Oncofusion

start / graph / gene-paf1

Proteinstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q5c-interactome2

PAF1 — pan-essential subunit of a complex the fusion needs, where the window is in the interface and not in the protein

also called PAF1 · PD2 · RNA polymerase II-associated factor 1 homolog · PAF1C · PAFc

PAF1 is part of a machine that travels with RNA polymerase, and the KMT2A fusion grabs onto it using a piece of KMT2A that the break always keeps. Every cell needs PAF1, so you cannot block the protein — but you may be able to block only the handshake between it and the fusion, which is the same logic that makes menin a target.

The same thing, for a scientist

PAF1 (Q8N7H5, 531 aa) is a core subunit of the RNA polymerase II-associated factor complex (PAF1C: PAF1, CTR9, LEO1, CDC73, WDR61). PAF1C binds MLL through sequences flanking the CxxC domain, a region retained in every KMT2A fusion, and synergises with MLL fusion proteins at HOX loci. It is DepMap common-essential (mean Chronos -0.83, depleted in 93.3% of lines), so the protein itself has no therapeutic window. The important result is that disrupting the MLL-PAFc *interface* selectively inhibits MLL-fusion AML while sparing normal HSC function, which makes PAF1C the second published example of this project's central interface thesis.

What it looks like
PAF1531 aa
Coiled coilCoiled coil · residues 352–4000200400
loading AF-Q8N7H5-F1…
AlphaFold AF-Q8N7H5-F1 ↗
colourshape

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

very high (90+)confident (70–90)low (50–70)very low (<50)
I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
Why this entry matters more than its druggability suggests

PAF1C binds MLL directly, through sequences flanking the CxxC domain — and that is exactly the part of KMT2A that survives every breakpoint. Muntean and colleagues showed PAFc synergises with MLL fusion proteins at HOX loci to promote leukemogenesis 1; Milne and colleagues independently mapped multiple recruitment interactions at the HOXA9 locus in the same issue-year 2.

Then came the result that should matter most to this project. PAFc is required indiscriminately for AML proliferation — it is not an MLL-selective dependency. But a dominant-negative MLL fragment that blocks the MLL–PAFc interface selectively kills MLL-rearranged cells while leaving normal HSC function intact in reconstitution assays 3.

That is the menin argument, reproduced at a second interface, by a different group, in a different complex. It is independent support for the graph's central hypothesis that the therapeutic window in this disease lives in specific protein–protein interfaces rather than in the proteins themselves. It is also a reason to look harder at the MLL–PAFc interface as a target in its own right.

The protein has no window at all

PAF1 is DepMap common-essential: mean Chronos -0.832 across 1,258 lines, -0.822 across leukemia lines, depleted in 93.3% of all lines 4. Leukemia-restricted correlation with MEN1 is -0.095, confidence interval crossing zero 5 — the expected flat-profile result for a pan-essential gene, and confirmation that a co-dependency scan cannot see genes like this.

Human constraint tells the opposite story and the mismatch is instructive: pLI 0.011, LOEUF 0.611, 31 observed against 68.4 expected loss-of-function variants 4. A gene can be required by every cell and still tolerate heterozygous loss in a person; cellular essentiality and haploinsufficiency are different measurements. Any triage rule that treats LOEUF as a general tolerance proxy would pass PAF1 as GREEN on constraint while DepMap flags it RED, and the DepMap call is the right one here.

Expression is uniform and unhelpful for selectivity: bone marrow 34.8, liver 44.7, testis 61.4 nTPM, low tissue specificity 4.

PAF1C is not a one-way activator

The simple picture of PAFc as an activator that the fusion recruits is incomplete. PAF1 complex interactions with SETDB1 mediate promoter H3K9 methylation and transcriptional repression, including of HOXA9 and MEIS1 6. The same complex can therefore push the output programme in either direction depending on partner.

That is a warning about interface-targeting strategies generally. An agent that disrupts MLL–PAFc removes both the activating and the repressive arm at once, and the net effect on HOXA9 is not predictable from the activating arm alone.

