Oncofusion

start / graph / interface-aff1-rbm15-proximity

Interfaceweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q20-proximity-interactome

AFF1-RBM15 proximity: the m6A-complex adaptor, itself a paediatric leukaemia fusion partner, has AFF1 in its BioID neighbourhood

also called RBM15 BioID AFF1 · OTT1 proximity AFF1

RBM15 directs a chemical tag onto RNA and is itself broken and fused in a leukaemia of infants. Tagging RBM15 and listing its neighbours in a leukaemia cell turns up AFF1, KMT2A and DOT1L.

The same thing, for a scientist

RBM15 (Q96T37) is the adaptor that targets the METTL3/METTL14 m6A writer complex to specific transcripts, and **RBM15::MKL1 (OTT1-MAL) is the defining fusion of infant acute megakaryoblastic leukaemia**. As a BioID2 bait in K562 it recovers **529 preys**, among them **AFF1, CCNT1, KMT2A, DOT1L, PSIP1, KAT7, ICE2, SETD1A, SETD1B, HCFC1, CHD8, IGF2BP1, MSI2, HNRNPA2B1, YTHDC1 and SBDS** [E1]. It is one of AFF1's four proximity baits [E2]. The entry records this because it is a second, independent route into the m6A layer that this graph already holds - and because a proximity list containing both AFF1 and the m6A reader YTHDC1 is the kind of thing that will be mis-cited as an AFF1-m6A mechanism if it is not written down carefully.

Go deeper
Two paediatric fusion partners in one neighbourhood

RBM15 and AFF1 are both 3' partners of recurrent paediatric leukaemia fusions - RBM15::MKL1 in infant AMKL, KMT2A::AFF1 in infant B-ALL. That they are proximity neighbours in a third leukaemia line is a coincidence worth one sentence and no more; proximity in K562 says nothing about either fusion.

What is mechanically more interesting is the company: RBM15 recovers DOT1L, KAT7, SETD1A/B, HCFC1 and PSIP1 alongside AFF1 1. That is a chromatin-writer neighbourhood, not an RNA-processing one, and RBM15's own annotated job is RNA methylation targeting. Whether that means RBM15 works at chromatin or simply that everything nuclear is near everything nuclear is exactly what this dataset cannot decide 3.

What must not be concluded

This does not establish that the AFF1 transcript is m6A-modified, that AFF1 protein binds RBM15, or that the m6A machinery regulates the fusion. This graph already holds a separate and much better-supported finding that RNA-binding proteins of the KHDRBS and IGF2BP families control MLL-AF4 translation 4 - that is a functional result with mutant constructs and in-vivo transformation behind it. A proximity co-recovery is not in the same class of evidence and should never be cited beside it as if it were.

Retained region

Retained region: not determinable. AFF1 was a prey, never a bait, so the labelled molecule is endogenous full-length wild-type AFF1. Proximity labelling biotinylates accessible lysines anywhere on the prey and no site-level biotinylation was reported, so the record carries no positional information and cannot distinguish the AFF1 N-terminus that KMT2A::AFF1 discards from the C-terminal half it keeps.

Facts and figures
partner agene-aff1
partner bgene-rbm15
methodBioID2 proximity-dependent biotinylation in K562 (PSI-MI MI:1314), spoke-expanded
baitRBM15 (Q96T37)
preyAFF1 (P51825)
reciprocalfalse
reciprocity noteAFF1 was never a bait in any proximity experiment in IntAct, so no reciprocal recovery is possible.
druggabilityNone. No affinity, no interface, no evidence of direct contact.
n preys total529
interface residues a
[]
interface residues b
[]
structures
[]
How this could bite us back
  • A proximity co-recovery sitting in the same folder as functional RNA-layer results will be quoted alongside them. It is graded weak and its caveat says so, but grading is not protection against summarisation.
  • RBM15 is required for normal megakaryopoiesis; it is not a target.
Still open
  1. Is the AFF1 transcript m6A-modified, and does RBM15 target it? The proximity record cannot say; an RBM15 CLIP track over AFF1 would.
  2. RBM15 recovers a chromatin-writer neighbourhood (DOT1L, KAT7, SETD1A/B, HCFC1) rather than an RNA-processing one. Is that a real localisation of RBM15 to chromatin, or the dataset's nuclear background?
  3. Does RBM15::MKL1, the infant AMKL fusion, retain whatever puts RBM15 in this neighbourhood?
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    As a BioID2 bait in K562, RBM15 recovers 529 distinct preys including AFF1, CCNT1, KMT2A, DOT1L, PSIP1, KAT7, ICE2, SETD1A, SETD1B, HCFC1, CHD8, IGF2BP1, MSI2, HNRNPA2B1, YTHDC1 and SBDS.

    What this evidence cannot show

    Retrieved from IntAct 2026-09-19; the deposited pairs are the authors' curated bona fide subset. A 529-prey neighbourhood contains a slice of most nuclear processes. K562 carries BCR::ABL1 and no KMT2A rearrangement; the bait is an overexpressed BioID2 fusion.

    Evidence that argues against this (2)
    • In the largest human BioID survey (192 baits, HEK293) a prey's bait profile is dominated by subcellular compartment rather than by binding, so co-recovery of nuclear proteins by nuclear baits is the expected background. Go CD et al. 2021
    • BioID and affinity purification of chromatin-associated complexes recover FEW SHARED partners and enrich for different biological processes; a proximity hit is a different measurement from a co-purification hit and cannot be imported into a composition claim. Lambert JP et al. 2015
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    RBM15 is one of exactly four baits, out of fifty, whose neighbourhood contains AFF1; the others are NPM1, SBDS and ZRANB2.

    What this evidence cannot show

    Four of fifty is the 43.7th percentile of this dataset's baits-per-prey distribution and is not a specificity signal. The count is over the curated pair list, not the raw hit list.

    Evidence that argues against this (1)
    • The set of baits that reach AFF1 is determined by which fifty proteins the authors chose, all of them RNA-binding proteins selected for an m6A- and splicing-oriented study. A different fifty baits would produce a different four, and nothing about AFF1 is being sampled preferentially. Khoroshkin M et al. 2024
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Proximity biotinylation and affinity purification of chromatin-associated complexes recover few shared partners and enrich for distinct biological processes; BioID preferentially recovers lower-abundance proteins.

    PMID 25281560DOI
    What this evidence cannot show

    Methods comparison in HEK293 on histones and Mediator; cited to constrain interpretation rather than to support a biological claim.

    Evidence that argues against this (1)
    • The same study concludes both approaches recover biologically meaningful interactions and that BioID complements AP-MS, so it does not license discarding proximity results. Lambert JP et al. 2015
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    RNA-binding proteins of the KHDRBS and IGF2BP families control the oncogenic activity of MLL-AF4 by recognising AU-rich sites in its transcript and inhibiting its translation, with synonymous mutations that remove the sites restoring translation and producing transformation in vivo.

    PMID 36335100DOI
    What this evidence cannot show

    Murine retroviral system; abstract read, full text not read. Cited here as the contrast case - a functional RNA-layer result - not as support for anything about RBM15.

    Evidence that argues against this (1)
    • A second group reports IGF2BP3 LOWERING KMT2A::AFF1 protein abundance rather than raising it, so even the well-supported RNA-layer results disagree on sign; a proximity co-recovery has no standing beside either. IntAct molecular interaction database (EMBL-EBI) 2026
Connected to