Oncofusion

start / graph / interface-kmt2a-drc5-declined-an-axonemal-protein-from-a-centrosome-cilium-bioid-map

Interfaceweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.q160-the-prey-only-interactome

KMT2A-DRC5 - DECLINED. A ciliary dynein-regulatory-complex protein, recovered by BioID in a centrosome-cilium map, and the only one of 135 uncovered partners never mentioned anywhere in this corpus

also called KMT2A-DRC5 · DRC5 · dynein regulatory complex subunit 5

The one candidate in this whole list that nobody here has ever written a word about turns out to be a cilium protein, picked up as a neighbour of KMT2A in a map of the cell's microtubule-organising centre.

The same thing, for a scientist

**Verdict: DECLINED.** IntAct holds two identical KMT2A-DRC5 (Q03164-Q5JU00) rows: **proximity-dependent biotin identification**, interaction type *proximity*, HEK293, KMT2A as prey, intact-miscore **0.27** - the lowest in the shortlist - from **PMID 26638075** [E1]. That is Gupta et al., *Cell* 2015: a BioID map of the **centrosome-cilium interface** using **58 bait proteins** and generating a topology network of **>7,000 interactions**, analysed for roles in centriole duplication, ciliogenesis and centriolar satellite biogenesis [E2]. **KMT2A is not a bait, is not in the abstract, and has no ciliary function.** DRC5 is a subunit of the axonemal nexin-dynein regulatory complex. **This row is recorded because of what it is, not what it shows: it is one of only 19 of the 135 uncovered prey-only partners that a word-boundary scan finds nowhere in this corpus** [E3] - and the reason it is unmentioned is that it is a cilium protein in a leukaemia graph, not that it was overlooked.

Go deeper
'Never mentioned anywhere' turned out to mean 'correctly ignored'

The census flagged 19 of the 135 uncovered partners as never mentioned anywhere in the corpus. CORRECTED before shipping: this entry first said DRC5 was the highest-ranked of the 19. It is not - SMAD9, FAM9B, FOXE1, FOXK2, HMGN5, KLHL22, MAGEB2, MRPS26, MYO1E, NEDD4 and four SLC transporters all rank above it, and three of those (HMGN5, KLHL22, FOXK2) are written up elsewhere in this beat. DRC5 was picked out of the 19 for a different reason: under the PSICQUIC grading this beat started from, it was the ONLY never-mentioned uncovered partner, which is exactly what a novelty signal is supposed to look like.

It was not. Opening the source resolved it in one step: a ciliary axonemal protein, in a centrosome BioID map, at the score floor, with KMT2A as an incidental neighbour.

The lesson is about the metric rather than the protein. corpus_mentions == 0 was being used as a novelty signal - it appears in the census output and in interactome-bait-versus-prey before it. It is not a novelty signal; it is a relevance signal running backwards. The partners nobody has mentioned are disproportionately the ones with no plausible connection, because 2,931 entries of active curation have already absorbed everything with one. A ranking that promotes unmentioned partners promotes irrelevance.

Proximity is not contact, and two rows are not two experiments

The interaction type on both rows is proximity, not association or physical association. BioID biotinylates within roughly 10 nm of the bait; a proximity record is a statement about neighbourhood, and in a crowded nucleus or around a centrosome the neighbourhood is large.

And the two rows are one experiment. They are byte-identical in bait, method, type, host and score. This is the row-versus-experiment trap the corpus already records for the four AFF4 two-hybrid partners: a MITAB or REST record is a PSI-MI assay annotation, and a ranking by record count double-counts it. The census ranks by distinct PMIDs for exactly this reason, and DRC5's PMID count is one.

What would be needed to reopen it: a reason for KMT2A to be at a centrosome. The literature does contain reports of chromatin regulators at centrosomes, so it is not absurd; there is simply nothing here, and no experiment in this record set addresses it.

