Oncofusion

start / graph / gene-afg2b

Geneweak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.Q104-interactome-read-queue

AFG2B / SPATA5L1 (Q9BVQ7) - the second 55LCC ATPase, the only read-queue member with an annotated nuclear pool, pan-essential in 95.5% of lines, and the one that touches PSIP1

also called AFG2B · SPATA5L1 · AAA ATPase AFG2B · ATPase family gene 2 protein homolog B · DFNB119 gene · NEDHLS gene

AFG2B is the partner engine to AFG2A in a four-part machine that clears stuck proteins off DNA as it is copied, and losing it causes recessive deafness or a severe childhood brain disorder. One 2024 experiment placed it next to both KMT2A and LEDGF, two proteins this project cares about - but in cervical cancer cells, and it is the only read-queue member that UniProt even places in the nucleus.

The same thing, for a scientist

AFG2B (Q9BVQ7, 753 aa; HGNC-approved symbol AFG2B, previous official symbol **SPATA5L1**) is the second AAA+ ATPase of the **55LCC complex** (AFG2A/SPATA5, AFG2B/SPATA5L1, C1orf109, CINP), a chromatin-associated machine whose ATPase activity is specifically enhanced by replication fork DNA and which drives cysteine-protease-dependent cleavage of replisome substrates; loss causes ubiquitin-independent proteotoxicity, replication stress and severe chromosome instability [E1]. It reaches this project through exactly two IntAct rows, both from that paper: **KMT2A and PSIP1/LEDGF**, tandem affinity purification, **HeLa S3**, IntAct MI-score 0.35, AFG2B the bait and the scaffold protein the prey in both [E2]. **Two facts distinguish it from its partner and from the rest of the queue.** First, it is the **only member of the fifteen-gene read queue with any annotated nuclear localisation at all** - UniProt Q9BVQ7 carries a Nucleus comment, although only at ECO:0000250, by similarity to a rodent orthologue [E5]; its partner AFG2A has none. Second, its scaffold hit is **PSIP1**, the LEDGF protein that tethers the KMT2A complex to chromatin and which this graph treats as a dedicated axis - so between the two subunits the 55LCC purification touches KMT2A, MEN1 and PSIP1, three of this project's six scaffold proteins, in one experiment. **The complex-coherence test still fails**: C1orf109 and CINP are absent from the entire 217-partner union [E3]. Human genetics: biallelic variants cause **DFNB119** (autosomal recessive non-syndromic deafness) and **NEDHLS** (neurodevelopmental disorder with hearing loss and spasticity) [E5]. DepMap 26Q2: mean gene effect **-1.3319** across 1,258 screens, **95.5% of lines below -0.5**, and the leukaemia mean is **-1.4064**, i.e. slightly MORE negative than the panel - pan-essential [E4]. `SPATA5L1[tiab] AND leukemia[tiab]` = 0, a parsed zero, and the symbol-generation trap recurs: `AFG2B[tiab]` = 2 against `SPATA5L1[tiab]` = 14 [E6].

What it looks like
loading AF-Q9BVQ7-F1…
AlphaFold AF-Q9BVQ7-F1 ↗
colourshape

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

very high (90+)confident (70–90)low (50–70)very low (<50)
I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
The two rows, and the PSIP1 one is the interesting half
sourcescaffoldmethodhostrole of AFG2BMI-score
PMID 38554706KMT2A (Q03164)tandem affinity purification (MI:0676)HeLa S3bait0.35
PMID 38554706PSIP1/LEDGF (O75475)tandem affinity purification (MI:0676)HeLa S3bait0.35

One publication, one purification, two prey hits. The read-queue criterion that admitted this gene - two or more baits - is satisfied by two prey proteins in a single experiment, not by replication.

