Oncofusion

start / graph / target-phase2-psip1-ibd-dossier

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

Phase-2 dossier: the LEDGF/PSIP1 IBD - the best therapeutic index in the interactome, a solved structure, two hotspot residues, and three compounds in the world

also called LEDGF IBD dossier · PSIP1 phase-2 plan · the other half of the clamp

LEDGF is the protein that anchors the fusion complex to the right place on DNA. Cells that carry the fusion need it badly while normal cells barely notice losing it, which is exactly the profile a good drug target should have - and almost nobody has made chemistry against it.

The same thing, for a scientist

PSIP1/LEDGF p75 is the second grip of the menin-LEDGF-KMT2A clamp. Its integrase-binding domain (IBD, ~347-435) binds a KMT2A segment at 140-160 through two aromatic hotspots, F148 and F151, and the interaction is required for MLL-AF9 transformation in colony assays. Its dependency profile is the most favourable of any partner in this graph: DepMap Chronos mean +0.028 across 1,258 lines (i.e. neutral) falling to -0.444 in KMT2A-rearranged lines. ChEMBL holds three potent rows against it, all from HIV-1 integrase antiviral chemistry, best IC50 470 nM. There is no oncology programme. The combination of a solved interface, named hotspots, a validated genetic requirement, a near-neutral essentiality profile and essentially zero competition is the strongest phase-2 case in this beat.

Go deeper
What structure exists, and what does not

Exists, verified against RCSB this cycle: 2MSR, solution structure of the LEDGF/p75 IBD in complex with the MLL1 140-160 peptide, NMR 1; 3U88, human menin with MLL1 and LEDGF, X-ray 3.0 A 1. The PWWP-nucleosome contact is separately solved by cryo-EM 3.

Does not exist: a structure of the IBD with a small molecule bound; a structure of the full LEDGF protein in a productive complex. The AlphaFold model of PSIP1 is largely disordered and is not a usable docking template outside the IBD and PWWP modules.

What interface would be engaged, and what makes it tractable

The IBD-KMT2A contact is primarily sustained by two aromatic residues of KMT2A, F148 and F151, inserting into the IBD 2. Two-hotspot aromatic interfaces are the most tractable class of protein-protein interaction for small molecules and macrocycles, because a ligand only has to reproduce two anchor points rather than distribute affinity across a flat surface. The same IBD surface is the binding site for HIV-1 integrase, PogZ and JPO2, and binding is mutually exclusive 2 - which is why the only chemistry that exists against it came from antiviral work.

The proof that blocking this interface does something: a cyclic peptide (CP65) developed against the LEDGF-integrase interaction selectively impaired clonogenic growth of primary murine MLL-AF9 blasts 2. That is a tool-compound-level demonstration, in mouse cells, with a peptide - which is exactly the level of evidence that a phase-2 programme exists to convert or kill.

Essentiality and the normal-tissue picture: why this is the best index here

DepMap Chronos mean across 1,258 lines is +0.028 - knocking PSIP1 out does nothing to the average cancer cell. In KMT2A-rearranged lines it falls to -0.444, and in KMT2A::AFF1 lines to -0.346, a selectivity of -0.472 against all lines 4. In this project's own lineage-conditioned analysis the PSIP1-MEN1 co-dependency retained 97% of its pooled correlation after within-tissue centring (0.529 to 0.515) and rose to 0.686 within leukaemia lines 5, so the signal is not a lineage artefact.

GTEx v8: expressed in all 54 tissues, 6.71 TPM in whole blood, minimum 4.53 TPM in liver, maximum 126.0 TPM in EBV-transformed lymphocytes 6. gnomAD pLI 0.9999984, LOEUF 0.373 - constrained, but measurably less so than MEN1 (0.204) or KAT7 (0.207) 7.

Chemical matter: the gap that is the opportunity

ChEMBL target CHEMBL3988590 holds three activity rows at pChEMBL 6 or above, best IC50 470 nM (CHEMBL3981070) 8. All are LEDGIN-class antiviral compounds assaying the LEDGF-integrase interface. There is no oncology series, no co-crystal with a small molecule, and no clinical candidate.