Facts and figures
partner agene-kmt2a
partner bDRC5
partner b uniprotQ5JU00
partner b length AA501
partner a uniprotQ03164
partner a length AA3969
verdictDECLINED
adjudicated byq160, cycle 146, 2026-09-24
druggabilityNot assessable and not relevant.
n methods1
n platforms1
n publications1
n records2
endogenous sideno
intact miscore0.27
interaction type miproximity
cell statethe source manipulates ciliogenesis; this row's condition is not recorded
epigeneticsNone.
decline reason classincidental proximity neighbour in an organelle-focused BioID atlas; compartment and function both unrelated
evidence statenot-worth-measuring
evidence state noteNo experiment is proposed. This entry exists so that the next database sweep does not re-surface DRC5 as an unexamined top-ranked gap.
evidence state why not absence classThis entry is not an absence. It reached for absence_class to record an evidence state that vocabulary cannot express - 'not-worth-measuring' - so the value moved to evidence_state rather than being folded into an absence token that would misdescribe it. See q190: only 69 of the 529 entries carrying absence_class were absence entries.
interface residues a
[]
structures
[]
ligands
[]
records
[
  "KMT2A-DRC5 x2 (identical), proximity-dependent biotin identification, PROXIMITY, HEK293, KMT2A prey, miscore 0.27, PMID 26638075"
]
cell systems
[
  "HEK293 embryonic kidney"
]
localization
{
  "compartment": "axoneme / centrosome-cilium interface",
  "subnuclear": [],
  "shown_for_fusion": "NO - nobody has looked, and there is no reason to."
}
measurements
[
  {
    "quantity": "count",
    "subject": "uncovered human prey-only partners with zero corpus mentions",
    "value": 19,
    "units": "partners",
    "n": 135,
    "universe": "the 135 uncovered partners in the 2026-09-24 census, word-boundary scan of 2,931 node files on the current HGNC symbol only",
    "null": null,
    "note": "DRC5 is one of the 19 and is a cilium protein. Under the earlier PSICQUIC grading it was the only one. Zero mentions is not a novelty signal."
  }
]
How this could bite us back
  • Declining the only never-mentioned candidate makes the beat's headline gap smaller and is therefore the kind of decline that flatters the writer. It is recorded with the evidence so the next reader can disagree with it cheaply.
Still open
  1. Are the other 18 never-mentioned partners the same shape - correctly ignored rather than overlooked? Only DRC5 has been opened, and if the pattern holds then 'never mentioned in the corpus' should be dropped from every ranking in this project rather than kept as a novelty column.
  2. How many rows in this corpus's interaction analyses are duplicate records of a single experiment? DRC5 is two; the AFF4 two-hybrid set is twelve rows from one screen. Nothing counts this systematically.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    IntAct REST returns two KMT2A-DRC5 (Q03164-Q5JU00) records, identical in every recorded field: detection method 'proximity-dependent biotin identification', interaction type 'proximity', host Homo sapiens HEK293 embryonic kidney cell, KMT2A as PREY, intact-miscore 0.27, both from PMID 26638075. 0.27 is the lowest score in this beat's shortlist.

    What this evidence cannot show

    Two rows, one experiment. They differ only in interaction AC and are counted once in this beat's PMID-based ranking; any ranking by record count would double them. Interaction type 'proximity' is a neighbourhood statement, not a contact statement.

    Evidence that argues against this (1)
    • BioID has recovered genuine, hard-to-capture associations that co-IP misses, and dismissing proximity records as a class would discard the method's main advantage. The objection here is specific: a ciliary axonemal protein, a centrosome-focused bait panel, and no route by which KMT2A would be there. Gupta GD et al. 2015
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    PMID 26638075 (Gupta et al., Cell 2015) used proximity-dependent biotinylation with 58 bait proteins to map the centrosome-cilium interface, generating a protein topology network comprising more than 7,000 interactions, and analysed it for protein modules in centriole duplication, ciliogenesis and centriolar satellite biogenesis, including dynamic changes during ciliogenesis. KMT2A is not among the named baits and does not appear in the abstract.

    A Dynamic Protein Interaction Landscape of the Human Centrosome-Cilium Interface. · Gupta GD, Coyaud E, Goncalves J, Mojarad BA, Liu Y, Wu Q, Gheiratmand L, Comartin D, et al. (20 authors), Pelletier L · Cell 2015
    PMID 26638075DOI
    What this evidence cannot show

    Abstract read in full; the >7,000-interaction table was not opened. A 58-bait, 7,000-interaction map is an atlas; this beat's >=15-partner heuristic did not flag it because it touches only one partner in this census, which is the same detector failure recorded against PMID 25609649.

    Evidence that argues against this (1)
    • A large BioID map is a hypothesis-generating resource and its authors make no claim about KMT2A; treating its incidental rows as claims by the paper would be unfair to it. The decline is of the database row, not of the study. Gupta GD et al. 2015
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    A word-boundary scan of all 2,931 node files on 2026-09-24 finds DRC5 mentioned zero times. It is one of 19 uncovered human prey-only partners with zero corpus mentions; the others are SMAD9, FAM9B, FOXE1, FOXK2, HMGN5, KLHL22, MAGEB2, MRPS26, MYO1E, NEDD4, SLC15A1, SLC25A22, SLC35G3, SLC7A6, FHIP1B, RAVER1, SMC5 and SOX7. Under the PSICQUIC grading this beat started from, DRC5 was the only one.

    What this evidence cannot show

    The scan matches the current HGNC symbol only. DRC5's HGNC aliases and previous symbols were NOT scanned, so a mention under the retired symbol would have been missed - the exact failure scripts/is-symbol-covered.py exists to prevent, and this scan does not do what that tool does.

    Evidence that argues against this (1)
    • Zero mentions was treated as a novelty signal by this beat's own ranking and by the analysis it inherited from. Opening the source shows it is closer to the opposite: in a heavily curated corpus, an unmentioned partner is more likely to be irrelevant than to be overlooked. this project 2026
Connected to