Oncofusion

start / graph / ana-the-six-open-queue-members-split-three-and-three-and-the-dependency-data-was-wrong-about-which

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.

The six open read-queue members split three rejected and three earning entries - and the flattest dependency in the set, ASB7, turns out to be a negative regulator of H3K9me3 via SUV39H1 degradation

also called six open queue members adjudicated · ASB7 degrades SUV39H1 · ASB7 negative regulator of H3K9me3 · read queue closed · function beats dependency

Six candidate partners were left undecided because the quick numbers could not settle them. Reading what each protein actually does settled all six, and the one with the least impressive numbers turned out to be the most interesting.

The same thing, for a scientist

Cycle 58 adjudicated two of eight read-queue members and left **six explicitly open**, on the stated grounds that a flat CRISPR gene effect is not a rejection ground by itself and that cycle 57's rejections combined **localisation AND UniProt function AND dependency** while that pass had measured only accession structure and dependency. This is the owed third field. **Three are rejected, each on a ground this project has used before:** - **H2BC21** - a histone, and the object argument is the real one. Q104 rejected H3C1 because ten gene names sit on one accession, but noted that *the reason is the object, not the arithmetic*. H2BC21 has its own accession (Q16778, 126 aa) and so escapes the arithmetic - and its **entire UniProt FUNCTION comment is antimicrobial**: *"Has broad antibacterial activity. May contribute to the formation of the functional antimicrobial barrier of the colonic epithelium, and to the bactericidal activity of amniotic fluid."* No chromatin or transcriptional function is curated for it at all. Its locations are **uncoded**. A histone recovered in a nuclear-complex pulldown is a substrate and a structural component, not a private partner. - **PIH1D1** - nuclear on **ECO:0000250 only**, which is by-similarity and is precisely the ground Q104 used to reject FGF12 (*nucleus at ECO:0000305 only*). Gene effect −0.0748, 1.4% of lines below −0.5. Its curated function is C/D box snoRNP assembly, SWI/SNF recruitment to rRNA gene promoters, and R2TP-mediated MTOR/SMG1 stability - real biology, unconnected to this complex. - **CD226** - **Cell membrane** on ECO:0000269, a cell-surface immune receptor for PVR/CD155 and NECTIN2/CD112 driving NK and CTL cytotoxicity, expressed by peripheral-blood T lymphocytes, with a **positive** gene effect (+0.1691, no line below −0.5). Same ground as Q104's MDK rejection: a surface receptor in a nuclear-complex co-IP. **Three earn entries, and the dependency data pointed at the wrong ones:** - **ASB7** is the finding. On the numbers it was the second-flattest member of the set (−0.0457, 0.5% of lines below −0.5) and I was about to reject it. Its UniProt FUNCTION reads: substrate-recognition component of a **cullin-5 RING E3 ubiquitin ligase** (ECS/CRL5) mediating ubiquitination and proteasomal degradation of LZTS1, PSRC1 and **SUV39H1** - and *"The ECS(ASB7) complex acts as a negative regulator of H3K9me3 histone mark by mediating ubiquitination and degradation of SUV39H1: ASB7 is recruited to heterochromatin by CBX5 (HP1 alpha)"*. Chromosome is **ECO:0000269**. So the flattest candidate in the queue is a **direct negative regulator of H3K9me3**, and this graph has an open conflict about exactly that mark at exactly this locus. - **HMGA1** - binds preferentially the **minor groove of A+T-rich regions** of double-stranded DNA, implicated in nucleosome phasing, mRNA 3'-end processing and transcription of genes in or near A+T-rich regions. Gene effect −0.8382 with 90.9% of lines below −0.5. The A+T-rich detail matters here: this graph's break model includes **scaffold-attachment regions**, which are A+T-rich, in the KMT2A breakpoint cluster. - **NUCKS1** - **coded** nuclear and chromosome localisation (ECO:0000269), a chromatin-associated protein promoting **homologous recombination** that binds double-stranded DNA and secondary structures including **D-loops**, with lower affinity than RAD51AP1. Gene effect −0.0091, i.e. flat - and again the function, not the number, is what earns it. **The lesson is the one this project keeps relearning.** Ranked by gene effect the set orders HMGA1 (−0.838), H2BC21 (−0.703), PIH1D1 (−0.075), ASB7 (−0.046), NUCKS1 (−0.009), CD226 (+0.169). Ranked by whether they earn an entry the order is roughly **inverted at both ends**: the second-most-essential is rejected and two of the three flattest are kept.