PSIP1 is why this row is worth a second look. LEDGF/p75 is the chromatin tether for the menin-KMT2A complex and has its own dedicated beat in this graph (ledgf-axis). Combined with the AFG2A rows against KMT2A and MEN1, one HeLa S3 TAP recovered KMT2A, MEN1 and PSIP1 - three of the six scaffold proteins this project tracks - with the two 55LCC ATPases as baits.

And the alternative explanation is the obvious one. KMT2A, menin and LEDGF are a physically associated module. A bait that touches one of them in a nuclear extract can recover all three without touching any of them specifically, because they come out together. So "three of six scaffolds" is close to "one module", and the number is less impressive than it looks. Recording it that way is the honest version.

The nuclear pool: the only one in the queue, and it is uncoded

This matters because the queue's rejections turn on it. The cycle-54 adjudication rejected CAMKV and EPB41L1 substantially on the ground that neither has an experimentally established nuclear pool, and the same test was applied to every member of this queue.

AFG2B is the only one that passes even nominally. UniProt Q9BVQ7 location comments 5:

locationevidence
CytoplasmECO:0000269, PubMed 18445686
Cytoplasm, cytoskeleton, spindleECO:0000269, PubMed 18445686
NucleusECO:0000250 - by similarity to D4A2B7, a rodent orthologue

So the nuclear assignment is an inference from a rodent protein, not a human measurement, and the two experimentally coded locations are both cytoplasmic and both from a 2008 paper that this beat did not open and which predates the complex's discovery by sixteen years. Per rule 91 an evidence code establishes that a curator read an experiment, not that the paper is about the fact.

Compare the partner. AFG2A (Q8NB90) has no nuclear location comment at all - cytoplasm and spindle coded, mitochondrion by similarity. Two obligate subunits of one complex, one placed in the nucleus by similarity and the other not placed there at all, while the FUNCTION comment on both calls the complex chromatin-associated. That is an internal inconsistency in the database, not in the biology, and it is worth reporting as a defective record in the direction of under-annotation: a 2024 Cell paper established a chromatin-associated function for both and neither location comment was updated.

Human genetics: two distinct disorders from one gene

UniProt records two disease entries on Q9BVQ7 5:

  • DFNB119 - deafness, autosomal recessive 119: non-syndromic sensorineural hearing loss, mild to profound.
  • NEDHLS - neurodevelopmental disorder with hearing loss and spasticity: autosomal recessive, with hearing loss, global developmental delay, impaired intellectual development, hypotonia, spastic-dystonic cerebral palsy, focal or generalised epilepsy and microcephaly.

The pair is informative on its own. One gene producing both an isolated sensory phenotype and a severe multisystem neurodevelopmental one is the signature of an allelic series: hypomorphic alleles give deafness, more severe alleles give the encephalopathy. That matters for any intervention argument, because it says partial inhibition of this protein has a known human consequence and it is hearing loss.

And the parallel with AFG2A is exact. AFG2A's disorder (NEDHSB / AFG2A-related encephalopathy) has hearing loss in 93.62% of 51 reported individuals alongside microcephaly and drug-resistant epilepsy. Both subunits of 55LCC therefore converge on a deafness-plus-encephalopathy phenotype, which is independent genetic evidence that the two proteins work in one pathway - stronger evidence for the complex than the interaction data is, and it does not involve HeLa cells.

Why this is not a target

DepMap 26Q2, 1,258 screens: mean gene effect -1.3319, minimum -2.856, 95.5% of lines below -0.5; leukaemia mean -1.4064 over 57 screens 4.

AFG2B is the more essential of the two subunits by a clear margin (-1.33 against AFG2A's -0.83), and it is the third most essential gene this beat measured, behind RPS11 (-2.84) and ahead of PES1 (-1.02). The leukaemia mean is 0.07 more negative than the panel mean, which is a difference of no consequence at this magnitude and is reported only so that nobody later reads it as selectivity.

The verdict is the standard one for an essential machine: pan-essential, no window, and the human loss-of-function phenotype is deafness at the mild end and a spastic encephalopathy at the severe end.

