Oncofusion

start / graph / ana-proximity-interactome-enumeration

Analysisestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.Q20-proximity-interactome

Proximity-interactome beat: every query run, every paper rejected, and the one resource not reachable

also called Q20 enumeration · proximity beat enumeration

A record of exactly what was searched, what was found, what was read and set aside, and what could not be reached - so that anyone repeating this next year repeats the decisions and not only the searches.

The same thing, for a scientist

**Searched:** 13 PubMed queries (proximity vocabulary x KMT2A/MLL/AFF1/AFF4/ELL/EAF/MLLT/CDK9/DOT1L/MEN1/leukaemia/fusion) with query translations read, plus 14 IntAct PSICQUIC queries by accession and by PSI-MI method code, plus one Europe PMC full-text retrieval and one NCBI elink to GEO [E1]. **Found and used:** 8 proximity publications contributing records for this graph's proteins, of which 4 are written up as entries [E2]. **Rejected with reasons:** 9 papers and 1 whole query class [E3]. **Not reachable:** 2 resources, both named with the reason [E4]. **Unreadable: none.** Every paper cited here returned either an abstract or full text; the one paper the beat depends on most was read in full from PMC [E2].

Go deeper
Queries run, with counts

All PubMed via E-utilities esearch on 2026-09-19; PL abbreviates (BioID OR TurboID OR miniTurbo OR APEX2 OR "proximity labeling" OR "proximity labelling" OR "proximity-dependent biotinylation").

queryrecordson-beat after title check
PL AND (KMT2A OR MLL)00
PL AND (AFF1 OR AF4 OR MLLT2)00
PL AND (AFF4 OR "super elongation complex" OR SEC)120 (SEC is a contaminated alias - Sec bodies, Sec61, cyanobacteria)
PL AND (ELL OR ELL2 OR EAF1 OR EAF2)00
PL AND (MLLT1 OR ENL OR MLLT3 OR AF9 OR MLLT10 OR AF10)10
PL AND (CDK9 OR "cyclin T1" OR CCNT1 OR "P-TEFb")41 (method only)
PL AND (DOT1L OR MEN1 OR menin)30
PL AND (leukemia OR leukaemia)383
PL AND ("fusion oncoprotein" OR "fusion protein" AND oncogen*)71
"proximity-dependent biotinylation" AND "RNA binding proteins"151
BioID AND MLL / BioID AND KMT2A / TurboID AND KMT2A0 / 0 / 0-
nine-method OR x five-disease OR pooled query0-
("proximity proteomics" OR "proximity interactome" OR "biotin ligase") AND (KMT2A OR MLL OR AFF1 OR AFF4)0-

Controls, to show the zeros are measurements: BioID 853, "proximity labeling" 1,455, "proximity labelling" 186, "proximity proteomics" 158, KMT2A 3,216, MLL 7,520, BioID AND leukemia 14, TurboID AND leukemia 12, (BioID OR TurboID) AND leukemia 26. Query translations were read for the zero-returning terms; "BioID"[All Fields] AND "MLL"[All Fields] is the literal translation and it returns 0 1.

IntAct PSICQUIC, same date: detmethod:"MI:1314" 64,945 rows total; LMNA (P02545) 259 as positive control; then per accession for AFF1, AFF4, ELL, ELL2, MLLT1, MLLT3, MLLT10, CDK9, CCNT1, DOT1L, MEN1, KMT2A; plus pubid:39251607 for the full 19,403-row deposition 1.

Found and used
PMIDwhat it contributeswritten up
3925160750-RBP BioID2 map in K562; AFF1's entire modern proximity recordana-rbp-bioid2-proximity-map-k562 and 8 more
35140242109-TF BioID network; KMT2A x11, MLLT10 x3, DOT1L x2, MLLT1 x1ana-tf-bioid-network-reaches-kmt2a-and-af10
3693125914-3-3 proximity study; KMT2A prey of six isoformsana-kmt2a-14-3-3-proximity-two-studies
38061017TurboID fused to three AML oncofusion cDNAs in primary murine cellsana-turboid-of-aml-oncofusions-already-done
32664447BioID of NUP98::HOXA9 and SET::NUP214same entry
29568061MAC-tag; AFF4, MEN1, MLLT1 as prey of a histone H3.1 bait; ELL of CDK7census only
26638075centrosome-cilium landscape; KMT2A prey of TCTE1, MLLT10 of CPAPcensus only
35384245RTK interactome atlas; KMT2A prey of EPHB4census only

Also used for method: 22412018 (BioID), 30125270 (TurboID), 25281560 (BioID vs AP-MS on chromatin), 34079125 (Human Cell Map) 2.

