": "class IIA",
  "RGFP966": "HDAC3-selective - exists, never tested in KMT2A-rearranged ALL",
  "RGFP233": "HDAC1/2-selective"
}

selective tools that exist but were never tested here

[
  "RGFP966 (HDAC3)",
  "RGFP233 (HDAC1/2)"
]

third instance of the pattern

{
  "CK2": "CX-4945/TBB/DMAT are ATP-competitive and pan-catalytic-subunit; CSNK2A1 is itself an alias symbol of CSNK2A2",
  "BET": "I-BET151/JQ1/I-BET762 occupy a conserved acetyl-lysine pocket; the selectivity achieved is BD1/BD2, which cuts across paralogues",
  "HDAC": "the zinc-binding pharmacophore is conserved across the class"
}

bounded absences

{
  "HDAC6": "0 records against MLL/KMT2A, query translation read, no [UID] rewrite; complex-level queries also returned nothing. Floor, not a zero - aliases HD6 and KIAA0901 not searched.",
  "SIRT6": "1 record, off-beat on title check (a DPY30 glucose-homeostasis study)",
  "KAT6B": "not written; this graph's own per-gene analysis places it at DepMap rank 8,339 against KAT6A's 106"
}

read and rejected

[
  "PMID 32429384 - EGCG and histone acetylation in human endothelial cells, not leukaemia",
  "PMID 28778402 - epigenetics in MDS, a review of a different disease",
  "PMID 25016802 - MLL in multiple myeloma, Japanese-language review",
  "PMID 15774615 - MLL partial tandem duplication, a different lesion",
  "PMID 12384433 - HDAC inhibitor AN-9, 2002, never pursued clinically, no KMT2A data"
]

How this could bite us back

The field's response to pan-HDAC toxicity has been isoform-selective compounds. In this disease the isoform-selective compounds that exist are sub-class-selective, and all three tested failed in vivo at maximum achievable dose - so the escape route from the toxicity problem has already been tried here and did not work.
Because no HDAC trial in this disease has measured a pharmacodynamic marker, a member-level rationale cannot even be falsified by a trial built on it. The rationale and the readout are both at the class level and the claim is at the member level.
The one per-enzyme attribution available, HDAC3, is required in two unrelated malignancies. A trial of an HDAC3-selective agent in KMT2A-rearranged leukaemia would very likely be positive and would mean nothing about the fusion - the exact trap this graph recorded for MOF, now available in a drug that could actually be dosed.
HDAC7 and HDAC9 are class IIa paralogues that share alias symbols, and one is claimed to need raising while the other is claimed to need lowering. A single database lookup error inverts the therapeutic recommendation with nothing in the returned record looking wrong.

Still open

RGFP966 (HDAC3-selective) and RGFP233 (HDAC1/2-selective) both exist and neither has been tested in a KMT2A-rearranged model. Why not, and what would they show?
Can an HDAC degrader achieve the isoform selectivity an inhibitor cannot? Degraders can exploit surface differences outside the conserved catalytic pocket, which is exactly where these enzymes differ.
Does the conserved-pocket generalisation predict other families in this graph where member-level claims are currently being made on family-level tools?
Given that HDAC7 and HDAC9 alias each other and point in opposite therapeutic directions, how many published class IIa results in this disease have the two transposed?

Every source behind this entry (7)