Facts and figures
uniprotQ9BVQ7
length AA753
PDB noteBoth PDB cross-references are shared with AFG2A (Q8NB90) and are presumably the 55LCC structures. NEITHER WAS FETCHED, so no chain, resolution, residue range or interface is claimed from them.
alphafold IDAF-Q9BVQ7-F1
domains noteUniProt Q9BVQ7's feature table carries NO Domain features, so no AAA+ ATPase boundary is annotated and none is invented. The lane is deliberately EMPTY, per the integer-residue rule.
accession resolutionResolved live 2026-09-22: gene_exact:AFG2B AND organism_id:9606 AND reviewed:true returns EXACTLY ONE reviewed row - Q9BVQ7, AFG2B_HUMAN, gene_primary AFG2B, synonym SPATA5L1, 753 aa. The migration SPATA5L1 -> AFG2B is confirmed at UniProt, HGNC (fetch/symbol/AFG2B returns ENSG00000171763, 'AAA ATPase AFG2B') and Open Targets (approvedSymbol AFG2B). LENGTH DISCRIMINATION NOTE: AFG2A is 893 aa and AFG2B is 753 aa, 140 residues apart, so verify-accessions.py CAN distinguish a swap between the two subunits - unusually for a paralogue pair in this graph.
localizationCytoplasm and cytoplasm/cytoskeleton/spindle, both ECO:0000269 from PubMed 18445686 (2008, not opened by this beat); Nucleus, ECO:0000250 ONLY - by similarity to rodent D4A2B7. AFG2B is the ONLY member of the fifteen-gene read queue with any annotated nuclear localisation, and it is by similarity rather than measured. The complex is called chromatin-associated in the FUNCTION comment, citing the 2024 paper, which the location comments do not reflect [E5].
subnuclear localizationUNPUBLISHED, and the prior question is only weakly settled: the nuclear pool itself is ECO:0000250 by similarity to a rodent orthologue [E5]. The nearest positional statement available is functional - 55LCC acts on the replisome and its ATPase is specifically stimulated by replication fork DNA [E1] - which implies replication-fork chromatin rather than a subnuclear body. That implication is recorded as an inference from function, NOT a measurement: no replication-focus, nuclear-body or chromatin-domain co-localisation has been reported for this protein in any cell.
cell typeHeLa S3 (epithelioid cervical carcinoma) for both IntAct rows [E2]; unspecified human cell lines for the 55LCC structural and functional work [E1]; human patients for the two disease entities, which are auditory and neurological [E5]. NO haematopoietic cell, no B-lineage cell, no infant cell, no primary leukaemia sample and no KMT2A::AFF1-expressing cell anywhere in the record.
cell stateS phase and replication stress: 55LCC processes replisome substrates in S phase and its ATPase is stimulated by fork DNA [E1]. Also mitotic - a spindle localisation is coded from a 2008 paper [E5]. NOT assayed in any differentiation state and specifically not in the differentiation-blocked state this project's fusion imposes.
epigenetic readoutNONE DIRECTLY. AFG2B writes, reads and erases no histone modification and has no reader module. Its chromatin connection is the 'chromatin-associated' designation of 55LCC and the replisome as substrate [E1][E5]. Recorded explicitly: the epigenetics dimension here is replication-coupled chromatin proteostasis, not a mark.
shown for fusionNO - nobody has looked in any KMT2A fusion, in any system. Both IntAct rows are against wild-type KMT2A and wild-type PSIP1 in HeLa S3 with AFG2B as bait [E2]; AFG2B/SPATA5L1 is absent from the Skucha seven-fusion AP-MS network; and SPATA5L1[tiab] AND leukemia[tiab] is a parsed zero [E6].
intact independenceONE PUBLICATION, ONE EXPERIMENT. Both rows are prey hits from the same HeLa S3 tandem affinity purification.
PDB IDS
[
  "8RHN",
  "9LGO"
]
domains
[]
complex membership
{
  "complex": "55LCC - AFG2A (SPATA5) + AFG2B (SPATA5L1) + C1orf109 + CINP; heterohexameric, chromatin-associated",
  "function": "replisome proteostasis; ATPase enhanced by replication fork DNA, coupled to cysteine-protease-dependent cleavage of replisome substrates; loss causes ubiquitin-independent proteotoxicity, replication stress and severe chromosome instability [E1]",
