Oncofusion

start / graph / target-phase2-kat7-dossier

Targetmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q14-phase2-readiness

Phase-2 dossier: KAT7/HBO1 - the only enzyme here with a real active site, a co-crystal inhibitor and a co-dependency that survives lineage adjustment almost intact

also called HBO1 dossier · MYST2 phase-2 plan · KAT7 readiness

KAT7 is an enzyme that puts a chemical tag on histones which leukaemia stem cells need to keep their key genes switched on. It has a proper drug-shaped pocket, a published inhibitor and a crystal structure showing exactly how that inhibitor binds - and unlike most of the others here, the evidence that leukaemia needs it more than normal cells do holds up when you correct for cell type.

The same thing, for a scientist

KAT7 (HBO1, MYST2) is a MYST-family acetyltransferase that deposits H3K14ac in complex with a BRPF-family scaffold. An shRNA and CRISPR domain screen in an ex vivo murine leukaemia-stem-cell model identified the HAT domain as required for LSC maintenance through H3K14ac sustaining Hoxa9 and Hoxa10, and the same work produced a potent acetyl-CoA-competitive inhibitor with a co-crystal structure. In this project's DepMap work KAT7's co-dependency with MEN1 retains 97% of its pooled value on within-tissue centring (0.214 to 0.208) and rises to 0.517 within leukaemia lines, and its lineage-controlled selectivity of -0.242 ranks fifth in the interactome. ChEMBL holds 98 activity rows at pChEMBL 6 or above, best reported Ki 8 nM. It has no physical interaction with KMT2A - membership in this interactome is circuit-level, through shared HOXA9 output - which is the principal reason to treat it as a combination partner rather than a stand-alone programme.

Go deeper
What structure exists, and what does not

6MAJ, the KAT7 HAT domain with a bound inhibitor and the BRPF2/BRD1 scaffold fragment, X-ray 2.14 A, verified against RCSB this cycle 1. A full set of acyl-CoA-bound structures (acetyl, propionyl, butyryl, crotonoyl, succinyl) is recorded in gene-kat7. Structure is not a limiting factor.

Does not exist: any structure or evidence of a direct KAT7-KMT2A contact. There is no interface to drug here; there is an enzyme to inhibit.

What pocket is engaged

The acetyl-CoA site of the MYST HAT domain, engaged by a competitive cofactor analogue 2. Cofactor-competitive inhibition is the easiest chemistry in this dossier set and the hardest selectivity problem: the MYST family (KAT6A, KAT6B, KAT8, TIP60) shares the fold, and KAT6A/KAT6B inhibitors already exist clinically for other indications. A MYST cross-reactivity panel is a G1 requirement, not a later refinement.

Essentiality and the normal-tissue picture

DepMap Chronos mean -0.300 across 1,258 lines, not flagged common-essential; -0.629 in KMT2A-rearranged lines; lineage-controlled selectivity -0.242, fifth in the interactome 3. The co-dependency with MEN1 retains 97% of its pooled value after within-tissue centring and reaches 0.517 in leukaemia lines 4 - among the most robust signals in the executed analysis after KMT2A, DOT1L and PSIP1.

GTEx v8: all 54 tissues, 6.75 TPM whole blood, minimum 3.28 TPM liver, maximum 34.1 TPM testis 5. gnomAD pLI 1.0, LOEUF 0.207 - as constrained as MEN1 6. The mouse knockout phenotype is not established in this graph by a verified citation, and that is a real hole: KAT7 is a developmental acetyltransferase and the germline phenotype is the single most informative missing datum for a paediatric programme.

Chemical matter

ChEMBL target CHEMBL3774299: 98 activity rows at pChEMBL 6 or above, best reported Ki 8 nM (CHEMBL5928174) 7, plus the inhibitor co-crystal 6MAJ. No clinical candidate. This is the best chemical-matter position of any non-clinical target in the beat.

What would have to be true for a phase-2 programme

Three things. (1) The dependency must be human and KMT2A::AFF1, not murine and MLL-AF9: the founding evidence is an ex vivo mouse LSC system 2. (2) The inhibitor must produce the mark change it claims - H3K14ac loss - in a KMT2A::AFF1 cell at a concentration that does not kill it by other means. (3) The MYST family must be spared, or the programme becomes a pan-MYST toxicity study.

Because KAT7 has no physical contact with the fusion, the honest framing is that this is a downstream-output target of the same class as DOT1L. That class has failed once already as monotherapy, which is why this dossier's verdict is 'combination partner', not 'lead programme'.

