Oncofusion

start / graph / gene-parp16

Geneweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q75-dot1l-partners

PARP16 - graded A- in the DOT1L pulldown, the best of the six, and a tail-anchored ER membrane protein that cannot be in the nucleoplasm

also called ARTD15 · C15orf30 · Protein mono-ADP-ribosyltransferase PARP16 · pART15

PARP16 is an enzyme that sits permanently stitched into the membrane of the endoplasmic reticulum, a compartment outside the nucleus. A DOT1L pulldown recovered a small amount of it, which is hard to reconcile with DOT1L living inside the nucleus.

The same thing, for a scientist

PARP16 (Q8N5Y8, 322 aa) is the only mono-ADP-ribosyltransferase of the PARP family that is an integral membrane protein: UniProt records it as **endoplasmic reticulum membrane, single-pass type IV (tail-anchored)**, on two experimental codes [E3]. It is also the **best-graded of the six unwritten DOT1L partners - class A-** [E2], which makes it the one that most deserves an explanation rather than a dismissal. The numbers are otherwise unremarkable: interaction stoichiometry 0.0218 (1 PARP16 per 46 DOT1L), abundance stoichiometry 3.4, enrichment 2.09 SD, prey-bait profile correlation **0.043**, `core.interactor` not set, and the reported -log10 p is Inf and therefore unusable [E2]. **The compartment argument is the substantive one and it is a prior, not a proof**: a tail-anchored ER protein and a nucleoplasmic methyltransferase have no shared compartment in which to meet, and a detergent lysate abolishes exactly that separation. `DOT1L[tiab] AND PARP16[tiab]` and `(MLL OR KMT2A)[tiab] AND (PARP16 OR ARTD15)[tiab]` both return 0 [E4].

Go deeper
Why an A- grade on a compartment-incompatible prey is the interesting part

Five of the six unwritten DOT1L spokes are class B or B-. PARP16 is A-, one step below the DotCom subunits, and the class is the paper's composite of enrichment, p value and prey-bait correlation 2. Two readings, and the entry does not choose between them:

  1. The class is carried by the p value, which is degenerate. PARP16's -log10.p.value is reported as Inf in the source table - the Welch test returned an infinite value, which happens when the control samples have zero variance for that protein, i.e. when it was never quantified in the controls at all. A composite score that includes an infinite term will rank a prey highly for a reason that is an artefact of the control, not a strength of the signal. EBNA1BP2's p value is also Inf, and EBNA1BP2 is graded B-, so the p value is clearly not the only term - but it is the one that differs here, and the correlation (0.043) and enrichment (2.09 SD) are both in the B- range.
  2. The interaction is real and compartmentally surprising. PARP16 has a documented role in the unfolded protein response and in ADP-ribosylating PERK and IRE1, and tail-anchored proteins are inserted post-translationally, so a cytosolic pre-insertion pool exists. DOT1L itself has an AF17-dependent cytoplasmic pool in this graph's own record 5. A cytoplasmic encounter is not impossible.

What settles it is a reciprocal experiment, and PARP16 was not a bait in this study 6. Recorded as unresolved rather than argued to a conclusion.

The compartment argument, stated at the strength it actually has

Four of the six unwritten DOT1L spokes are annotated to compartments with no nuclear component at all - PARP16 (ER membrane), NIN (centrosome), BRINP1 (cytoplasm) and IBTK (cytoplasm and membrane, except its nuclear isoform 2). That is a pattern worth recording and it is also exactly the kind of reasoning this graph has already caught itself getting wrong.

localization-partner-baseline-compartment-is-often-miscalled exists because baseline compartment annotations in this corpus are frequently incomplete: UniProt gives AF9 nucleus and chromosome and no cytoplasm at all, and AF9 has a documented cytoplasmic pool. So 'UniProt says ER, therefore not a nuclear partner' is the same inference in the same direction that the graph has already shown to fail.

The defensible version is narrower. A single-pass type IV membrane protein is not a soluble protein with an under-annotated second location - it is physically embedded in a lipid bilayer by a C-terminal transmembrane helix, and its topology is a structural fact rather than a compartment call 3. That makes the prior lower than for a merely cytoplasmic annotation. It still does not exclude a cytosolic pre-insertion pool, and it says nothing at all about what happens after a cell is lysed in detergent.