  "second_function": "cytoplasmic maturation of pre-60S ribosomal particles, promoting release of RSL24D1/RLP24 (UniProt, cited to PubMed 35354024; not independently verified here) - the likely explanation for the -1.33 versus -0.83 essentiality asymmetry between the two subunits",
  "coherence_test": "FAILS. Two of four subunits are in the 217-partner union; C1orf109 and CINP are absent entirely and neither has a node [E3].",
  "other_subunits_in_this_graph": "gene-afg2a (written by this beat). gene-c1orf109 and gene-cinp do not exist.",
  "scaffolds_touched_across_both_subunits": "KMT2A and MEN1 via AFG2A, KMT2A and PSIP1 via AFG2B - three of this project's six scaffold proteins in ONE HeLa S3 purification. Discounted in the body: KMT2A, menin and LEDGF are one physically associated module, so three hits is close to one."
}
intact rows
[
  {
    "scaffold": "KMT2A",
    "scaffold_uniprot": "Q03164",
    "pmid": "38554706",
    "method": "tandem affinity purification (MI:0676)",
    "host": "HeLa S3",
    "role_of_this_gene": "bait",
    "role_of_scaffold": "prey",
    "intact_miscore": 0.35
  },
  {
    "scaffold": "PSIP1",
    "scaffold_uniprot": "O75475",
    "pmid": "38554706",
    "method": "tandem affinity purification (MI:0676)",
    "host": "HeLa S3",
    "role_of_this_gene": "bait",
    "role_of_scaffold": "prey",
    "intact_miscore": 0.35
  }
]
human genetics
{
  "disorders": [
    {
      "name": "DFNB119 - deafness, autosomal recessive 119",
      "description": "non-syndromic sensorineural hearing loss, mild to profound",
      "source": "UniProt disease comment on Q9BVQ7; no clinical paper fetched"
    },
    {
      "name": "NEDHLS - neurodevelopmental disorder with hearing loss and spasticity",
      "description": "autosomal recessive; hearing loss, global developmental delay, impaired intellectual development, hypotonia, spastic-dystonic cerebral palsy, focal or generalised epilepsy, microcephaly",
      "source": "UniProt disease comment on Q9BVQ7; no clinical paper fetched"
    }
  ],
  "interpretation": "THIS BEAT'S INFERENCE, NOT A PUBLISHED CLAIM: one gene giving both an isolated sensory phenotype and a severe multisystem one is the signature of an allelic series, with hypomorphs producing deafness. Flagged as inference because no paper was read for either disorder.",
  "convergence_with_partner": "AFG2A's disorder (NEDHSB / AFG2A-related encephalopathy) has hearing loss in 93.62% of 51 individuals with microcephaly and drug-resistant epilepsy. Both 55LCC ATPases converge on deafness-plus-encephalopathy, which is independent genetic evidence that the two work in one pathway - and it does not involve HeLa cells, so it is stronger evidence for the complex than the interaction data is.",
  "cancer": "No malignancy reported for either disorder in the records read here."
}
depmap
{
  "release": "26Q2 (meta.dataVersion year 26 month 06)",
  "fetched": "2026-09-22",
  "screens": 1258,
  "mean_gene_effect": -1.3319,
  "min_gene_effect": -2.856,
  "fraction_below_minus_0_5": 0.955,
  "leukaemia_screens": 57,
  "leukaemia_mean_gene_effect": -1.4064,
  "partner_comparison": "AFG2A -0.8275; AFG2B is 0.50 more negative, a large asymmetry for two obligate subunits of one complex",
  "verdict": "pan-essential; the leukaemia-versus-panel difference of 0.07 at this magnitude is of no consequence and must not be read as selectivity"
}
bounded literature
{
  "queried": "2026-09-22",
  "names_tried": [
    "AFG2B",
    "SPATA5L1",
    "55LCC",
    "leukemia"
  ],
  "not_tried": [
    "SPATA5L1 paired with KMT2A",
    "SPATA5L1 paired with MLL",
    "C1orf109",
    "CINP",
    "MLL-AF4",
    "MLL-AF9",
    "\"t(4;11)\""
  ],
  "counts": {
    "AFG2B[tiab]": 2,
    "SPATA5L1[tiab]": 14,
    "\"55LCC\"[tiab]": 1,
    "SPATA5L1[tiab] AND leukemia[tiab]": 0
  },
  "querytranslations": {
    "SPATA5L1[tiab] AND leukemia[tiab]": "\"SPATA5L1\"[Title/Abstract] AND \"leukemia\"[Title/Abstract]"
  },
  "warninglist_on_every_zero": [
    "No items found."
  ],