Facts and figures
target genegene-kat7
uniprotO95251
pocketacetyl-CoA site of the MYST HAT domain; engaged by a competitive cofactor analogue
contact with fusionnone documented - membership in this interactome is circuit-level, through shared HOXA9/HOXA10 output
known gapNo verified germline mouse knockout phenotype is recorded in this graph. For a developmental acetyltransferase being considered in infants, this is the most informative single missing fact and is a G1 item.
phase2 verdictSTART, as a combination partner and only after the LEDGF and menin programmes. Best chemistry of the non-clinical candidates, but no physical contact with the fusion places it in the same downstream-output class as DOT1L, which has already failed once.
compute backendModal or laptop for the ChEMBL/DepMap re-derivations; Boltz for MYST-family selectivity co-folds; no cluster time required.
family selectivity liability
[
  "KAT6A",
  "KAT6B",
  "KAT8",
  "KAT5/TIP60"
]
structures verified this cycle
[
  {
    "pdb": "6MAJ",
    "method": "X-ray",
    "resolution_A": 2.139,
    "ligand": "inhibitor (JAV)",
    "partner": "BRPF2/BRD1 31-80"
  }
]
essentiality
{
  "chronos_mean": -0.3,
  "chronos_mean_kmt2ar": -0.629,
  "selectivity_vs_leukemia": -0.242,
  "rank_in_interactome": 5,
  "depmap_common_essential": false,
  "pooled_r_vs_MEN1": 0.2141,
  "within_tissue_r_vs_MEN1": 0.2077,
  "leukemia_only_r_vs_MEN1": 0.5168
}
normal tissue gtex v8 tpm
{
  "whole_blood": 6.753,
  "min": {
    "tissue": "Liver",
    "tpm": 3.277
  },
  "max": {
    "tissue": "Testis",
    "tpm": 34.134
  },
  "n_tissues": 54
}
gnomad
{
  "pLI": 1,
  "loeuf": 0.207
}
chemical matter
{
  "chembl_target": "CHEMBL3774299",
  "rows_pchembl_ge_6": 98,
  "best": {
    "molecule": "CHEMBL5928174",
    "type": "Ki",
    "value_nM": 8,
    "pchembl": 8.1
  },
  "inhibitor_cocrystal": "6MAJ"
}
go no go
{
  "G1_feasibility": {
    "question": "Selectivity and the missing knockout",
    "GO": "A MYST-family biochemical counter-screen (KAT6A, KAT6B, KAT8, KAT5) shows at least 30-fold selectivity, and a verified germline knockout phenotype is located in the literature",
    "STOP": "Under 10-fold selectivity against KAT6A/KAT6B - the programme would be indistinguishable from an existing KAT6 programme"
  },
  "G2_biology": {
    "question": "Does the mark change, and does the programme follow?",
    "GO": "H3K14ac reduced by at least 50% and HOXA9 mRNA by at least 50% in SEM or MV4-11 at 10x the biochemical Ki, below the 72 h GI20",
    "STOP": "H3K14ac reduced by at least 50% with HOXA9 unchanged - the mark is not load-bearing in this fusion",
    "ambiguous_zone": "HOXA9 reduction 20-50%",
    "tie_breaker": "KAT7 HAT-domain CRISPR knock-in of a catalytically dead allele in SEM; binding"
  },
  "G3_window": {
    "question": "Normal HSC cost, and the paediatric development question",
    "GO": "Under 30% loss of healthy-donor CD34+ CFU at the G2-effective concentration, and no germline knockout phenotype that implicates skeletal or neural development",
    "STOP": "An embryonic-lethal or severe developmental knockout phenotype combined with more than 50% CFU loss"
  }
}
How this could bite us back
  • KAT7 has no physical contact with the fusion. Inhibiting it suppresses a downstream output, which is escapable by any route that restores HOXA9 - the same structural weakness that sank pinometostat.
  • The MYST family is closely related; a non-selective compound is a pan-acetyltransferase agent with unpredictable developmental toxicity in a child.
  • KAT7 is LOEUF 0.207, as constrained as MEN1, and its knockout phenotype is unrecorded in this graph. Starting medicinal chemistry before that phenotype is known risks discovering the toxicity after the money is spent.
  • The founding evidence is a murine LSC screen; murine LSC dependencies have repeatedly failed to transfer to human infant ALL.
Still open
  1. What is the germline Kat7 knockout phenotype in mouse, and does it implicate tissues that are still developing in an infant?
  2. Is the KAT7 dependency specific to KMT2A-rearranged disease or general to HOXA-high AML stem cells - and if general, is there any window at all against normal HSC?
  3. Can a cofactor-competitive MYST inhibitor achieve KAT7 selectivity over KAT6A in cells, where acetyl-CoA concentration differs from the biochemical assay?
Every source behind this entry (7)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    An inhibitor-bound KAT7 HAT domain structure exists (PDB 6MAJ, X-ray 2.139 A), verified by direct query of the RCSB data API this cycle.

    What this evidence cannot show

    The structure contains the BRPF2/BRD1 scaffold fragment, not the full physiological complex.