Facts and figures
why this was runCycle 58 left six of eight read-queue members explicitly open, recording that a non-dependency is not a rejection ground on its own and that cycle 57's rejections combined localisation, UniProt function and dependency while that pass had measured only two of the three. This supplies the third field for all six.
methodgene_exact UniProt query per symbol with the reviewed-record count asserted as exactly one before use, reading cc_subcellular_location with its evidence codes and cc_function verbatim; combined with the Chronos figures measured in cycle 58 with all tissue blocks flattened and a payload guard.
rule 83the Open Targets depMapEssentiality payload carries no release identifier, so no version is implied; the Chronos figures are those measured in cycle 58 in a single pull with all tissue blocks flattened.
absence classmeasured-not-analysed
rejected with the measurement
{
  "H2BC21": {
    "accession": "Q16778, H2B2E_HUMAN, 126 aa, Histone H2B type 2-E",
    "locations": "Nucleus [UNCODED]; Chromosome [UNCODED]",
    "uniprot_function_verbatim": "Has broad antibacterial activity. May contribute to the formation of the functional antimicrobial barrier of the colonic epithelium, and to the bactericidal activity of amniotic fluid",
    "gene_effect": "-0.7027, 76.1% of 1,140 lines below -0.5",
    "ground": "THE OBJECT ARGUMENT. It escapes the accession-ambiguity ground that caught H3C1 and H2BC10 - it has its own accession - but it is a histone, and the only function UniProt curates for it is antimicrobial. A histone recovered in a nuclear-complex pulldown is a substrate and a structural component. Q104 said explicitly that the reason to exclude substrates is the object and not the arithmetic; this is that ground applied on its own for the first time.",
    "what_would_overturn_it": "a demonstration that this specific H2B variant, rather than nucleosomal H2B generally, is required for a fusion-complex function"
  },
  "PIH1D1": {
    "accession": "Q9NWS0, PIHD1_HUMAN, 290 aa, PIH1 domain-containing protein 1",
    "locations": "Nucleus [ECO:0000250] - BY SIMILARITY ONLY, no experimental localisation",
    "uniprot_function": "C/D box snoRNP particle assembly; recruits SWI/SNF to the core promoter of rRNA genes and enhances pre-rRNA transcription; mediates TELO2 interaction with the R2TP complex, necessary for MTOR and SMG1 stability; positively regulates mTORC1 assembly and activity",
    "gene_effect": "-0.0748, 1.4% of lines below -0.5",
    "ground": "nuclear localisation is by-similarity only, which is the ground Q104 used to reject FGF12 (nucleus at ECO:0000305 only), combined with a flat gene effect and a curated function - nucleolar ribosome biogenesis and mTORC1 - with no connection to this complex.",
    "the_link_worth_keeping": "mTORC1. Q106 wrote gene-sesn2 as the ISR-to-mTORC1 handoff this cycle, so PIH1D1 sits one step from a beat this graph now holds. Rejected as an interactome candidate, NOT as uninteresting - if an mTORC1 beat is opened it should be reconsidered there."
  },
  "CD226": {
    "accession": "Q15762, CD226_HUMAN, 336 aa, CD226 antigen",
    "locations": "Cell membrane [ECO:0000269]",
    "uniprot_function": "cell-surface receptor in intercellular adhesion, lymphocyte signalling, cytotoxicity and lymphokine secretion by cytotoxic T cells and NK cells; costimulatory receptor binding PVR/CD155 and NECTIN2/CD112 on target cells",
    "tissue": "expressed by peripheral blood T lymphocytes",
    "gene_effect": "+0.1691, ZERO lines below -0.5",
    "ground": "a cell-surface immune receptor in a nuclear-complex co-immunoprecipitation, with a positive gene effect. Same ground as Q104's MDK rejection.",
    "and_it_belongs_somewhere_else": "CD226/DNAM-1 is a real immune-checkpoint-adjacent receptor on T and NK cells and this is a leukaemia. It belongs in an immunology beat, not in the interactome."
  }
}
earn an entry
{
  "ASB7": {
    "accession": "Q9H672, ASB7_HUMAN, 318 aa, Ankyrin repeat and SOCS box protein 7",
