Oncofusion

start / graph / target-phase2-enl-yeats-dossier

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

Phase-2 dossier: the ENL YEATS pocket - two co-crystals, nanomolar chemistry, and a co-dependency that only appears inside leukaemia

also called MLLT1 YEATS dossier · ENL phase-2 plan · SGC-iMLLT readiness

ENL is a reader protein that recognises a chemical tag on histones and helps keep the fusion programme switched on. Unusually for this field it already has a published inhibitor with a solved structure, and a degrader exists too - so the question is not whether we can block it but whether blocking it matters.

The same thing, for a scientist

MLLT1/ENL reads acylated histone H3 through its YEATS domain. Two co-crystals were verified this cycle: 5J9S (ENL YEATS with H3K27ac, X-ray 2.70 A) and 6HT1 (ENL YEATS with the chemical-probe series compound SGC-iMLLT, X-ray 2.10 A). ChEMBL holds 144 activity rows at pChEMBL 6 or above with a best reported Ki of 1.43 nM. The dependency case is more equivocal: ENL's co-dependency with MEN1 is weak when pooled across all lineages (r 0.130) and retains only 72% of that on within-tissue centring, but rises to 0.426 within leukaemia lines - a target whose relevance is lineage-conditional. This is the archetype for a phase-2 question that is answerable cheaply and quickly, because both the probe and the degrader already exist.

Go deeper
What structure exists, and what does not

Exists, verified against RCSB this cycle: 5J9S, ENL YEATS in complex with histone H3 acetylated at K27, X-ray 2.70 A; 6HT1, MLLT1 (ENL) YEATS domain with SGC-iMLLT (compound 92), X-ray 2.10 A 1. The 6HT1 structure is the important one - it is an inhibitor-bound reader pocket at good resolution, which is the single most useful asset a structure-based campaign can start from.

Does not exist: a structure of ENL in the context of the super elongation complex, or bound to a KMT2A fusion protein. All design work is on the isolated YEATS domain.

What pocket is engaged

The YEATS aromatic sandwich that accommodates the acylated lysine side chain. Reader pockets of this class are shallow but genuinely ligandable, and the SGC-iMLLT series proves it. The relevant selectivity liability is the YEATS family itself: AF9 (MLLT3), GAS41 (YEATS4) and YEATS2 share the fold, and AF9 in particular is the paralogue this graph's biology entries single out as not a dependency while ENL is. A selectivity gate against MLLT3 is therefore mandatory, not optional.

Essentiality and the normal-tissue picture

In this project's lineage-conditioned analysis, ENL's correlation with MEN1 is 0.130 pooled, 0.094 within-tissue (72% retained) and 0.426 within leukaemia lines 2. Read plainly: ENL is not a general partner of the menin module, but it becomes one inside blood cancer. That is an attractive shape for selectivity and a fragile one for statistics, since it rests on 54 leukaemia lines.

GTEx v8: broadly and highly expressed - 40.3 TPM in whole blood, minimum 18.3 TPM in heart left ventricle, maximum 141.7 TPM in cerebellum, all 54 tissues 3. ENL is the most abundantly expressed candidate in this beat, which is a poor sign for a therapeutic window. gnomAD pLI 0.9999984, LOEUF 0.330 4.

Chemical matter

ChEMBL target CHEMBL4295801 (Protein ENL): 144 activity rows at pChEMBL 6 or above, best reported Ki 1.43 nM (CHEMBL4210431) 5. There is also a bifunctional degrader reported in the literature for ENL, which is recorded in gene-mllt1. For a phase-2 plan this matters more than the inhibitor: reader domains frequently tolerate ligand occupancy without losing function, because the protein's job is scaffolding as much as reading, and only degradation removes the scaffold.

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

Two things, in order. First, degradation must beat occupancy: if the SGC-iMLLT-class probe does not suppress HOXA9 but the degrader does, the programme is a degrader programme and the inhibitor chemistry is a distraction. Second, the window must be real: at 40 TPM in whole blood, ENL is abundant in exactly the tissue we cannot afford to damage, so the CD34+ colony assay is the gate that decides this target, not the potency assay.

