Oncofusion

start / graph / ana-aff1-proximity-neighbourhood

Analysismoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q20-proximity-interactome

AFF1's modern proximity neighbourhood is four RNA-binding proteins - NPM1, RBM15, SBDS, ZRANB2 - and it barely overlaps AFF4's

also called AFF1 BioID neighbourhood · AFF1 proximity partners

The first modern map of what AFF1 sits next to inside a living human cell finds four neighbours, all of them proteins that handle RNA. Its close relative AFF4 has four neighbours too, and only one of them is the same.

The same thing, for a scientist

In the 50-bait BioID2 map of K562, AFF1 (P51825) is recovered as prey by **four baits: NPM1, RBM15, SBDS and ZRANB2** [E1]. AFF4 (Q9UHB7) is recovered by four as well - **GPKOW, HNRNPC, PPIL4 and ZRANB2** - and the two sets share **only ZRANB2** [E2]. Both counts sit at the **43.7th percentile** of the dataset's baits-per-prey distribution, so neither is a specific hit and neither is unusually promiscuous; the interesting quantity is the **non-overlap**, not the size [E1]. All four AFF1 baits are RNA-binding proteins with splicing, m6A-deposition, ribosome-maturation or nucleolar functions. **None of this can be mapped to a region of AFF1 that the fusion retains**, because AFF1 was never a bait and the data carry no positional information.

Go deeper
The four, and what each bait is
baitUniProtwhat it doesalso recovers
ZRANB2O95218zinc-finger splicing regulatorAFF4, ELL, MLLT1, CCNT1, KMT2A, DOT1L, BRD4, WDR5, NELFE, POLR2A
NPM1P06748nucleophosmin; nucleolar chaperone, itself a leukaemia driverELL, MLLT1, CCNT1, KMT2A, MEN1, DOT1L, WDR5, BRD4
RBM15Q96T37m6A methyltransferase-complex adaptor; the RBM15::MKL1 fusion drives infant AMKLCCNT1, KMT2A, DOT1L, PSIP1, KAT7
SBDSQ9Y3A560S ribosome maturation; mutated in Shwachman-Diamond syndromeCDK9, KMT2A, PSIP1, WDR5, PAF1, ICE1

Each of the four has its own entry or interface entry in this folder 1.

AFF1 and AFF4 do not share a neighbourhood, and that is the finding

AFF1 and AFF4 are the two mutually exclusive scaffolds of the super elongation complex. If a single SEC particle were the dominant species in K562, a bait that labels one scaffold should usually label the other. It does not. Four baits reach AFF1, four reach AFF4, and one bait reaches both 2.

That pattern is what the SEC-family literature predicts - AFF1-containing and AFF4-containing complexes are distinct particles with distinct partners - and it is the first time this graph has evidence for it from a method that does not lyse the cell first. It is also the cleanest internal control available in this dataset: two proteins of comparable abundance and identical compartment, sampled by the same 50 baits, come back with disjoint profiles. That is difficult to explain by abundance alone, which is the standing objection to every proximity result.

But it is four against four. With a median of five baits per prey and 2,682 preys, the probability of a one-element overlap between two random four-element subsets of fifty is not small, and no statistical test is offered here because the sampling is not random. Treat this as a hypothesis with a stated test, not a result 2.

What module do the authors assign AFF1 to? None, and the reason matters

The study's headline product is an assignment of each RBP to a regulatory module with an annotated function and target regulon. Those assignments are computed for the 50 baits, using integrated distances across BioID, Perturb-seq and eCLIP 3.

AFF1 is a prey. It has no module, no Perturb-seq perturbation and no eCLIP track in this study. The question the dispatch asked - what module do the authors put AFF1 in - has the answer they do not put it in one, and any statement to the contrary would be reading the bait table as if it contained preys.

Retained region

Can this be mapped to a region the fusion retains? No, and the reason is structural, not a gap in the reading. AFF1 was a PREY in this experiment, not a bait. What was labelled is endogenous, full-length, wild-type AFF1 in a cell with no KMT2A rearrangement. Proximity labelling biotinylates accessible lysines anywhere on the prey and the mass spectrometry was not run for site-level biotinylation, so the record carries no positional information at all. Nothing here distinguishes a contact with the AFF1 N-terminus, which KMT2A::AFF1 discards, from one with the C-terminal half it keeps. Any entry that infers otherwise is over-reading the data.