    Evidence that argues against this (1)
    • An inhibitor-bound HAT-domain structure at 2.14 A is a strong starting point, and the caution for acetyltransferases is selectivity across the family rather than structure availability — KAT7 shares substrate chemistry with KAT6A/B, which this graph records as the pair whose joint inhibition overcomes menin resistance. RCSB PDB 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KAT7/HBO1 and members of its complex are required for leukaemia stem cell maintenance; the HAT domain deposits H3K14ac which sustains Hoxa9 and Hoxa10 expression, and a potent acetyl-CoA-competitive inhibitor was developed with efficacy across human AML lines and primary patient cells.

    HBO1 is required for the maintenance of leukaemia stem cells · MacPherson L, Anokye J, Yeung MM, et al. · Nature 2019
    PMID 31827282DOI
    What this evidence cannot show

    The discovery screen is an ex vivo murine leukaemia-stem-cell model and the genetic work is mouse; the human evidence is cell-line and primary-cell sensitivity, not KMT2A::AFF1-specific. No germline knockout phenotype is cited here.

    Evidence that argues against this (1)
    • KAT7's lineage-controlled selectivity of -0.242 is real but modest, and KAT7 has no documented physical interaction with KMT2A or any KMT2A fusion, so the dependency could be a general AML-stem-cell requirement rather than a fusion-specific one - in which case the therapeutic window against normal HSC may be small. source 2026
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    KAT7 is not a DepMap common-essential gene (Chronos mean -0.300 across 1,258 lines) but falls to -0.629 in KMT2A-rearranged lines, giving a lineage-controlled selectivity of -0.242 that ranks fifth in this graph's interactome.

    What this evidence cannot show

    Few KMT2A-rearranged lines; DepMap release version unverified.

    Evidence that argues against this (1)
    • A Chronos mean of -0.300 across 1,258 lines falling to -0.629 in KMT2A-rearranged lines is exactly the comparison this project has shown to be confounded: the KMT2A-rearranged panel is myeloid-enriched, so the difference mixes lineage with fusion status and must be lineage-split before it is used as a go/no-go input. Oncofusion knowledge graph (internal analysis) 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    KAT7's co-dependency with MEN1 retains 97% of its pooled value after within-tissue centring (0.214 to 0.208) and rises to 0.517 within leukaemia lines.

    What this evidence cannot show

    54 leukaemia lines; a correlation of 0.21 is small in absolute terms even when it is robust to lineage adjustment.

    Evidence that argues against this (1)
    • Retaining 97% of a co-dependency after within-tissue centring controls for TISSUE and not for LINEAGE WITHIN leukaemia, which is the axis that matters here. The centring is the right instinct applied one level too coarsely. Oncofusion knowledge graph (internal analysis) 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    KAT7 is expressed in all 54 GTEx v8 tissues, 6.75 TPM in whole blood, minimum 3.28 TPM in liver, maximum 34.1 TPM in testis.

    What this evidence cannot show

    Adult bulk tissue; no developmental or paediatric samples, which is the population where a developmental acetyltransferase matters most.

    Evidence that argues against this (1)
    • GTEx TPM is BULK tissue RNA. It gives neither protein abundance nor cell-type resolution within a tissue, so a low whole-blood figure is compatible with high expression in the progenitor compartment that matters here, and a high figure in a bulk tissue does not localise the risk. GTEx Portal 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    KAT7 is as constrained against loss of function as MEN1 (pLI 1.0, LOEUF 0.207 versus MEN1's 0.204).

    What this evidence cannot show

    As with MEN1, severe constraint did not forbid a tolerable drug; this number is a prompt to look for developmental toxicity, not a stopping rule.

    Evidence that argues against this (1)
    • A gnomAD constraint metric is not a safety signal for a drug, and this beat uses it as one across every dossier. pLI and LOEUF measure intolerance to HETEROZYGOUS loss-of-function across a whole lifetime from conception, in a population sample. A drug gives partial, reversible occupancy in an adult for weeks. High constraint says the gene matters in development; it does not predict on-target toxicity of transient pharmacological inhibition, and this project has already had to make the same correction for germline MEN1 penetrance in the endocrine beat. KAT7 matching MEN1 on constraint (LOEUF 0.207 vs 0.204) is presented as reassurance by analogy, which inherits whatever is wrong with using constraint for menin in the first place. gnomAD, Broad Institute 2026
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ChEMBL holds 98 activity rows against KAT7 (CHEMBL3774299) at pChEMBL 6 or above with a best reported Ki of 8 nM.

    What this evidence cannot show

    Cofactor-competitive biochemical potency does not predict cellular potency when intracellular acetyl-CoA is high, and no MYST-family counter-screen data accompanies these rows.

    Evidence that argues against this (1)
    • A ChEMBL activity-row count measures how much medicinal chemistry has been DONE, not how tractable the target is. The best reported affinity comes from the most optimised compound in whichever series was pursued, so it is a property of the investment rather than of the site, and comparing row counts across targets ranks attention rather than druggability. ChEMBL, EMBL-EBI 2026
Connected to