This is the definition of open ground, and it is also the definition of risk: three compounds in a public database usually means either nobody looked, or several groups looked and failed without publishing. The phase-2 programme must be designed so that the second possibility is discovered in month three, not year three.

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

One thing, and it is testable within a cycle: blocking the IBD-KMT2A contact must be sufficient to collapse the fusion output programme in a human KMT2A::AFF1 cell. Everything published is either genetic (mutant LEDGF rescue) or murine (CP65 in MLL-AF9 blasts). If a compound or peptide that verifiably displaces the KMT2A 140-160 peptide in vitro does not move HOXA9 in SEM or MV4-11, the whole target is an interface that can be blocked without consequence, and the programme stops.

Correction: the ligand is bipartite and the affinity is known

This dossier described the pocket as accepting a single KMT2A motif at 140-160 with two aromatic hotspots. PMID 25305204 shows the ligand is bipartite - IBM1 (124-133, F129) and IBM2 (147-152, F148/L149/F151) - and both are required for high-affinity binding.

For a druggability dossier that changes the problem. A competitor has to occupy, or allosterically defeat, two binding events rather than one - and the paper's own short IBM2 peptide inhibits both the IBD-MLL/menin and IBD-integrase interactions, so a single-motif competitor is a demonstrated starting point rather than a hypothetical one.

And the selectivity liability is now quantified. The IBM2 site on IBD overlaps the HIV-1 integrase binding site, and integrase efficiently sequesters the IBD from the menin-MLL complex. Anything that binds this pocket is, by construction, also an HIV integrase-LEDGF inhibitor - which is an opportunity in one field and an off-target in the other.