    "locations": "Chromosome [ECO:0000269]; Nucleus [ECO:0000250]",
    "uniprot_function_verbatim_extract": "Substrate-recognition component of a cullin-5-RING E3 ubiquitin-protein ligase complex (ECS complex, also named CRL5 complex), which mediates the ubiquitination and subsequent proteasomal degradation of target proteins, such as LZTS1, PSRC1 and SUV39H1. The ECS(ASB7) complex acts as a negative regulator of H3K9me3 histone mark by mediating ubiquitination and degradation of SUV39H1: ASB7 is recruited to heterochromatin by CBX5 (HP1 alpha)",
    "gene_effect": "-0.0457, 0.5% of lines below -0.5 - the second flattest in the set",
    "why_it_earns_one": "a direct negative regulator of H3K9me3, acting by degrading the H3K9me3 writer SUV39H1, recruited to heterochromatin by HP1 alpha, with CODED chromosome localisation. This graph holds an unresolved conflict about whether H3K9 methylation restrains or promotes the fusion programme at this locus, and a beat measuring H3K9me3 at the KMT2A breakpoint region is running as this is written.",
    "what_this_entry_does_NOT_claim": "that ASB7 is a member of the KMT2A::AFF1 complex. Its interaction evidence is one bait, two publications, anti-tag co-IP, MI-score 0.35. The FUNCTION is what earns the entry; the interaction is what surfaced it."
  },
  "HMGA1": {
    "accession": "P17096, HMGA1_HUMAN, 107 aa, High mobility group protein HMG-I/HMG-Y",
    "locations": "Nucleus [UNCODED]; Chromosome [UNCODED]",
    "uniprot_function": "binds preferentially the minor groove of A+T-rich regions in double-stranded DNA; suggested to function in nucleosome phasing and 3'-end processing of mRNA transcripts; involved in transcription regulation of genes containing or close to A+T-rich regions",
    "gene_effect": "-0.8382, 90.9% of 1,258 lines below -0.5 - a harder dependency than MEN1, KMT2A or DOT1L",
    "why_it_earns_one": "a genuine chromatin architectural protein with a nuclear and chromosomal assignment and a defined DNA-binding preference. The A+T-rich preference is specifically interesting here because this graph's break model includes scaffold-attachment regions, which are A+T-rich, within the KMT2A breakpoint cluster.",
    "the_caveat_that_travels_with_it": "its locations are UNCODED, and cycle 58 already recorded that a pan-panel mean is the wrong statistic for comparing a broad dependency with a selective one - so 'harder than MEN1' must not be read as more important than MEN1."
  },
  "NUCKS1": {
    "accession": "Q9H1E3, NUCKS_HUMAN, 243 aa, Nuclear ubiquitous casein and cyclin-dependent kinase substrate 1",
    "locations": "Nucleus [ECO:0000269 x2]; Chromosome [ECO:0000269] - the only member of the six with fully CODED nuclear and chromosomal localisation",
    "uniprot_function": "chromatin-associated protein involved in DNA repair by promoting homologous recombination; binds double-stranded DNA and secondary DNA structures such as D-loop structures, but with less affinity than RAD51AP1",
    "gene_effect": "-0.0091, 1.4% of lines below -0.5 - the flattest in the set",
    "why_it_earns_one": "coded chromatin localisation plus a homologous-recombination function and D-loop binding. This graph holds a break-repair machinery beat and an R-loop beat, and D-loop binding sits between them.",
    "the_flatness_is_not_a_rejection": "a non-dependency can be a real complex member; the ground for rejection was never the number.",
    "CORRECTED_same_cycle_by_gene_nucks1": "This block justified NUCKS1 partly on D-loop binding and said it 'sits between the break-repair beat and the R-loop beat'. Writing gene-nucks1 measured that: (NUCKS1 OR NUCKS) AND \"D-loop\" returns 0 and AND \"R-loop\" returns 0, both GENUINE zeros with parsed querytranslations, and the primary paper's abstract (PMID 26323318) does not mention D-loops - the phrase is UniProt's. The R-loop link was mine and has an empty literature. What the entry actually stands on is the RAD51AP1-paralogue relationship with reduced DNA-binding affinity, impaired homologous recombination, and GREATLY INCREASED MITOMYCIN-C SENSITIVITY with more spontaneous and MMC-induced chromatid breaks - which connects it to this graph's ICL/Fanconi beat, not its R-loop beat."