E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Vorinostat, romidepsin, belinostat and panobinostat are FDA-approved for haematological malignancies and inhibit class II and/or class I HDACs, including HDAC1, 2, 3 and 6.
Functional-genetic dissection of HDAC dependencies in mouse lymphoid and myeloid malignancies. · Matthews GM, Mehdipour P, Cluse LA, Falkenberg KJ, Wang E, Roth M, et al. · Blood 126(21):2392-2403 2015
PMID 26447190 DOI
What this evidence cannot show
Taken from the introduction of a primary paper rather than from labels or a pharmacology review, and cited only for the uncontroversial multi-isoform fact. Approval indications have changed since 2015 and this citation should not be used for regulatory status. Chidamide's isoform coverage (HDAC1, 2, 3, 10) is not from this source and is stated in this entry as background.
Evidence that argues against this (1)
- That these four drugs are approved and inhibit class I and II HDACs is exactly why an approval cannot license an enzyme-level claim. The same paper combined genetic and pharmacological approaches specifically to ask whether suppressing individual or multiple Hdacs phenocopies broad-acting HDAC inhibitors, which is a question that would not need asking if the approved compounds told you which enzyme mattered. Matthews GM et al. 2015
E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The HDAC4/5-selective inhibitor LMK-235 eradicates KMT2A-rearranged ALL cell lines and primary patient samples in vitro, but in a KMT2A-rearranged ALL xenograft model the maximum achievable dose was insufficient to induce anti-leukaemic effects in vivo, with similar results for the class IIA inhibitors MC1568 and TMP195, and LMK-235 with venetoclax did not prolong survival.
Selective inhibition of HDAC class IIA as therapeutic intervention for KMT2A-rearranged acute lymphoblastic leukemia. · Verbeek TCAI, Vrenken KS, Arentsen-Peters STCJM, Castro PG, van de Ven M, van Tellingen O, et al. · Communications Biology 7(1):1257 2024
PMID 39362994 DOI
What this evidence cannot show
Even the 'selective' compounds here are sub-class-selective, reaching HDAC4, 5, 7 and 9 rather than one enzyme, which is the point this entry cites them for. The authors attribute the in vivo failure to compound stability and efficiency rather than to the biology, and no target-engagement measurement is reported to distinguish those. Single xenograft model. Abstract read; PMCID PMC11450098, full text not fetched.
Evidence that argues against this (1)
- The in-vitro-to-xenograft failure is not necessarily an argument against the target. A compound that eradicates cells in culture and fails in vivo more often indicts exposure, half-life or protein binding than the enzyme, and the paper reports the failure at the maximum achievable dose - which is a statement about the molecule's pharmacology, not about whether class IIA HDACs matter in this disease. de Barrios O et al. 2025
E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In a functional-genetic dissection depleting individual Hdacs in three genetically distinct murine malignancies, Hdac3-depleted cells were selected against in competitive assays for all three tumour types - MLL-AF9;Nras(G12D) AML, PML-RARa APL and Emu-Myc lymphoma - and genetic co-depletion of Hdac1 with Hdac2 was required for a pro-apoptotic response, phenocopied by the HDAC1/2-specific agent RGFP233.
Functional-genetic dissection of HDAC dependencies in mouse lymphoid and myeloid malignancies. · Matthews GM, Mehdipour P, Cluse LA, Falkenberg KJ, Wang E, Roth M, et al. · Blood 126(21):2392-2403 2015
PMID 26447190 DOI
What this evidence cannot show
Mouse, shRNA, and the MLL-AF9 arm carries a cooperating Nras(G12D) mutation. This is the only per-enzyme HDAC attribution in an MLL-fusion model found in this beat, and the fact that it exists at all is what makes the surrounding pharmacological silence interpretable rather than merely absent. Abstract read; PMCID PMC4653767, full text not fetched.
Evidence that argues against this (1)
- The per-enzyme answer and the fusion-specificity answer are the same experiment. Because Hdac3 was required in an APL and a B-lymphoma as well as in the MLL-AF9 model, the only enzyme-level HDAC attribution available in this disease is simultaneously the evidence that the enzyme is not a fusion dependency - which is the same shape as the KAT8/MOF finding this graph already holds, arriving from the eraser half of the cycle. Valerio DG et al. 2017
E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
HDAC7 is epigenetically silenced by EZH2 and the KMT2A::AFF1 fusion protein in t(4;11) infant B-ALL, where low HDAC7 expression is associated with poor therapeutic response, and restoring HDAC7 expression with the menin inhibitor MI-538 enhanced by chidamide impairs proliferation in an HDAC7-dependent manner.
HDAC7 induction combined with standard-of-care chemotherapy provides a therapeutic advantage in t(4;11) infant B-cell acute lymphoblastic leukemia. · de Barrios O, Ocon-Gabarro I, Gusi-Vives M, Collazo O, Meler A, Romecin PA, et al. · Biomarker Research 13(1):99 2025
PMID 40722046 DOI
What this evidence cannot show
Cited in this entry specifically as a single-enzyme claim that does NOT come from an inhibitor, which is the audit's point. The therapeutic effect is stated to be only partially mediated by HDAC7. A directly opposing report holds that knocking HDAC7 down kills the same cells - see gene-hdac7 - and this entry does not resolve that. Abstract read; PMCID PMC12305908, full text not fetched.
Evidence that argues against this (1)