  "symbol_generation_recall": "AFG2B retrieves 2 of the 14 records SPATA5L1 retrieves (14.3%). Both AFG2B ids are a subset of the SPATA5L1 set, so the current symbol is impoverished but NOT contaminated - unlike AFG2A, where 4 of 7 records are aflatoxin papers. And unlike AFG2A, AFG2B DOES retrieve the source paper 38554706.",
  "acknowledged_gap": "SPATA5L1 was not paired with KMT2A or MLL by this beat; the partner gene's parsed zeros were carried across, which is an inference across genes and is recorded as such in E6's counter-evidence.",
  "bounded_by": "titles and abstracts only"
}
identifier traps
[
  "AFG2B[tiab] = 2 records against SPATA5L1[tiab] = 14 (14.3% recall). Impoverished, not contaminated: both AFG2B ids are a subset of the SPATA5L1 set.",
  "Its partner AFG2A behaves DIFFERENTLY - AFG2A[tiab] is 57% aflatoxin papers and does NOT retrieve the source paper, while AFG2B does. Two subunits of one complex, two different symbol failure modes; a beat must test each symbol separately rather than assuming the complex behaves as one.",
  "Length discrimination WORKS for this pair: AFG2A 893 aa versus AFG2B 753 aa, 140 residues apart, so an accession swap between the two subunits is catchable by verify-accessions.py - noted because it is unusual in this graph.",
  "No row for AFG2B or SPATA5L1 exists in data/symbol-collisions.json (147 rows, 2026-09-22), checked before the first query. NEW trap."
]
How this could bite us back
  • AFG2B is required in 95.5% of DepMap lines with a mean gene effect of -1.33, and leukaemia lines are if anything slightly more dependent. There is no therapeutic window; an inhibitor would be a general anti-proliferative.
  • The human loss-of-function phenotypes are recessive deafness at the mild end and a spastic encephalopathy with epilepsy and microcephaly at the severe end. Partial systemic inhibition has a known human consequence and it is hearing loss - a particularly poor trade in an infant population.
  • Inhibiting replisome proteostasis raises replication stress and chromosome instability. In a disease caused by a translocation, that is a mechanistically direct route to further rearrangement and to therapy-related second malignancy.
  • The protein carries a second essential function in pre-60S ribosome maturation, separate from the replisome one, so any inhibitor has two independent essential-pathway liabilities and no existing tool separates them.
Still open
  1. Run SPATA5L1[tiab] AND KMT2A[tiab] and SPATA5L1[tiab] AND MLL[tiab]. This beat carried its partner gene's zeros across instead of running them, and that is the one gap in this entry's bounded absence.
  2. Do C1orf109 and CINP co-purify with PSIP1, KMT2A or menin? The 55LCC coherence test fails on two of four subunits and this is the measurement that would settle whether the rows record a complex or two loose ATPases in a TAP eluate - and it should be run against IntAct REST, which returns 114 rows for PSIP1 where the instrument used here returns 50.
  3. Why is AFG2B 0.50 more essential than AFG2A if they are obligate partners? UniProt's separate pre-60S maturation function is the likely answer and nothing in this graph's interaction record reflects it.
  4. Does the LEDGF/PSIP1 tether have a replication-coupled function? PSIP1 is the only scaffold protein AFG2B touches that is not KMT2A itself, PSIP1 has its own beat here, and a LEDGF-replisome connection would be a genuinely new axis rather than a HeLa artefact.
  5. Is there a KMT2A-rearranged-specific replication-stress phenotype that a 55LCC hypomorph modifies? The cheapest route from a HeLa co-purification to a leukaemia-relevant result, and unattempted.
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The 55LCC complex comprises the AAA+ ATPases SPATA5 (AFG2A) and SPATA5L1 (AFG2B) together with the heterodimer C1orf109-CINP; the ATPases' N-terminal domains interact with C1orf109-CINP to form a funnel-like structure above a cylindrical ATPase motor; 55LCC deficiency elicits ubiquitin-independent proteotoxicity, replication stress and severe chromosome instability; and its ATPase activity is specifically enhanced by replication fork DNA and coupled to cysteine-protease-dependent cleavage of replisome substrates in response to fork damage.