Bounded rejections - found, on-topic-looking, and set aside

Three lines each: identifier, reason, what would change my mind 3.

  • PMID 34215314 (AF10/MLLT10 prevents somatic cell reprogramming through DOT1L-mediated H3K79). Returned by the MLLT proximity query; it is not a proximity-labelling study - the term matched incidentally. Would change my mind: if its methods contain a BioID or APEX arm not reflected in the abstract.
  • PMID 35690146 (Dot1L recruits OGT to chromatin). On-topic for DOT1L, not a proximity study of the elongation complex, and O-GlcNAc on DOT1L is a different beat. Would change my mind: if this graph opens a DOT1L post-translational-modification beat.
  • PMID 26836506 (validation of the mass-extraction window in LC-HRMS). A pure analytical-chemistry paper returned by the MEN1 term. Rejected as a symbol artefact.
  • PMID 36012204 (CFTR interactome by differential proximity labelling). Contributes two IntAct rows placing CDK9 near a CFTR bait. Rejected as biologically uninterpretable here: an epithelial chloride channel bait says nothing about P-TEFb in leukaemia. Would change my mind: nothing short of a nuclear CFTR result.
  • PMID 28330616 (human protein phosphatase interactome). One row, CDK9 prey of PPP5C. Kept in the census, not written up: a single row from a phosphatase survey does not support an entry. Would change my mind: independent evidence of PPP5C acting on CDK9's T-loop.
  • PMID 26638075 and PMID 35384245 (centrosome-cilium landscape; RTK interactome). Each contributes one or two rows placing KMT2A or MLLT10 near a centrosomal or receptor-kinase bait. Rejected as compartment artefacts of abundant nuclear preys in non-nuclear bait surveys. Would change my mind: a replicated cytoplasmic pool of KMT2A.
  • PMID 31381561 (PAX5 is part of a functional transcription factor network targeted in lymphoid leukaemia). This is the closest rejection and the one most likely to be wrong. It is a BioID study in B-lineage leukaemia identifying 239 PAX5-associated proteins including IKZF1 and RUNX1 - a modern proximity map in the right disease. It was not written up because PAX5 is not in this beat's protein set and IntAct serves no AFF1, AFF4 or KMT2A row from it. Would change my mind: if its 239-protein list contains any SEC or KMT2A-complex member, which this beat did not check because the list is in supplementary material. This is the first thing a repeat of this beat should do.
  • PMID 39706164 (TurboCas: locus-specific labelling of genomic regions and isolation of their protein interactome). A method paper, not a result about this biology. Recorded as a route: TurboCas at a HOXA9 enhancer in a t(4;11) line is a different and possibly better experiment than tagging the fusion. Would change my mind: nothing - it is already recorded as an opportunity rather than rejected.
  • The whole class of gene-symbol-plus-method PubMed queries is rejected as an instrument for this question, on the evidence of this beat: 13 such queries return 0-1 on-beat records while the structured database holds 46 KMT2A proximity rows.
Not reachable, with the reason
  • The Human Cell Map (192 BioID baits in HEK293, PMID 34079125). Its web application at humancellmap.org is a JavaScript single-page app; two candidate REST paths were probed on 2026-09-19 and both returned the app shell rather than data. Not queried. Whether AFF1 or KMT2A appears as a prey there is unknown, and every IntAct-bounded zero in this beat is caveated accordingly 4.
  • BioGRID, which curates proximity data under its own method vocabulary (Proximity Label-MS). Its REST API requires an access key that was not available in this session. Not queried 4.

Neither is an absence claim. Both are named so that a repeat of this beat starts there.

Unreadable: none

No paper in this beat qualifies for the unreadable register. PMID 39251607, on which most of the beat rests, was read in full: Europe PMC search?query=EXT_ID:39251607&resultType=core gives PMC11385195, and PMC11385195/fullTextXML returned HTTP 200 with a <body> element and 141,821 characters of text. Every other paper cited returned an abstract through efetch. Papers read at abstract level only are marked as such in their own citation caveats 2.