- This result points the OPPOSITE way from the rest of this beat and is the strongest single objection to the framing. If HDAC7 is silenced by EZH2 and the fusion, and LOW HDAC7 tracks poor therapeutic response, then the therapeutic move is to INDUCE HDAC7 - which is what the paper's own title proposes - and a pan-HDAC inhibitor would push the axis the wrong way in exactly the patients it is aimed at. de Barrios O et al. 2025
E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 211 paediatric patients with acute leukaemia, a distinctive signature for KMT2A (MLL) rearrangement was identified consisting of high HDAC9 and MEF2D expression regardless of age, KMT2A partner and lineage, with HDAC9 overexpression carrying adverse prognostic value regardless of KMT2A rearrangement; the authors frame the work as choosing more specific HDAC inhibitors for selected candidates because HDAC inhibitor toxicity due to lack of specificity has limited their use.
Paediatric patients with acute leukaemia and KMT2A (MLL) rearrangement show a distinctive expression pattern of histone deacetylases. · Vega-Garcia N, Malatesta R, Estella C, Perez-Jaume S, Esperanza-Cebollada E, Torrebadell M, et al. · British Journal of Haematology 182(4):542-553 2018
PMID 29978456 DOI
What this evidence cannot show
mRNA expression only - not protein, not activity, and an expression signature is not a dependency. Cited here because it is a single-enzyme claim arrived at without an inhibitor, and because the authors state the attribution problem this entry audits in their own opening sentence. Retrospective cohort; full text not retrieved.
Evidence that argues against this (1)
- An expression signature is not a dependency. High HDAC9 and MEF2D identify KMT2A-rearranged cases but say nothing about whether either is required, and the same authors frame their motivation as choosing a more specific inhibitor for selected candidates - a stratification argument, not a mechanistic one. HDAC9's MITR form is catalytically inactive, so a high transcript can mean more scaffold rather than more enzyme. Matthews GM et al. 2015
E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
A gene_exact UniProt query for HDAC7 in reviewed human records returns two rows with HDAC9 (Q9UKV0, 1011 aa) first, because HDAC9 carries HDAC7 and HDAC7B as gene synonyms and its protein name includes 'Histone deacetylase 7B (HD7) (HD7b)'; the intended record is Q8WUI4 (952 aa). The same query form for HDAC1, HDAC2, HDAC3, HDAC5, HDAC9, KAT5, KAT2A, KAT2B, BRD2 and BRD3 returned exactly one reviewed row each on the same day.
UniProtKB REST search, gene_exact:<symbol> AND organism_id:9606 AND reviewed:true with fields accession,id,gene_primary,gene_synonym,protein_name,length, run for eleven symbols on 2026-09-20 · UniProt Consortium · UniProtKB 2026
What this evidence cannot show
Row ordering in a UniProt TSV is not contractual and may change between releases; the transferable finding is that the query returns more than one reviewed row and that one of them is the same-subclass paralogue, not the specific ordering seen on this date. This is a live instance of rule 62 rather than a new rule, but it is a worse instance than the ones already logged because the wrong record's own PROTEIN NAME contains the searched symbol.
Evidence that argues against this (1)
- A database query returning the wrong gene first is a property of the query interface on the date it was run, not a stable fact - ranking changes as records are curated. The durable point is that HDAC9's protein name contains 'Histone deacetylase 7B', and that is a nomenclature collision which will still mislead a reader after the ranking changes. Vega-Garcia N et al. 2018
E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database
A title/abstract search for HDAC6 against MLL or KMT2A returns 0 records with a clean query translation and no [UID] rewrite, and a search for SIRT6 returns 1 record which is a DPY30 glucose-homeostasis study and off-beat on title check.
PubMed E-utilities esearch: (MLL[tiab] OR KMT2A[tiab]) AND HDAC6[tiab] -> count 0, warninglist 'No items found', translation read; and (MLL[tiab] OR KMT2A[tiab]) AND SIRT6[tiab] -> count 1 (PMID 30448059), title-checked. Run 2026-09-20. · NCBI E-utilities · PubMed 2026
What this evidence cannot show
Query translations were read for both, per this project's rule that a surprising zero must be checked for a parser rewrite; neither contained a [UID] term. Aliases were not exhaustively searched - HDAC6 also publishes as HD6 and KIAA0901, SIRT6 as SIR2L6 - so both counts are floors. Complex-level phrase searches on NuRD, Sin3, CoREST, SAGA, NuA4 and the BET phrase forms were run separately in this beat and none returned an HDAC6 or SIRT6 result either, which is what makes these absences worth recording rather than merely unsearched.
Evidence that argues against this (1)
- A subunit-keyed absence count is structurally wrong for a complex, as this project established when a paper knocking down Ikka, Ikkb, Ikkg and IRAK3 proved unreachable from any of those four symbols because its abstract framed everything at 'IKK/NF-kB'. Class I HDACs work in NuRD, Sin3 and CoREST, so a paper could knock down HDAC1 and HDAC2 while naming only the complex. The complex-level queries run in this beat returned 8 records for NuRD/Sin3/CoREST against MLL or KMT2A, none of which was an HDAC knockdown in a KMT2A fusion on title check - so the absence survives the complex-level test, but it is the complex-level test and not the symbol count that makes it defensible. NCBI E-utilities 2026