Facts and figures
target genegene-mllt1
uniprotQ03111
pocketYEATS aromatic sandwich that reads acylated histone H3 lysine
phase2 verdictSTART, third in order. Cheap and fast because the reagents exist, but the therapeutic window is the real uncertainty and the abundance in blood is a genuine warning.
compute backendBoltz co-folding for YEATS-family selectivity profiling (ENL vs AF9 vs GAS41) at low credit cost; no cluster time required.
essentiality
within_tissue_r_vs_MEN10.094within_tissue_r_vs_MEN1: 0.0936pooled_r_vs_MEN10.130pooled_r_vs_MEN1: 0.1295leukemia_only_r_vs_MEN10.426leukemia_only_r_vs_MEN1: 0.4257fraction_retained0.723fraction_retained: 0.723
paralogue selectivity liability
[
  "MLLT3 (AF9)",
  "YEATS4 (GAS41)",
  "YEATS2"
]
structures verified this cycle
[
  {
    "pdb": "5J9S",
    "title": "ENL YEATS in complex with histone H3 acetylation at K27",
    "method": "X-ray",
    "resolution_A": 2.702
  },
  {
    "pdb": "6HT1",
    "title": "MLLT1 (ENL) YEATS domain in complex with SGC-iMLLT (compound 92)",
    "method": "X-ray",
    "resolution_A": 2.1
  }
]
essentiality — the numbers behind the chart
{
  "pooled_r_vs_MEN1": 0.1295,
  "within_tissue_r_vs_MEN1": 0.0936,
  "leukemia_only_r_vs_MEN1": 0.4257,
  "fraction_retained": 0.723
}
normal tissue gtex v8 tpm
{
  "whole_blood": 40.304,
  "min": {
    "tissue": "Heart_Left_Ventricle",
    "tpm": 18.273
  },
  "max": {
    "tissue": "Brain_Cerebellum",
    "tpm": 141.675
  },
  "n_tissues": 54
}
gnomad
{
  "pLI": 0.9999984,
  "loeuf": 0.33
}
chemical matter
{
  "chembl_target": "CHEMBL4295801",
  "rows_pchembl_ge_6": 144,
  "best": {
    "molecule": "CHEMBL4210431",
    "type": "Ki",
    "value_nM": 1.43,
    "pchembl": 8.85
  }
}
go no go
{
  "G1_feasibility": {
    "question": "Can occupancy and degradation be compared head to head?",
    "GO": "A YEATS probe and an ENL degrader both available and both showing target engagement in a KMT2A::AFF1 line by cellular thermal shift or western",
    "STOP": "Neither reagent can be obtained or synthesised within the cycle"
  },
  "G2_biology": {
    "question": "Does removing ENL collapse the fusion output programme?",
    "GO": "ENL degradation reduces HOXA9 mRNA by at least 50% at 72 h, with less than 20% reduction in a non-KMT2A-rearranged control line",
    "STOP": "Below 20% HOXA9 reduction at more than 90% protein loss - which would mean ENL is dispensable to the programme",
    "ambiguous_zone": "20-50%",
    "tie_breaker": "MLLT1 CRISPR knockout in SEM with MLLT3 knockout as the paralogue comparator; binding"
  },
  "G3_window": {
    "question": "Can normal haematopoiesis survive ENL loss?",
    "GO": "Less than 50% loss of healthy-donor CD34+ colony formation at the G2-effective concentration",
    "STOP": "More than 50% CFU loss - a plausible outcome given 40 TPM expression in whole blood, and the reason this gate, not potency, decides the target"
  }
}
How this could bite us back
  • ENL is the most abundantly expressed candidate in this beat in whole blood (40 TPM). Systemic ENL removal is most likely to injure the tissue we are trying to rescue.
  • YEATS domains are a small paralogous family; a non-selective ligand hits AF9, GAS41 and YEATS2, whose combined loss is unstudied.
  • Reader domains commonly tolerate occupancy; a potent probe that does nothing in cells would consume a cycle and produce an ambiguous rather than a negative result, which is why the degrader comparison is placed at G1 rather than later.
Still open
  1. Does YEATS-pocket occupancy phenocopy ENL loss, or does ENL's scaffolding role mean only degradation works?
  2. Is the leukaemia-only co-dependency with MEN1 real, or an artefact of computing a subgroup correlation on 54 lines after the pooled correlation came out near zero?
  3. Can a ligand distinguish ENL from AF9 given that the graph's own biology entries say AF9 is not a dependency while ENL is - and if not, does hitting both matter?
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Two ENL YEATS co-crystals were verified by direct query of the RCSB data API this cycle: 5J9S (ENL YEATS with H3K27ac, X-ray 2.70 A) and 6HT1 (MLLT1 ENL YEATS with SGC-iMLLT, X-ray 2.10 A).

    What this evidence cannot show

    Both are isolated YEATS domains; neither reports the domain in the context of the super elongation complex or a fusion protein.

    Evidence that argues against this (1)
    • YEATS co-crystals with acylated histone peptides establish the reader pocket. The therapeutic question for ENL is different: this graph records that ENL PERSISTS at oncoprotein target loci after the oncoprotein and menin are undetectable, so occupying the reader pocket may not remove what ENL is doing there. RCSB PDB 2026
  2. E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset

    ENL's co-dependency with MEN1 is weak across all lineages (pooled r 0.130, within-tissue 0.094) but substantially stronger within leukaemia lines (r 0.426), making it a lineage-conditional rather than a general member of the menin module.

    What this evidence cannot show

    A leukaemia-only correlation computed on 54 lines with a pooled correlation near zero is exactly the pattern that can arise by chance; this number should not be used as a gate without replication in an independent line panel.

    Evidence that argues against this (1)
    • The same analysis showed that a pooled correlation can invert on lineage adjustment (MYB: 0.105 pooled to -0.052 within-tissue). A leukaemia-only enrichment computed post hoc deserves the same scepticism. source 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    MLLT1 is highly and broadly expressed in normal human tissue - 40.3 TPM in whole blood, 18.3 TPM minimum in heart left ventricle, 141.7 TPM maximum in cerebellum, across all 54 GTEx v8 tissues.

    What this evidence cannot show

    Bulk adult tissue; high transcript abundance does not by itself predict toxicity, since protein turnover and functional reserve are not measured.

    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. MLLT1 at 40.3 TPM in whole blood with a floor of 18.3 across tissues is a broadly expressed target, which is the relevant fact and is being reported alongside constraint as if the two were independent lines of evidence. GTEx Portal 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    MLLT1 is constrained against loss of function in humans (pLI 0.9999984, LOEUF 0.330).

    What this evidence cannot show

    Germline lifetime constraint; a weak predictor of tolerability for a reversible, time-limited inhibitor.

    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. gnomAD, Broad Institute 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ChEMBL holds 144 activity rows against ENL (CHEMBL4295801) at pChEMBL 6 or above with a best reported Ki of 1.43 nM, so this target has real, published chemical matter.

    What this evidence cannot show

    Reader-domain ligands frequently show excellent biochemical affinity and poor cellular consequence; potency here is not evidence of function.

    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. A best Ki of 1.43 nM against ENL is a real chemical achievement and says nothing about whether reader-pocket occupancy produces the phenotype. ChEMBL, EMBL-EBI 2026
Connected to