Facts and figures
target genegene-psip1
uniprotO75475
pocketIBD (approx. 347-435) surface that accepts a BIPARTITE KMT2A ligand: IBM1 at 124-133 (F129) and IBM2 at 147-152 (F148/L149/F151). Kd 14.7 uM binary, 1.4 uM against the menin-MLL complex (PMID 25305204). The entry previously described a single motif at 140-160, which is IBM2 alone.
phase2 verdictSTART FIRST. Highest expected information per dollar of any target in this beat: the essentiality profile is already the best available, the interface is solved, the hotspots are named, and the decisive experiment is a peptide-displacement assay plus one qPCR readout.
compute backendBoltz credits for IBD-peptide and IBD-macrocycle co-folds; Explorer Slurm GPU only if a de novo binder design campaign is approved after G2.
interface nodes
[
  "interface-psip1-ibd-kmt2a",
  "interface-psip1-pwwp-h3k36me-nucleosome"
]
structures verified this cycle
[
  {
    "pdb": "2MSR",
    "title": "Solution structure of LEDGF/p75 IBD in complex with MLL1 peptide (140-160)",
    "method": "NMR"
  },
  {
    "pdb": "3U88",
    "title": "Crystal structure of human menin in complex with MLL1 and LEDGF",
    "method": "X-ray",
    "resolution_A": 3
  }
]
structures absent
[
  "no small-molecule co-crystal with the IBD",
  "no structure of full-length LEDGF in a productive complex"
]
essentiality
{
  "chronos_mean_all_lines": 0.028,
  "chronos_mean_kmt2ar": -0.444,
  "chronos_mean_kmt2a_aff1": -0.346,
  "selectivity_vs_all_lines": -0.472,
  "depmap_common_essential": false,
  "lineage_conditioned_r_vs_MEN1": 0.515,
  "leukemia_only_r_vs_MEN1": 0.686
}
normal tissue gtex v8 tpm
{
  "whole_blood": 6.713,
  "min": {
    "tissue": "Liver",
    "tpm": 4.527
  },
  "max": {
    "tissue": "Cells_EBV-transformed_lymphocytes",
    "tpm": 125.996
  },
  "n_tissues": 54
}
gnomad
{
  "pLI": 0.9999984,
  "loeuf": 0.373
}
chemical matter
{
  "chembl_target": "CHEMBL3988590",
  "rows_pchembl_ge_6": 3,
  "best": {
    "molecule": "CHEMBL3981070",
    "type": "IC50",
    "value_nM": 470,
    "pchembl": 6.33
  },
  "note": "all rows are LEDGIN-class antiviral chemistry against the LEDGF-integrase interface; no oncology series exists"
}
go no go
{
  "G1_feasibility": {
    "question": "Can we assay IBD-KMT2A displacement?",
    "GO": "Fluorescence-polarisation displacement of a labelled KMT2A 140-160 peptide from recombinant IBD, Z-prime above 0.6, with F148A/F151A peptide as negative control",
    "STOP": "The labelled peptide does not bind recombinant IBD with a measurable Kd"
  },
  "G2_biology": {
    "question": "Is blocking the IBD sufficient to collapse the fusion programme in a human KMT2A::AFF1 cell?",
    "GO": "A displacer with biophysical IC50 below 1 uM reduces HOXA9 mRNA by at least 50% at 72 h at a concentration below the 72 h GI20 in SEM or MV4-11",
    "STOP": "HOXA9 reduction below 20% at any non-cytotoxic concentration, with confirmed target engagement",
    "ambiguous_zone": "20-50% HOXA9 reduction",
    "tie_breaker": "CRISPR knock-in of KMT2A F148A/F151A into SEM; if the genetic version also gives under 50%, the interface is not sufficient and the programme stops"
  },
  "G3_window": {
    "question": "Therapeutic index",
    "GO": "Less than 30% loss of healthy-donor CD34+ colony formation at the G2-effective concentration - expected to pass, given the +0.028 Chronos mean",
    "STOP": "More than 60% CFU loss, which would mean the favourable DepMap profile does not transfer to primary cells"
  }
}
How this could bite us back
  • LEDGF is the cellular cofactor HIV-1 integrase uses to target integration. A systemic LEDGF-IBD blocker is, by construction, an antiviral-adjacent agent; in a child receiving one it could alter the behaviour of any integrating virus or integrating gene-therapy vector, which matters in a field that also uses lentiviral CAR-T.
  • Multivalent chromatin anchoring means avidity may defeat a single-site inhibitor even at full IBD occupancy.
  • The IBD surface is shared with PogZ and JPO2; a displacer will release those too, and the consequences of that are entirely unstudied.
  • The favourable DepMap profile comes from immortalised lines. Primary infant progenitors may depend on LEDGF more than a cancer cell line does, and GTEx contains no infant tissue at all.
Still open
  1. Does the multivalency of LEDGF chromatin binding mean that blocking the IBD alone leaves the complex tethered through PWWP and the disordered linker, so that only a dual PWWP+IBD blockade releases it?
  2. Why has no oncology chemistry appeared against this interface in the twelve years since the hotspots were published - was it tried and abandoned, and by whom?
  3. Is the LEDGF requirement the same for KMT2A::AFF1 as for KMT2A::MLLT3, given that the published functional work is almost entirely MLL-AF9?
Every source behind this entry (8)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The LEDGF IBD-KMT2A peptide complex (2MSR, NMR) and the menin-MLL1-LEDGF ternary complex (3U88, X-ray 3.0 A) were verified by direct query of the RCSB data API this cycle.

    What this evidence cannot show

    2MSR is an NMR structure of an isolated domain with a 21-residue peptide; it has no resolution value and does not report the interface in a chromatin context.

    Evidence that argues against this (1)
    • The IBD-KMT2A complex is NMR and the ternary complex is 3.0 A, so neither is at a resolution that supports structure-based design of a small molecule against a protein-protein interface. The interface has nonetheless been validated as a target in published work, which is the stronger argument and is not the one the entry makes. Cancer Research 2014
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The LEDGF IBD-MLL interaction is sustained primarily by two aromatic MLL residues, F148 and F151, is mutually exclusive with HIV-1 integrase, PogZ and JPO2 binding, is essential for transformation in colony-forming assays in MLL-AF9-positive cells, and is druggable at tool level by the cyclic peptide CP65.

    PMID 25082813DOI
    What this evidence cannot show

    The functional work is in murine MLL-AF9 cells, not human KMT2A::AFF1; CP65 is a cyclic peptide repurposed from antiviral work, with no stated cell-permeability or exposure data suitable for a therapeutic claim. Twelve years on, no follow-on oncology chemistry has appeared, which is itself a warning.