Facts and figures
uniprotQ8N5Y8
length AA322
gene symbolPARP16
protein nameProtein mono-ADP-ribosyltransferase PARP16 (ARTD15)
uniprot resolutionaccession:Q8N5Y8 AND organism_id:9606 AND reviewed:true; one reviewed hit; primary gene name read back as PARP16; length 322 aa checked
shown for fusionNO - nobody has looked in any fusion. Zero PubMed records pair MLL or KMT2A with PARP16 or ARTD15.
interaction with dot1l
{
  "source_pmid": "26496610",
  "source_class": "proteome-scale AP-MS atlas",
  "intact_interaction_ac": "EBI-11060394",
  "imex_id": "IM-24272-464",
  "detection_method": "psi-mi:MI:0007 anti tag coimmunoprecipitation",
  "interaction_type": "psi-mi:MI:0914 association",
  "expansion": "psi-mi:MI:1060 spoke expansion - binary pair inferred, not observed",
  "intact_miscore": 0.35,
  "author_confidence_class": "A-",
  "author_confidence_note": "the highest grade among the six unwritten DOT1L partners, and unexplained by anything visible in the table",
  "core_interactor": false,
  "interaction_stoichiometry_prey_per_bait": 0.02183,
  "interaction_stoichiometry_as_ratio": "1 PARP16 per 46 DOT1L",
  "log10_prey_bait_ratio": -1.661,
  "abundance_stoichiometry_prey_per_bait": 3.433,
  "log10_prey_bait_expression_ratio": 0.5356944,
  "enrichment_sd": 2.09,
  "neg_log10_p": null,
  "neg_log10_p_note": "reported as Inf in the source table and therefore unusable",
  "prey_bait_correlation": 0.043,
  "valid_values": 3,
  "hela_copies_per_cell": 12034.43,
  "independent_sources": 0,
  "reciprocal_tested": false,
  "reciprocal_note": "PARP16 is not one of the study's 1,330 bait cell lines"
}
localization
{
  "subcellular": "Endoplasmic reticulum membrane, single-pass type IV (tail-anchored) membrane protein, on two UniProt experimental codes, with no nuclear annotation. Topology is a structural fact rather than a compartment call, which makes the compartment mismatch with a nucleoplasmic methyltransferase stronger than a merely cytoplasmic annotation would be - but a transient cytosolic pre-insertion pool exists for every tail-anchored protein.",
  "source": "UniProt Q8N5Y8, fetched 2026-09-21; ECO:0000269 on PMID 23103912 and PMID 22701565, neither opened (brief rule 91)"
}
absence claim
{
  "queries_run": [
    "DOT1L[tiab] AND PARP16[tiab]",
    "(MLL OR KMT2A)[tiab] AND (PARP16 OR ARTD15)[tiab]"
  ],
  "records": [
    0,
    0
  ],
  "operand_controls": {
    "DOT1L[tiab]": 710
  },
  "names_tried": [
    "PARP16",
    "ARTD15"
  ],
  "names_not_tried": [
    "pART15",
    "C15orf30",
    "mono-ADP-ribosyltransferase as a phrase"
  ],
  "date": "2026-09-21",
  "tool": "PubMed E-utilities esearch",
  "bounded_by": "titles and abstracts only; supporting figures not searched; PARP-family work is often framed by pathway rather than paralogue symbol (brief rule 71)"
}
How this could bite us back
  • PARP16 is a druggable enzyme and 'DOT1L partner' would make it look like a target in this disease. It rests on one A- spoke of one screen with a correlation of 0.043 and an unusable p value, and no reciprocal arm exists.
  • If the graph adopts the source confidence class as a filter, PARP16 passes and EBNA1BP2 - which has ten times the stoichiometry - fails. Any filter must be justified against both cases or it is just picking the column that gives the wanted answer.
  • Dismissing this on compartment grounds repeats the exact inference localization-partner-baseline-compartment-is-often-miscalled was written to warn against, and the entry is deliberately written to stop short of that.
Still open
  1. Why is PARP16 graded A- when its enrichment and prey-bait correlation sit in the B- range? The class threshold definitions are in the paper's unreadable methods, and resolving this would tell the graph how much weight the class deserves as a filter.
  2. Is PARP16 a bait in BioPlex or any other AP-MS atlas, and does DOT1L appear in its prey list? That is the reciprocal test this pair does not have, and it is one database query away.
  3. Does DOT1L have a cytosolic pool in cells where AF17 is expressed, and does anything membrane-associated come down with it? The AF17-dependent cytoplasmic pool in this graph is inferred by similarity and has never been measured in a human cell.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The entire recorded interaction evidence for PARP16 with DOT1L is one row of one IntAct interaction record - EBI-11060394 / IM-24272-464, spoke-expanded (MI:1060) from a single anti-tag co-immunoprecipitation of LAP-tagged DOT1L in HeLa, interaction type 'association' (MI:0914), intact-miscore 0.35, author-confidence A- - from PMID 26496610; PARP16 appears exactly once across all 49 IntAct rows involving DOT1L from seven publications.

    What this evidence cannot show

    One database on one day. Spoke expansion means the binary pair was inferred by a curator from co-purification and never observed as a binary interaction. HeLa cervical carcinoma only - no leukaemia cell, no KMT2A fusion, no primary sample (brief rule 73).

    Evidence that argues against this (1)
    • The same pulldown recovered the genuine DotCom core at A+, so the experiment works; a single appearance in IntAct bounds the curated record rather than the literature, and an uncurated supplementary table would be invisible to this count. Hein MY et al. 2015
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    In the source study's interactions table the DOT1L-PARP16 spoke is class A- - the highest grade of the six unwritten partners - with interaction stoichiometry 0.02183 (1 PARP16 per 46 DOT1L, log10 prey:bait -1.661), abundance stoichiometry 3.433, enrichment 2.090 SD, prey-bait profile correlation 0.043, core.interactor not set, and a -log10 p value reported as Inf; PARP16 is present in the paper's HeLa proteome at 12,034 copies per cell.