Facts and figures
uniprotQ8N7H5
gene symbolPAF1
ensemblENSG00000006712
length AA531
alphafold IDAF-Q8N7H5-F1
PDB notePAF1 is resolved only within Pol II elongation-complex cryo-EM structures, not alone; 6TED (Pol II-DSIF-PAF-SPT6, 3.1 A), 7OPC (Pol II-CSB-CRL4CSA-UVSSA-SPT6-PAF, 3.0 A), 8A3Y (Pol II-DSIF-SPT6-PAF1-TFIIS-hexasome, 3.3 A).
complexPAF1C (PAF1, CTR9, LEO1, CDC73, WDR61)
rolePAF1C binds MLL via CxxC-flanking sequences retained in all KMT2A fusions; pan-essential protein, selective interface
druggabilityThe protein: none, pan-essential. The MLL-PAFc interface: unproven but genetically validated as selective and HSC-sparing, and currently unaddressed by any small molecule. No ChEMBL count quoted; an interface-specific count could not be determined.
subnuclear localizationCONTESTED - an uncited UniProt note places PAF1 in nucleoplasmic puncta and explicitly excludes nucleoli, while a 2024 nucleolus-restricted proximity map recovers PAF1 at intergenic rDNA and shows PAF1 loss disrupts nucleolar organisation
shown for fusionNO for localisation - PAF1 has never been imaged in a KMT2A-rearranged cell, although MLL-AF4 cooperates with PAF1 and FACT at high-density enhancer interactions (PMID 37626123)
PDB verified
[
  "6TED",
  "7OPC",
  "8A3Y"
]
domains
[
  {
    "name": "Coiled coil",
    "start": 352,
    "end": 400,
    "pfam": null
  }
]
essentiality
{
  "depmap_common_essential": true,
  "chronos_mean": -0.8316,
  "chronos_mean_leukemia_lines": -0.8221,
  "n_lines": 1258,
  "fraction_lines_depleted": 0.9329,
  "gnomad_pLI": 0.0106,
  "gnomad_loeuf": 0.611,
  "gnomad_obs_lof": 31,
  "gnomad_exp_lof": 68.4,
  "mouse_ko": "Not established in this entry; no Paf1 knockout citation was verified and none should be asserted. The genetically tested PAF1C subunit in this disease is CDC73.",
  "hpa_ntpm": {
    "testis": 61.4,
    "liver": 44.7,
    "spleen": 42.6,
    "heart_muscle": 40.4,
    "naive_B_cell": 36.7,
    "bone_marrow": 34.8,
    "thymus": 28.9,
    "lymph_node": 27.4
  },
  "hpa_specificity": "Low tissue specificity"
}
codependency men1
{
  "r_leukemia_lines": -0.0953,
  "ci_leukemia": [
    -0.318,
    0.1491
  ],
  "crosses_zero": true,
  "p_permutation": 0.5027
}
localization
{
  "subcellular": "nucleus (UniProt Q8N7H5, ECO:0000269 for PMID 16491129 and 30367041). NOTE the identifier trap: gene_exact:PAF1 returns two reviewed human records, Q8N7H5 (PAF1, 531 aa) and P28328 (PEX2, 305 aa, peroxisomal, carrying PAF1 as a synonym).",
  "subnuclear": "CONTESTED. UniProt carries an UNCITED note - 'punctuate distribution throughout the nucleus except in nucleoli and the perinuclear chromatin' - while compartment-enriched proximity labelling of the nucleolar Pol II interactome recovers PAF1 at intergenic rDNA, and PAF1 deficiency disrupts nucleolar organisation and rRNA biogenesis, rescued by repressing unscheduled R-loops (PMID 39505901). Possible reconciliation: bulk distribution versus a transient functional minority pool. No source reconciles them.",
  "how_shown": "immunofluorescence of unknown stringency (PMID 16491129, full text not read); compBioID with nucleolar Pol II plus an R-loop repression rescue (PMID 39505901)",
  "systems": "pancreatic cancer cells and NIH 3T3 for the older work; human cell lines for the nucleolar study; nothing haematopoietic or leukaemic",
  "contradiction": "nucleolar exclusion versus a functional nucleolar pool; resolving experiment is endogenous PAF1 co-stained with fibrillarin and nucleophosmin at exposure adequate for a minority pool",
  "node": "localization-paf1-nucleoplasmic-puncta-versus-a-nucleolar-pool"
}
cell type
{
  "evidence_class_key": "concept-expressed-required-maintains-three-evidence-classes",
  "identifier_note": "PAF1 returns TWO reviewed human records. The one meant here is Q8N7H5, PAF1, 531 aa; P28328 is PEX2 (305 aa) and carries PAF1 only as a synonym. HGNC fetched 2026-09-20: PAF1, HGNC ensembl ENSG00000006712, UniProt Q8N7H5.",
  "expressed_in": {
    "class": 1,
    "summary": "ubiquitous"
  },
  "required_in": {