Connected to

supports →HDAC3 - the only HDAC in this disease whose requirement was established genetically rather than by a drug, and the same experiment shows it is not fusion-specific
HDAC3 is the only per-enzyme HDAC attribution in this disease and it is genetic, not pharmacological
refutes →Vorinostat (SAHA) - the only HDAC inhibitor given to newly diagnosed infants with ALL, in a 50-patient trial whose endpoint was tolerability and which reported four treatment-related deaths, all infectious
pan-HDAC coverage means no vorinostat result can attribute to an enzyme
refutes →Panobinostat - the HDAC inhibitor this graph has reasoned about for cycles without having an entry for, with strong KMT2A-rearranged xenograft activity and a terminated adult trial
pan-HDAC coverage means no panobinostat result can attribute to an enzyme
analogous to →BRD3 - named in the founding BET paper's own mechanism and in one title since, and not once separated from BRD4 by any experiment in this disease
the same conserved-pocket problem makes every BET result family-level rather than paralogue-level
analogous to →Silmitasertib (CX-4945) - a CKII inhibitor that works on KMT2A-rearranged leukemia by strengthening the allele the fusion does not have
CX-4945 is pan-catalytic-subunit, making every CK2 result CK2-level - the first instance of this pattern in the graph
supports →HDAC7 - the eraser the index fusion silences, where the proposed therapy is to INDUCE the HDAC, and two 2024-2025 papers point in opposite directions
the HDAC7/HDAC9 alias collision means nomenclature cannot separate the two class IIa enzymes the pharmacology already cannot separate
supports →MOF/KAT8 is required by both KMT2A::MLLT3 and NUP98::HOXA9 — which is why it is the wrong kind of convergence
the eraser half reaches the same verdict the writer half did: the attributable enzyme is not fusion-specific
← supportsZero of fifteen, again: the same worklist, the same wrong intermediate answer of thirteen, and the third recurrence of a gap list that cannot see a decision
declines KAT6B and SIRT6 and lists HDAC6 under Enumerated and dropped as targets
← refutesTesting this corpus's own recorded refusals instead of its gap list overturned the first refusal tried, and the refusal had named the falsifier itself
the HDAC6 clause is overturned by the falsifier the entry itself named
← refutesHDAC6 - the one HDAC family member this graph refused, whose refusal named a falsifier that has been in the literature since 2020 and fires in the index fusion's own cell line
HDAC6-selective agents at nM potency in MV4-11, a KMT2A::AFF1 line