    Evidence that argues against this (1)
    • The absence of any oncology chemical series against this interface in ChEMBL after a decade is weak evidence that the interface is harder than it looks. source 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    LEDGF reads H3K36me2 and H3K36me3 nucleosomes through its PWWP domain with an additional contact at the acidic-patch/DNA interface, and its disordered linker contributes DNA-binding motifs - so the protein is a multivalent chromatin anchor, not a single-contact adaptor.

    Multivalency of nucleosome recognition by LEDGF · Koutna E, Lux V, Kouba T, et al. · Nucleic Acids Research 2023
    PMID 37615563DOI
    What this evidence cannot show

    Structural and biophysical; the relative contribution of each contact to fusion-complex tethering in a leukaemia cell is not measured.

    Evidence that argues against this (1)
    • Multivalency is bad news for a single-site inhibitor: if PWWP-nucleosome and linker-DNA contacts contribute substantial avidity, blocking the IBD alone may not release the complex from chromatin. source 2023
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    PSIP1 has a near-neutral DepMap Chronos mean of +0.028 across 1,258 lines but falls to -0.444 in KMT2A-rearranged lines, the most favourable dependency-versus-essentiality profile of any partner examined in this graph.

    What this evidence cannot show

    The number of KMT2A-rearranged lines in DepMap is small (single digits for KMT2A::AFF1 specifically), so the rearranged-line mean is fragile. DepMap release version is not stated by the API.

    Evidence that argues against this (1)
    • A near-neutral pan-cancer Chronos mean falling to -0.444 in KMT2A-rearranged lines is the selectivity claim, and it needs the lineage split before it can be used. PSIP1 does survive that split elsewhere in this graph, so the claim is likely right and is not established by the number quoted here. Oncofusion knowledge graph (internal analysis) 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In this project's executed lineage-conditioned analysis, the PSIP1-MEN1 co-dependency retained 97% of its pooled correlation after within-tissue centring (0.529 pooled to 0.515 within-tissue) and rose to 0.686 within leukaemia lines.

    What this evidence cannot show

    54 leukaemia lines; correlation of gene-effect profiles is not evidence of physical interaction, only of shared requirement.

    Evidence that argues against this (1)
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PSIP1 is expressed in all 54 GTEx v8 tissues, median 6.71 TPM in whole blood, minimum 4.53 TPM in liver, maximum 126.0 TPM in EBV-transformed lymphocytes.

    What this evidence cannot show

    Adult bulk tissue only; the EBV-transformed lymphocyte value is a cell line, not a tissue, and should not be read as a normal-blood number.

    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. LEDGF is also the HIV integration cofactor and is broadly required, so its normal-tissue profile is the expected one for a housekeeping chromatin adaptor. GTEx Portal 2026
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PSIP1 is constrained against loss of function (pLI 0.9999984, LOEUF 0.373) but measurably less so than MEN1 (LOEUF 0.204) or KAT7 (0.207).

    What this evidence cannot show

    LOEUF differences of this size between genes are within the noise of the estimator for genes of similar length; this should be read as 'not less constrained than the others by much', not as a safety claim.

    Evidence that argues against this (1)
    • A gnomAD constraint metric measures intolerance to heterozygous loss over a lifetime from conception, not tolerance of weeks of partial pharmacological inhibition in an adult, so ranking PSIP1 against MEN1 and KAT7 on LOEUF compares three germline developmental quantities and infers a safety ordering none of them measures. gnomAD, Broad Institute 2026
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ChEMBL holds only three activity rows against PSIP1 (CHEMBL3988590) at pChEMBL 6 or above, best IC50 470 nM, and all of them come from HIV-1 integrase antiviral chemistry rather than an oncology programme.

    What this evidence cannot show

    ChEMBL under-represents unpublished industry chemistry; an absence here is an absence in the public record only.

    Evidence that argues against this (1)
    • Three rows from HIV-1 integrase antiviral chemistry is not a tractability measurement: it records that the only campaign run against this protein targeted a different interface for a different disease. Absence of oncology chemistry means nobody tried, and the IBD interface has separately been validated as a target in published work, which is the stronger argument and is not the one this item makes. Cancer Research 2014
Connected to