    "class": 2,
    "floor_compressed": true,
    "pct_all_lines_depleted": 93.4,
    "all_1258_lines_mean_gene_effect": -0.83,
    "kmt2a_rearranged_lines_mean_gene_effect": -0.818,
    "myeloid_kmt2ar_mean": -0.762,
    "blineage_kmt2ar_mean": -0.916,
    "verdict": "PAN-ESSENTIAL. Depleted below -0.5 in 93.4% of 1,258 lines and the KMT2A-rearranged arm is +0.012 relative to all lines, i.e. indistinguishable. DepMap carries no cell-type information for PAF1.",
    "the_cell-type_evidence_that_does_exist": "the PAF1c subunit CDC73 is required for mouse haematopoietic stem cell MAINTENANCE while displaying leukaemia-SPECIFIC gene regulation (PMID 31031188) - a complex-level result that separates the normal-cell requirement from the leukaemic one, and the only such separation available for this complex",
    "rule_64_note": "PAF1's cell-state literature is reachable only through the COMPLEX name and through a partner subunit; a PAF1-symbol-only enumeration would have found two records and missed the CDC73 result entirely"
  },
  "context": "BOTH - normal murine HSC (via CDC73) and AML (via the KDM4A-PAF1 axis)",
  "graph_links": [
    "cell-type-hsc-bone-marrow"
  ]
}
cell state
{
  "maintains": {
    "class": 3,
    "states": [
      "cell-state-stem-like-self-renewal"
    ],
    "claim": "A KDM4A-PAF1-mediated epigenomic network is essential for acute myeloid leukaemia cell SELF-RENEWAL and survival.",
    "pmids": [
      "34083515",
      "31031188"
    ]
  },
  "bounded_query": {
    "date": "2026-09-20",
    "tool": "NCBI esearch db=pubmed",
    "query": "(PAF1[tiab] OR \"PAF1 complex\"[tiab] OR \"PAF complex\"[tiab] OR \"Paf1C\"[tiab]) AND <cell-state clause> AND <leukaemia clause>",
    "count": 2,
    "on_topic_after_title_check": 2,
    "controls": {
      "gene_clause_alone": 536,
      "state_clause_alone": 96743
    },
    "reading": "a small but fully on-topic literature - both records are real cell-state results. This is the opposite failure mode from FLT3's 617."
  }
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.8182,
  "chronos_mean_other_leukaemia": -0.7802,
  "difference": -0.0381,
  "pan_essentiality_all_lines": -0.8316,
  "selectivity_rank": 4529,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • Essential-function toxicity: PAF1 is required by 93% of all cell lines. Any agent that reduces PAF1 protein rather than one specific interface will be broadly cytotoxic.
  • Compensation in the wrong direction: PAFc also represses HOXA9 and MEIS1 through SETDB1, so removing PAFc from the fusion could in principle relieve repression as well as activation.
  • Delivery: no chemical matter exists for the MLL-PAFc interface, and the validating reagent is a dominant-negative protein fragment, which is a long way from a paediatric medicine.
  • Triage failure mode: PAF1 would pass a LOEUF-based tolerance filter (0.611, pLI 0.011) while being pan-essential in cells. Constraint and essentiality must both be checked, and neither alone is sufficient.
Still open
  1. Is the MLL-PAFc interface tractable to a small molecule? It is the strongest unexploited interface in this graph: genetically validated as selective, HSC-sparing, and located on a region of KMT2A that no breakpoint removes.
  2. Does disrupting MLL-PAFc also remove the SETDB1-dependent repressive arm, and if so does HOXA9 go up rather than down? The activating and repressive functions of PAFc have never been separated pharmacologically.
  3. Does MLL-PAFc disruption spare long-term HSC function under competitive transplantation, the stress test that exposed menin's HSC defect? The published reconstitution assays are less demanding than that.
  4. Is the MLL-PAFc interaction required in KMT2A::AFF1 B-ALL as it is in MLL-AF9 AML? Essentially all of the PAFc-MLL literature is myeloid.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The PAF complex binds MLL through sequences flanking the CxxC domain, a region retained in MLL fusion proteins, and synergises with MLL fusion proteins at HOX loci to promote leukemogenesis.