    What this evidence cannot show

    Downloaded 2026-09-21 from the Elsevier static CDN with magic bytes checked (PK). The Inf p value is recorded as null in data rather than dropped, and the inference that an infinite Welch statistic reflects zero control variance is MY reading of what produces an Inf, not a statement from the paper, whose methods section I could not read. The prey group is Q8N5Y8;Q8N5Y8-3;Q8N5Y8-2 and does not distinguish isoforms.

    Evidence that argues against this (1)
    • EBNA1BP2 also carries an Inf p value in the same table and is graded B-, so the Inf cannot by itself be what lifted PARP16 to A-. The degenerate-p explanation is therefore incomplete and the A- grade is not fully accounted for by anything visible in the table. Cell 163(3), supplementary material 2015
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProtKB Q8N5Y8 (PAR16_HUMAN) is human PARP16, 322 aa, 'Protein mono-ADP-ribosyltransferase PARP16 / ADP-ribosyltransferase diphtheria toxin-like 15', gene synonyms ARTD15 and C15orf30, with subcellular location 'Endoplasmic reticulum membrane (ECO:0000269, PMID 23103912); Single-pass type IV membrane protein (ECO:0000269, PMID 22701565)' and no nuclear annotation.

    What this evidence cannot show

    Resolved 2026-09-21 by accession with the primary gene name read back as PARP16 and the length checked. Per brief rule 91 the ECO:0000269 codes mean a curator read an experiment in PMID 23103912 and PMID 22701565; I did not open either, so the topology call is coded-but-untraced. 'Single-pass type IV' means tail-anchored - a C-terminal transmembrane helix inserted post-translationally - which implies a transient cytosolic pool before insertion and is not stated in the UniProt comment.

    Evidence that argues against this (1)
    • This graph's own localization-partner-baseline-compartment-is-often-miscalled records that UniProt baseline compartments in this corpus are frequently incomplete - AF9 is annotated nucleus and chromosome with no cytoplasm and has a documented cytoplasmic pool. An absent nuclear annotation is therefore not a measured absence of PARP16 from the nucleus. UniProt Consortium 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Counted absence, PubMed E-utilities esearch run 2026-09-21: DOT1L[tiab] AND PARP16[tiab] returns 0 records and (MLL OR KMT2A)[tiab] AND (PARP16 OR ARTD15)[tiab] returns 0 records, with the operand control DOT1L[tiab] returning 710.

    What this evidence cannot show

    Bounded by titles and abstracts (brief rule 80). Two aliases were run against the DOT1L limb and one spelling of the gene symbol; the retired-style alias pART15 and the spelled-out enzyme names were not run, though PARP16 is an unambiguous and current symbol with no known collision. A PARP-family paper framed at the level of 'mono-ADP-ribosylation' rather than the gene symbol would be unreachable - brief rule 71's vocabulary problem.

    Evidence that argues against this (1)
    • The whole PARP-inhibitor literature is framed by drug and pathway names rather than by paralogue symbols, which is exactly the vocabulary failure rule 71 documents, so this pair of zeros is a weaker bound than it looks for a PARP-family gene. NCBI PubMed (queries run 2026-09-21) 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    UniProtKB Q8TEK3 annotates DOT1L as Nucleus on an experimental code and Cytoplasm by similarity only, with the note that cytoplasmic localisation is dependent on and mediated by MLLT6/AF17 - so a cytoplasmic DOT1L pool exists in the record and is partner-determined rather than constitutive.

    What this evidence cannot show

    Fetched 2026-09-21. The cytoplasmic call is ECO:0000250, inferred by similarity to the mouse orthologue Q6XZL8, and is not a human measurement - this graph records that separately in localization-dot1l-nucleus-measured-cytoplasm-by-similarity. Using it to make a cytoplasmic PARP16 encounter plausible is therefore leaning on an inferred annotation.

    Evidence that argues against this (1)
    • MLLT6/AF17, the partner said to mediate DOT1L's cytoplasmic pool, is present in this same pulldown at A+ - so if a cytoplasmic DOT1L pool exists in these cells it is a shared route to PARP16 and to AF17 at once, and the pulldown cannot separate a genuine cytoplasmic contact from co-recovery of whatever else that pool touches. Cell 163(3), supplementary material 2015
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    PARP16 is not among Hein et al.'s 1,330 bait cell lines, so the DOT1L-PARP16 pair has no reciprocal arm anywhere in the study that reported it.

    What this evidence cannot show

    A bait list is an absence within one study. PARP16 may well be a bait in another AP-MS atlas (BioPlex was not checked in this beat), and that check is the cheapest next step for this entry.

    Evidence that argues against this (1)
    • The absence of a reciprocal arm is a gap in the evidence, not evidence against, and it applies equally to every prey in this pulldown except IBTK - so it cannot be used to rank PARP16 below the others. Cell 163(3), supplementary material 2015
Connected to