    PMID 38554706DOI
    What this evidence cannot show

    Read from the abstract only; the structural architecture, the fork-DNA stimulation and the protease coupling are the abstract's own statements, and no figure or PDB entry was opened. The two PDB cross-references on Q9BVQ7 (8RHN, 9LGO, shared with Q8NB90) were NOT fetched, so no chain, resolution, residue range or interface is claimed. SYSTEM: human cell lines including the HeLa S3 TAP that produced this entry's IntAct rows; no haematopoietic cell, no leukaemia, no fusion. NOMENCLATURE: the paper is titled with the PREVIOUS official symbols, which is why AFG2B[tiab] retrieves only 2 records where SPATA5L1[tiab] retrieves 14 - see E6.

    Evidence that argues against this (1)
    • The paper's stated clinical rationale is human neurodevelopmental disorders, not cancer. A chromosome-instability phenotype in cells that produces a non-neoplastic developmental syndrome in humans is a caution against the inference a leukaemia project would reach for, namely that 55LCC loss should promote translocation formation. Nou-Fontanet L et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    IntAct holds exactly two rows placing AFG2B (Q9BVQ7) against this project's scaffold proteins - one against KMT2A (Q03164) and one against PSIP1/LEDGF (O75475) - both from PMID 38554706, both tandem affinity purification (MI:0676), both in HeLa S3, both at IntAct MI-score 0.35, and in both rows the experimental role of AFG2B is bait while the scaffold protein is prey.

    What this evidence cannot show

    Queried live 2026-09-22; experimental role, host, detection method and MI-score read out of MITAB2.7 columns 19/20, 29, 7 and 15 rather than assumed. ONE PUBLICATION, ONE EXPERIMENT: both rows are prey hits from the same HeLa S3 purification, so the >=2-baits criterion that admitted this gene to the read queue is satisfied by two prey proteins in one experiment. MI-score 0.35, below IntAct's 0.6 conventional high-confidence cut. HOST HeLa S3: epithelioid cervical carcinoma, not haematopoietic, not B-lineage, no fusion. DIRECTION: AFG2B is the bait in both rows, so neither KMT2A nor PSIP1 has ever been used as bait in an experiment that recovered it. Rule 93 endpoint caveat: PSICQUIC returns 218 rows for KMT2A and 50 for PSIP1 against 376 and 114 from IntAct REST - PSIP1 is the worst-served bait of the six at 44%, so this count is a floor and a poor one.