    The PAF complex synergizes with MLL fusion proteins at HOX loci to promote leukemogenesis. · Muntean AG, Tan J, Sitwala K, Huang Y, Bronstein J, Connelly JA, Basrur V, Elenitoba-Johnson KSJ, Hess JL · Cancer Cell 2010
    PMID 20541477DOI
    What this evidence cannot show

    The mapped contact is MLL CxxC-flanking region to PAFc as a complex; the work does not resolve which PAFc subunit makes the direct contact, so subunit-level interaction claims are not supported by this paper. Mouse and cell-line models, predominantly MLL-AF9 rather than KMT2A::AFF1.

    Evidence that argues against this (1)
    • Binding through sequences flanking the CXXC domain is the load-bearing claim and it is shared: those sequences are retained in the fusion AND present in wild-type KMT2A, so PAF1 recruitment is not fusion-specific. Deletions abolishing the interaction remove targeting for both. Cell Reports 2014
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Multiple distinct interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus, and wild-type MLL1 pre-binding is required for stable recruitment of the fusion.

    Multiple interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus in leukemogenesis. · Milne TA, Kim J, Wang GG, Stadler SC, Basrur V, Whitcomb SJ, Wang Z, Ruthenburg AJ, Elenitoba-Johnson KSJ, Roeder RG, Allis CD · Molecular Cell 2010
    PMID 20541448DOI
    What this evidence cannot show

    Recruitment mechanics at a single locus in cell models; generalisation to the genome-wide fusion target set is an extrapolation.

    Evidence that argues against this (1)
    • Wild-type MLL1 pre-binding being required for stable fusion recruitment is a striking claim with a large consequence — it makes the remaining wild-type allele part of the mechanism — and it rests on chromatin occupancy in cell lines. The later finding that endogenous MLL1 is dispensable with MLL2 sustaining the disease pulls against it. Cell Reports 2014
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    PAFc is required indiscriminately for AML proliferation, but targeted disruption of the MLL-PAFc interaction selectively inhibits MLL-fusion-driven AML while leaving normal haematopoietic stem cell function intact.

    MLL fusion protein-driven AML is selectively inhibited by targeted disruption of the MLL-PAFc interaction. · Muntean AG, Chen W, Jones M, Granowicz EM, Maillard I, Hess JL · Blood 2013
    PMID 23900238DOI
    What this evidence cannot show

    Selectivity is demonstrated with a dominant-negative MLL fragment, a genetic tool rather than a drug; no small molecule disrupts this interface. Mouse AML models, not infant KMT2A::AFF1 B-ALL. The HSC-sparing result comes from reconstitution assays over a limited period and does not exclude a long-term defect of the kind competitive transplantation revealed for menin.

    Evidence that argues against this (2)
    • The selectivity claimed for the MLL-PAFc interface does not extend to the complex itself, and PAF1C is required by normal stem cells through a route the entry does not cover. PAF1C partners with SETD5 and HCF-1 to release promoter-proximal paused RNA polymerase II at E2F targets in haematopoietic stem cells; disrupting that axis leaves phenotypic HSC numbers intact or increased while destroying long-term self-renewal and multilineage reconstitution under transplantation pressure. Any agent that degrades or destabilises PAFc rather than precisely occluding the MLL contact surface will hit this. The distinction between 'the interface is selective' and 'the complex is selective' is the whole therapeutic argument here, and it has not been tested pharmacologically in a human. Li M et al. 2021
    • Two of the four PAF1C subunits in this beat are germline tumour suppressors in childhood or familial cancer: CTR9 in Wilms tumour and CDC73/parafibromin in hyperparathyroidism-jaw tumour syndrome. That is a property of the complex, not of one subunit, and it means the safety case for any PAFc-directed agent in a paediatric population has to be made against the possibility of inducing a second malignancy. This objection applies to E3's claim of selective inhibition as much as to the complex as a whole, because a compound that disrupts the MLL-PAFc interface will in practice be assayed and dosed against PAFc. Hanks S et al. 2014
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PAF1 is DepMap common-essential with mean Chronos -0.832 across 1,258 lines, -0.822 across leukemia lines and depletion in 93.3% of lines, while gnomAD reports pLI 0.011 and LOEUF 0.611 with 31 observed against 68.4 expected loss-of-function variants, and HPA reports low tissue specificity with bone marrow 34.8 nTPM.