Facts and figures
scriptnone (database query)
statusrun
aff1 uniprotP51825
aff1 n baits4
aff4 uniprotQ9UHB7
aff4 n baits4
percentile of prey promiscuity43.7
aff1 experimental roleprey (MI:0498) in all four rows
result summaryAFF1's proximity neighbourhood in K562 is NPM1, RBM15, SBDS and ZRANB2. AFF4's is GPKOW, HNRNPC, PPIL4 and ZRANB2. One bait in common. Neither count is specific in this dataset; the non-overlap is the observation worth testing.
inputs
[
  "IntAct PSICQUIC identifier:P51825 AND pubid:39251607",
  "IntAct PSICQUIC pubid:39251607 (full deposition)"
]
aff1 baits
[
  "NPM1",
  "RBM15",
  "SBDS",
  "ZRANB2"
]
aff4 baits
[
  "GPKOW",
  "HNRNPC",
  "PPIL4",
  "ZRANB2"
]
shared baits
[
  "ZRANB2"
]
How this could bite us back
  • The four named proteins are the four the 50-bait design happened to reach. Any of AFF1's real partners that no RBP bait sits near is invisible here, and the invisible set is much larger than the visible one.
  • Nominating any of these four as a target on the strength of a proximity hit would be nominating a neighbour. None of the four has been shown to bind AFF1, let alone to bind the fusion.
Still open
  1. Does the AFF1/AFF4 non-overlap survive in the raw PXD041608 hit lists before the bona fide filter, where both proteins should appear in more neighbourhoods?
  2. Is the non-overlap explained by abundance? Measuring AFF1 and AFF4 copy number in K562 would settle whether the comparison is between two proteins of comparable detectability.
  3. Do any of the four AFF1 baits touch the AFF1 region KMT2A::AFF1 retains? Only a reciprocal experiment with AFF1 as bait, and truncation constructs, can answer this.
  4. Is ZRANB2 - the only bait reaching both scaffolds - reaching the shared P-TEFb module rather than either scaffold, given that it also recovers CDK9's partner CCNT1?
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    In the K562 50-bait BioID2 deposition (IMEx IM-30059), AFF1 (P51825) is recovered as prey by exactly four baits - NPM1, RBM15, SBDS and ZRANB2 - each row PSI-MI MI:1314 with AFF1 in experimental role MI:0498 (prey). Four baits places AFF1 at the 43.7th percentile of the dataset's baits-per-prey distribution (median 5, mean 7.23, max 40, n = 2,682 preys).

    What this evidence cannot show

    Retrieved from IntAct 2026-09-19; these are the authors' curated bona fide pairs, not the raw hit list, so the true count before filtering may be higher. K562 carries BCR::ABL1 and no KMT2A rearrangement. Baits are overexpressed BioID2 fusions. Four of fifty is an unremarkable recovery rate in this dataset and must not be read as specificity.

    Evidence that argues against this (2)
    • In the largest published human BioID survey (192 baits, HEK293) a prey's bait profile is dominated by subcellular compartment rather than by binding. A nuclear elongation factor recovered by nuclear RNA-processing baits is what shared compartment occupancy alone predicts. Go CD et al. 2021
    • A head-to-head comparison of BioID against an affinity-purification protocol optimised for chromatin-associated complexes found the two identified FEW SHARED partners and enriched for distinct biological processes. A proximity hit is a different measurement from a co-purification hit, not a weaker one, and cannot be imported into a complex-composition claim. Lambert JP et al. 2015
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    AFF4 (Q9UHB7) is recovered by four baits in the same deposition - GPKOW, HNRNPC, PPIL4 and ZRANB2 - sharing only ZRANB2 with AFF1's four, despite AFF1 and AFF4 being the alternative scaffolds of the same complex family and being sampled by the same fifty baits in the same cells.

    What this evidence cannot show

    Four against four is a small comparison and no significance test is appropriate, because the baits are not a random sample of the proteome and the curated pair list has an unstated filtering threshold. The observation is a hypothesis with a stated test, not a result. Relative abundance of AFF1 and AFF4 in K562 was not measured here.

    Evidence that argues against this (2)
    • The AFF4 C-terminal homology domain both homodimerises and HETERODIMERISES with AFF1, AFF2 and AFF3. If AFF1-AFF4 heterodimers exist in these cells, the two scaffolds are sometimes in one particle and a disjoint proximity profile cannot mean they never co-occur. Chen Y et al. 2019
    • Mutual exclusivity is also the null for a small sample: with 50 baits, median 5 baits per prey and 2,682 preys, two independent four-element subsets overlap in one element often enough that the observation is compatible with no biological structure at all. IntAct molecular interaction database (EMBL-EBI) 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The study computes regulatory-module membership and functional annotation for its fifty BAITS by integrating BioID, Perturb-seq and ENCODE eCLIP distances; preys receive no module assignment, no perturbation and no eCLIP track, so AFF1 has no module in this study.

    PMID 39251607DOI
    What this evidence cannot show

    Full text read from PMC11385195. This is a statement about the study's design, not about AFF1's biology; the absence of a module is an absence of testing, not a negative result about AFF1's function.

    Evidence that argues against this (1)
    • An absence of testing is not a negative result. AFF1 has substantial RNA-associated biology of its own through the super elongation complex's control of pausing release, and a study that never perturbed it cannot be cited as evidence that it belongs to no regulatory module. Gu J et al. 2014
Connected to