    Evidence that argues against this (1)
    • KMT2A, menin and LEDGF/PSIP1 form a physically associated module, so a bait that contacts any one of them in a nuclear extract can recover all three without specificity for any. 'Three of six scaffold proteins across the two 55LCC subunits' therefore reduces close to 'one module', and should not be read as three independent associations. Krishnamoorthy V et al. 2024
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Neither C1orf109 nor CINP, the two heterodimeric partners that complete the 55LCC complex, appears anywhere in the 217-partner IntAct union of KMT2A, MEN1, PSIP1, AFF4, DOT1L and AFF1 - not among the 95 with their own entry, not among the 10 mentioned-only, and not among the 112 absent entirely - and no gene-c1orf109 or gene-cinp node exists in this graph.

    What this evidence cannot show

    Checked by this beat 2026-09-22 two ways: the 122-member all_missing list was searched for C1orf109 and CINP (absent), and the complete 2,019-id node listing was searched for gene-c1orf109 and gene-cinp (absent). THIS IS NOT a claim that those proteins do not bind these scaffolds; it is a claim about what the curated record contains, and IntAct curation of one paper reflects the depositors' submission. The endpoint is PSICQUIC, which returns 651 of the 1,015 rows IntAct REST gives for the same six accessions (64%), so a C1orf109 or CINP row could exist and be invisible here. The coherence test should be re-run on REST before it is treated as settled.

    Evidence that argues against this (1)
    • PSIP1 is the worst-served bait of the six under this endpoint - 50 PSICQUIC rows against 114 from REST, 44% - and PSIP1 is precisely the scaffold this entry's second row is against. So for this gene specifically, more than half of the relevant record is outside the instrument, and both the positive row count and the negative coherence test are drawn from the weaker half. EMBL-EBI IntAct 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    In DepMap release 26Q2 as served by Open Targets, AFG2B (ENSG00000171763) has a mean gene effect of -1.3319 across 1,258 screens, a minimum of -2.856, 95.5% of lines below -0.5, and a mean of -1.4064 across the 57 leukaemia-annotated screens - making it substantially more essential than its complex partner AFG2A (-0.8275).

    What this evidence cannot show

    RELEASE RECORDED per rule 83: meta.dataVersion = {year: 26, month: 06} = DepMap 26Q2, fetched 2026-09-22. 1,258 is this gene's own screen count. All 29 tissue blocks concatenated before averaging. Ensembl id resolved through the HGNC primary ensembl_gene_id per rule 78, readback control passed (Open Targets returned approvedSymbol AFG2B). The leukaemia subset is a substring match on diseaseFromSource, NOT stratified by KMT2A rearrangement. The leukaemia-versus-panel difference of 0.07 at a mean of -1.33 is of no consequence and is reported only to prevent it being read as selectivity later. No median-split proportion is reported, per rule 84.

    Evidence that argues against this (1)
    • The subunit asymmetry is itself a caution against reading 55LCC as a single unit in DepMap: -1.33 for AFG2B against -0.83 for AFG2A means the two obligate partners of one complex have materially different requirement profiles, which usually indicates that at least one of them has an essential function outside the complex. UniProt assigns both a separate pre-60S ribosome-maturation role, which is the likely explanation and which no interaction row in this graph reflects. UniProt Knowledgebase 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    UniProt resolves gene_exact:AFG2B in reviewed human records to exactly one entry, Q9BVQ7 (AFG2B_HUMAN, 753 aa, 'ATPase family gene 2 protein homolog B'), whose gene-names field carries the synonym SPATA5L1; its subcellular-location comments are Cytoplasm (ECO:0000269, PubMed 18445686), Cytoplasm/cytoskeleton/spindle (ECO:0000269, PubMed 18445686) and Nucleus (ECO:0000250, by similarity to D4A2B7) - the nuclear assignment being by similarity only; and it carries two disease comments, DFNB119 (autosomal recessive non-syndromic sensorineural deafness, mild to profound) and NEDHLS (neurodevelopmental disorder with hearing loss and spasticity, autosomal recessive, with global developmental delay, impaired intellectual development, hypotonia, spastic-dystonic cerebral palsy, focal or generalised epilepsy and microcephaly).