    What this evidence cannot show

    DepMap release version not stated by the API. The UniProt query for the symbol PAF1 returned two accessions; Q8N7H5 is the RNA polymerase II-associated factor, and the second hit is an unrelated protein sharing the alias. All figures here refer to Q8N7H5 / ENSG00000006712.

    Evidence that argues against this (1)
    • Pan-essentiality is the finding here and it argues against the target rather than for it. A gene depleted in more than 90% of all cell lines is required by the patient's normal marrow as much as by the leukaemia, so its appearance in a KMT2A complex is a statement about the complex and not a therapeutic opening. PAF1 at 93.3% of lines depleted with a pLI of 0.011 is also an instructive mismatch: cellular essentiality and germline constraint disagree here, which is why neither alone is a safety signal. gnomAD, Broad Institute 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    Across 54 DepMap leukemia lines PAF1 correlates with MEN1 at r = -0.095 with a confidence interval of -0.318 to 0.149 that crosses zero, the expected flat-profile result for a pan-essential gene.

    Leukemia-restricted co-dependency with MEN1, bootstrap and permutation testing; analyses/wrad-discrimination/ in this project · Broad Institute DepMap; Open Targets Platform; analysis run by this project · Database and derived analysis 2026
    What this evidence cannot show

    Our own computation, not peer reviewed, in 54 lines. A null correlation for a pan-essential gene is uninformative about biology: the method cannot detect co-dependency where there is no variance to correlate.

    Evidence that argues against this (1)
    • Same objection as for the other sub-threshold correlations in this beat: the deeper 2,743-gene variance-matched null with Holm correction leaves only KMT2A, PSIP1, DOT1L, FLT3, MEF2C and KAT7 standing, with MEIS1 already failing at Holm 0.067. LEO1's r = 0.302 at Holm 0.286 in the 16-gene panel is well below the genes that fail in the deep pool, so it should not be described as a signal of any kind. Broad Institute DepMap; Open Targets Platform; analysis run by this project 2026
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    PAF1 complex interactions with SETDB1 mediate promoter H3K9 methylation and transcriptional repression, including repression of HOXA9 and MEIS1, so PAFc is not a one-directional activator of the fusion output programme.

    PAF1 complex interactions with SETDB1 mediate promoter H3K9 methylation and transcriptional repression of Hoxa9 and Meis1 in acute myeloid leukemia. · Ropa J, Saha N, Chen Z, Serio J, Chen W, Mellacheruvu D, Zhao L, Basrur V, Nesvizhskii AI, Muntean AG · Oncotarget 2018
    PMID 29774127DOI
    What this evidence cannot show

    Published in Oncotarget, a journal with a weaker editorial reputation than the other citations here. The repressive arm is demonstrated in AML models; its weight relative to the activating arm in KMT2A::AFF1 B-ALL is unknown.

    Evidence that argues against this (1)
    • This bidirectionality is sharper than the entry states, and it has a clinical consequence. PAF1C's repressive arm is not confined to SETDB1: in normal haematopoietic stem cells PAFc works with SETD5 and HCF-1 to release paused RNA polymerase II at E2F targets and thereby maintain quiescence and long-term self-renewal, and disrupting that axis destroys reconstitution capacity while leaving phenotypic HSC counts intact or raised. An agent that reduces PAFc function to switch off the fusion's HOXA output therefore also removes a brake on HOXA9 and MEIS1 in the SETDB1 arm and removes a quiescence programme in the normal stem cell. The net effect on a HOXA-driven leukemia is not predictable from the activating arm alone. Li M et al. 2021
Connected to