  }
}
the ordering is inverted at both ends
{
  "by_gene_effect": [
    "HMGA1 -0.838",
    "H2BC21 -0.703",
    "PIH1D1 -0.075",
    "ASB7 -0.046",
    "NUCKS1 -0.009",
    "CD226 +0.169"
  ],
  "by_whether_they_earn_an_entry": [
    "ASB7 keep",
    "HMGA1 keep",
    "NUCKS1 keep",
    "H2BC21 reject",
    "PIH1D1 reject",
    "CD226 reject"
  ],
  "reading": "the second-most-essential member is rejected and two of the three flattest are kept. A CRISPR gene effect orders these candidates almost orthogonally to whether they are worth writing, which is the argument for having required the third field."
}
queue status after this
{
  "cycle_58_queue": 9,
  "adjudicated_cycle_58": [
    "H2BC10 (accession ambiguity)",
    "RPL13 (pan-essential ribosomal)"
  ],
  "rejected_here": [
    "H2BC21",
    "PIH1D1",
    "CD226"
  ],
  "earn_entries_and_still_unwritten": [
    "ASB7",
    "HMGA1",
    "NUCKS1"
  ],
  "so": "every member of the queue is now an entry, a documented rejection, or a named commitment to write - and the three commitments carry the reason each earned one."
}
How this could bite us back
  • Nothing here nominates a target. Three of the six are rejected, and of the three kept, one is a pan-panel dependency in 91% of screened lines (HMGA1) and two are non-dependencies - so none is a selective vulnerability and none has been shown to bind the fusion.
  • ASB7's function must not be turned into a therapeutic proposition. Raising H3K9me3 by blocking SUV39H1 degradation runs directly into this graph's own unresolved conflict about whether H3K9 methylation restrains or promotes the fusion programme, and the two sides predict opposite signs for the same intervention.
Still open
  1. Does the H2B type 2-E variant have a specific role, or is its antimicrobial-only FUNCTION comment an artefact of what got curated? This is the single judgement holding the H2BC21 rejection in place.
  2. Is ASB7-mediated SUV39H1 degradation active at the KMT2A locus, or only in HP1-alpha heterochromatin? The breakpoint element is an active CTCF/cohesin anchor in a transcribed gene, which is close to the opposite context.
  3. Does HMGA1's A+T-rich minor-groove preference put it on the scaffold-attachment regions in the KMT2A breakpoint cluster? This graph's break model includes SARs and nobody has asked.
  4. ANSWERED AND THE PREMISE WAS WRONG: (NUCKS1 OR NUCKS) AND "R-loop" is a genuine zero, as is AND "D-loop". The D-loop detail is a UniProt phrase not reachable at title-abstract level. The live question is whether the full text of PMID 26323318 supports it, and separately why gene-rad51ap1 - the stronger paralogue - is unwritten.
  5. Should PIH1D1 be reconsidered in an mTORC1 beat? It regulates mTORC1 assembly, and gene-sesn2 was written this cycle as the ISR-to-mTORC1 handoff.
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Three of the six open queue members are rejected on grounds this project has used before, each established from the UniProt record with its evidence codes. H2BC21 (Q16778, 126 aa, Histone H2B type 2-E) carries Nucleus and Chromosome as uncoded assignments and a FUNCTION comment that is entirely antimicrobial - broad antibacterial activity, a possible contribution to the antimicrobial barrier of the colonic epithelium and to the bactericidal activity of amniotic fluid - with no chromatin or transcriptional function curated. PIH1D1 (Q9NWS0, 290 aa) is nuclear on ECO:0000250 only, by similarity, with a function in C/D box snoRNP assembly, SWI/SNF recruitment to rRNA gene promoters and R2TP-mediated MTOR and SMG1 stability. CD226 (Q15762, 336 aa) is Cell membrane on ECO:0000269, a costimulatory cell-surface receptor for PVR/CD155 and NECTIN2/CD112 on T and NK cells, expressed by peripheral blood T lymphocytes.