Facts and figures
scriptnone (queries)
statusrun
n pubmed queries13
n intact queries14
n publications found10
n written up4
n rejected9
n unreachable resources2
n unreadable0
highest risk rejection31381561
first thing a repeat should doCheck PMID 31381561's 239-protein PAX5 BioID list for any SEC or KMT2A-complex member; it is the only modern proximity dataset found in B-lineage leukaemia.
result summary13 PubMed and 14 IntAct queries; 10 publications found, 4 written up, 9 rejected with reasons, 2 resources unreachable, 0 unreadable.
rejected pmids
[
  "34215314",
  "35690146",
  "26836506",
  "36012204",
  "28330616",
  "26638075",
  "35384245",
  "31381561",
  "39706164"
]
unreachable
[
  "Human Cell Map (humancellmap.org) - JavaScript SPA, no REST endpoint found",
  "BioGRID REST - access key required"
]
How this could bite us back
  • An enumeration entry makes a beat look complete. This one lists two resources it could not query and one rejection it believes may be wrong, and those three items are where the missing biology most likely is.
  • Recording rejections creates a record that future cycles may treat as settled. Each rejection here names the condition that reverses it precisely so it is not.
Still open
  1. Does PMID 31381561's 239-protein PAX5 BioID list contain any SEC or KMT2A-complex protein? This is the beat's single highest-value unchecked item.
  2. Does the Human Cell Map contain AFF1 or KMT2A as a prey, and would that change any zero in this beat?
  3. Does BioGRID's Proximity Label-MS vocabulary hold rows IntAct does not curate?
  4. Would a TurboCas experiment at a HOXA9 enhancer in a t(4;11) line be a better use of a proximity construct than tagging the fusion itself?
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Thirteen PubMed queries and fourteen IntAct PSICQUIC queries were run on 2026-09-19 with the counts tabulated in this entry; the zero-returning PubMed terms had their query translations read to confirm the zeros are not parser artefacts, and the IntAct method filter was controlled with 64,945 total MI:1314 rows and 259 for LMNA (P02545).

    What this evidence cannot show

    A counted absence is only as good as its symbol set. Retired and variant symbols searched: MLL, MLL1, mixed lineage leukemia, mixed-lineage leukemia, AF4, MLLT2, ENL, AF9, AF10. Not searched: AF4 as a bare token without a proximity term, or any non-English literature. The SEC alias is contaminated and its hits were title-checked rather than counted.

    Evidence that argues against this (1)
    • This beat's own central finding is that these queries are the wrong instrument: the same searches return zero while the structured database holds 46 curated KMT2A proximity rows. The query table is therefore a demonstration of blindness, not a literature census. IntAct molecular interaction database (EMBL-EBI) 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Eight publications contribute proximity-dependent biotin identification records for this graph's proteins: 39251607, 35140242, 36931259, 29568061, 26638075, 35384245, 28330616 and 36012204, with 38061017 and 32664447 contributing fusion-oncoprotein proximity maps found through the literature rather than the database; PMID 39251607 was read in full from PMC11385195 (HTTP 200, body element present, 141,821 characters).

    What this evidence cannot show

    The eight are what IntAct curates; the two fusion-oncoprotein papers are not in the IntAct proximity record for these accessions and were found only through the leukaemia query, which is why both routes were run. No claim is made that these ten are all the proximity literature touching this biology.

    Evidence that argues against this (1)
    • Two of the ten were found only by a title-and-abstract query and eight only by a method-code query. Neither route found the other's papers, so running one route alone would have produced a confident and incomplete answer in either direction. NCBI PubMed (E-utilities) 2026
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Nine identified papers and one whole query class were considered and rejected, each with a stated reason and a stated condition that would reverse the decision; the closest rejection is PMID 31381561, a BioID study of PAX5 in lymphoid leukaemia identifying 239 associated proteins including IKZF1 and RUNX1, set aside because PAX5 is outside this beat's protein set and its 239-protein supplementary list was not checked for SEC or KMT2A-complex members.

    PMID 31381561DOI
    What this evidence cannot show

    Abstract read; full text and supplementary data not read. Recording the rejection does not make it correct, and this one is explicitly flagged as the most likely to be wrong. The 239-protein list is the specific unchecked object.

    Evidence that argues against this (1)
    • A BioID study in B-lineage leukaemia is the closest published thing to what this beat was dispatched to find, and it was rejected on a protein-set criterion rather than on its content. If its 239-protein list contains an SEC member, the rejection was an error and the beat missed a directly relevant human B-ALL proximity dataset. Okuyama K et al. 2019
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Two proximity resources were identified and not queried: the Human Cell Map (192 BioID baits in HEK293), whose web application returned the single-page-app shell rather than data at two candidate REST paths on 2026-09-19, and BioGRID, whose REST API requires an access key unavailable in this session.

    PMID 34079125DOI
    What this evidence cannot show

    The citation establishes what the Human Cell Map is, not that it is unreachable; the unreachability is this session's observation and a browser-driven retrieval or a bulk download would very likely succeed. This is a not-queried, not an absence.

    Evidence that argues against this (1)
    • Both resources publish bulk downloads, so 'not reachable' means 'not reached by the routes tried in one session'. Every IntAct-bounded zero in this beat inherits that limit and should be read as bounding IntAct, not the field. IntAct molecular interaction database (EMBL-EBI) 2026
Connected to