    What this evidence cannot show

    Fetched live 2026-09-22. THE NUCLEAR COMMENT IS ECO:0000250 - BY SIMILARITY TO A RODENT ORTHOLOGUE (D4A2B7), NOT A HUMAN MEASUREMENT, and per rule 91 that is the weakest annotation class. The two experimentally coded locations are both cytoplasmic and both from PubMed 18445686, a 2008 paper this beat did NOT open and which predates the 55LCC complex's discovery by sixteen years. NOT VERIFIED HERE: the pre-60S maturation function (cited by UniProt to PubMed 35354024, not opened); the disease descriptions were read as UniProt's own text and no clinical paper for DFNB119 or NEDHLS was fetched, so the disorders' existence and descriptions rest on the database record alone and the allelic-series interpretation in the body is THIS BEAT'S INFERENCE from the two entries, not a published claim. The feature table has NO Domain features so no AAA+ boundary is asserted.

    Evidence that argues against this (1)
    • The record is internally inconsistent and the inconsistency runs against the interaction data: the FUNCTION comment calls 55LCC chromatin-associated, citing the 2024 paper, while the only experimentally coded locations are cytoplasm and spindle from 2008. Its obligate partner AFG2A has no nuclear comment at all. Two subunits of one complex annotated differently on the central question is a database defect in the direction of under-annotation, and the nuclear pool that the KMT2A and PSIP1 rows presuppose has never been measured in a human cell. UniProt Knowledgebase 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PubMed esearch 2026-09-22: AFG2B[tiab] returns 2 records against SPATA5L1[tiab] returning 14 - a 14.3% recall for the current symbol - and SPATA5L1[tiab] AND leukemia[tiab] returns 0, a parsed zero; both AFG2B records are also in the SPATA5L1 set, so unlike AFG2A the current AFG2B symbol is not contaminated, only impoverished, and it does retrieve the source paper 38554706.

    What this evidence cannot show

    The zero carries a non-empty querytranslation - `"SPATA5L1"[Title/Abstract] AND "leukemia"[Title/Abstract]` - and warninglist.outputmessages = ["No items found."]. Each operand separately tagged, no `(A OR B)[tiab]` form, per rule 94. The id sets for AFG2B (41390732, 37902276) and SPATA5L1 (14 records) were compared directly and AFG2B's two are a subset - that is the basis of the 'impoverished not contaminated' claim and it is a complete comparison, not a sample. NOT TRIED, and this is the material gap: SPATA5L1 was NOT paired with KMT2A or with MLL. Only the leukemia pairing was run for this gene, on the reasoning that its partner AFG2A returned parsed zeros for both KMT2A and MLL from the same single source paper - which is an inference across genes, not a measurement, and a beat extending this entry should run SPATA5L1[tiab] AND KMT2A[tiab] and SPATA5L1[tiab] AND MLL[tiab] directly. Bounded by titles and abstracts only. data/symbol-collisions.json (147 rows) grepped for AFG2B and SPATA5L1 before the first query per rule 88: no row, so this is a new trap.

    Evidence that argues against this (1)
    • This beat did not run SPATA5L1 against KMT2A or MLL and carried the partner gene's zeros across instead. That is exactly the shortcut this project's rules warn about, it is recorded here rather than hidden, and until it is run the claim that no literature links AFG2B to KMT2A rests on an inference from a different gene. NCBI E-utilities 2026
Connected to