    What this evidence cannot show

    These are readings of curated records, not of primary literature: no paper behind any FUNCTION or localisation comment was read here, so 'UniProt curates no chromatin function for H2BC21' is a statement about the record and not about the biology. Absence of a curated function is weak evidence of absence of function, and it is weakest exactly for histone variants, where the shared nucleosomal role may simply be annotated on the family rather than the variant. The rejections are decisions about what to write, not claims that these proteins do not bind the complex.

    Evidence that argues against this (2)
    • The H2BC21 rejection is the shakiest of the three and should be read as the weakest link in this analysis. Its gene effect is -0.703 with 76.1% of lines below -0.5, which is a real dependency, and its own accession makes it distinguishable - so the only thing standing between it and an entry is the judgement that a histone is a substrate. If the H2B type 2-E variant has a specific role, this rejection is wrong, and the antimicrobial-only FUNCTION comment is at least as likely to reflect what happened to get curated as what the protein does.
    • Rejecting CD226 on compartment grounds may discard something relevant to a different question. It is a costimulatory receptor on T and NK cells in a disease treated partly by immune mechanisms, and its recovery against a nuclear bait is far more likely a pulldown artefact than biology - but 'not an interactome member' is not 'not worth an entry', and this analysis chose the narrower reading.
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ASB7, the second-flattest dependency in the set, is a negative regulator of H3K9me3. UniProt records it as the substrate-recognition component of a cullin-5-RING E3 ubiquitin-protein ligase complex (the ECS or CRL5 complex) that mediates ubiquitination and subsequent proteasomal degradation of target proteins including LZTS1, PSRC1 and SUV39H1, and states that the ECS(ASB7) complex acts as a negative regulator of the H3K9me3 histone mark by mediating ubiquitination and degradation of SUV39H1, with ASB7 recruited to heterochromatin by CBX5 (HP1 alpha). Its chromosome localisation is ECO:0000269 and its nuclear localisation ECO:0000250. Its Chronos gene effect is -0.0457 with 0.5% of 1,258 lines below -0.5.

    What this evidence cannot show

    The FUNCTION comment cites four primary papers (PubMed 16325183, 27697924, 39039081, 40440427) and none was read here, so the strength, cell type and directness of the SUV39H1 degradation result are not established - this is a curated summary. Nothing connects ASB7 to the KMT2A locus or to a KMT2A-rearranged cell: its interaction evidence in this project's census is one bait, two publications, anti-tag co-IP and an MI-score of 0.35, well below IntAct's own 0.6 threshold. The entry this justifies is warranted by the protein's function, not by the interaction.

    Evidence that argues against this (2)
    • The timing invites a mistake this analysis should not make. A beat measuring H3K9me3 at the KMT2A breakpoint region is running as this is written, and it would be easy to promote ASB7 from 'a protein whose function touches that mark' to 'the explanation for whatever that beat measures'. Nothing supports that: ASB7 acts on SUV39H1 in heterochromatin via HP1 alpha, and the KMT2A breakpoint element is an active CTCF/cohesin anchor in a transcribed gene, which is close to the opposite chromatin context.
    • It also cuts against the mark-identity resolution this graph proposes. That resolution needs H3K9me1/2 and H3K9me3 to behave differently at the locus; ASB7 is evidence that H3K9me3 levels are set partly by writer DEGRADATION rather than by the demethylase/methyltransferase balance the mechanism entry models - so a third control layer exists that neither side of that conflict accounts for.
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    Ranked by CRISPR gene effect the six members order HMGA1 -0.8382, H2BC21 -0.7027, PIH1D1 -0.0748, ASB7 -0.0457, NUCKS1 -0.0091 and CD226 +0.1691; ranked by whether the localisation-and-function read justifies an entry, the three kept are ASB7, HMGA1 and NUCKS1 and the three rejected are H2BC21, PIH1D1 and CD226. So the second-most-essential member is rejected and two of the three flattest are kept, and NUCKS1 - the flattest of all six - has the only fully coded nuclear and chromosomal localisation in the set alongside a homologous-recombination and D-loop-binding function.

    What this evidence cannot show

    Both orderings are judgements applied to measurements, and the second one is mine - another coordinator applying the same three fields could keep H2BC21 and reject HMGA1 on the uncoded localisation. The claim that the orderings are 'inverted' is about the ends of the list and not a statistical statement; with six items no correlation is meaningful. The Chronos values carry no release identifier and are pan-panel means, which cycle 58 established is the wrong statistic for comparing selective with broad dependencies.

    Evidence that argues against this (1)
    • The inversion may be an artefact of the read-queue threshold rather than a lesson about dependency data. These six are what a two-publication-or-two-bait filter admitted from a single atlas-heavy census, so they are not a sample of candidate partners in general - and cycle 58 already noted that CD226, whose loss slightly helps 1,258 lines, cleared that threshold. A set assembled by a weak filter will order badly by any measure.
Connected to