RPA2 is the only compartment-compatible candidate of the six and the corpus already has a KMT2A resection thread it would fit — but it is an abundant ssDNA-binding subunit on 6 spectra in a benzonase-treated lysate with no RPA1 or RPA3 co-recovered — declined, thread flagged

Resolved by accession with the primary gene name read back (rules 62, 65); synonyms REPA2, RPA32, RPA34, and no row in data/symbol-collisions.json. Rule 91 applies: five ECO:0000269 citations means curators saw experiments in five papers, not that any of them is the founding or best localisation study, and none was opened here. A correct compartment is a necessary and very weak condition - most of the nuclear proteome is compartment-compatible with a nuclear bait.

• Compartment compatibility is the weakest possible point in a prey's favour and this entry should not be read as giving it weight. 37 of the 110 MLL-AF4-specific preys are nuclear; being one of them distinguishes RPA2 from ACADM and GLUD1 but not from 36 other unadjudicated proteins in the same tail. The reason RPA2 got its own entry is the pre-existing resection thread, not the annotation. This repository 2026

Read from the cached Supplementary Table 5 workbook with the column map asserted against the sheet's header row. RPA2 is not discussed anywhere in the paper. At 270 aa RPA2 is a small protein and 6 spectra from it is proportionally more than 6 spectra from a large one, which is a point in its favour that this entry does not quantify - no length-normalised spectral index (emPAI, NSAF) was computed for any prey in this beat.

• Spectral counts are not comparable across proteins of different size, and this entry compares them freely - RPA2's 6 spectra from 270 aa is a higher sequence coverage than GLUD1's 11 from 558 aa or RGPD8's 3 from 1765 aa. A length-normalised index would reorder the six candidates and might put RPA2 first. That calculation was not done, and if it had been it could have changed which of the six this beat spent effort on. Nature Communications 2018

'Absent from the network' means absent from the top-300-by-p list of every bait, not absent from the raw data - 4,600 proteins were recovered in total and 960 published, so RPA1 and RPA3 may have been detected below the cut. This is a measured-not-searched absence answerable from PXD009338. Symbols were matched both exactly and as substrings against the sheet's gene_name column and against its uniprot_id column with P27694 and P35244; the only substring hit was SNRPA1 (P09661), a different gene entirely, which is worth recording as the reason a substring sweep on a three-or-four-character symbol must be read and not trusted. The sheet does use multi-gene group strings, so a subunit inside such a string could still have been missed. The expected_absent_terms field records that the absence of RPA1 and RPA3 IS the claim, so an auditor should not flag their non-appearance.

• RPA1 and RPA3's absence is not decisive on its own. A top-300-per-bait cut is shallow, complex members are routinely recovered unevenly in AP-MS, and the study's GLM scores each prey independently so nothing enforces subunit coherence. This corpus has a complex-level coherence analysis for exactly this reason. What makes the argument here stronger than usual is RPA1's size: at 616 aa it offers roughly 2.3x RPA2's tryptic peptide space, so if the complex were present RPA1 should have been the easier subunit to detect, not the harder. UniProt Consortium 2026
• CORRECTION, cycle 165 - THIS ARGUMENT IS WEAKENED, NOT REFUTED. Q192 reported that PMID 12509449 contradicts the subunit-coherence argument because 'menin binds preferentially to free RPA2, so single-subunit recovery is the predicted outcome'. Re-read against the raw abstract, only half of that holds. In vitro, menin bound preferentially to free RPA2 over the RPA heterotrimer or an RPA2/RPA3 subcomplex, so a menin-bridged association need not bring RPA1 and RPA3 - which does weaken the expectation that an obligate heterotrimer should have brought RPA1. BUT the same paper reports that RPA1 co-precipitated with RPA2 from HeLa extracts using menin-specific antibodies, 'suggesting that menin binds to the RPA heterotrimer or a novel RPA1-RPA2-containing complex in vivo'. So the in-vivo evidence points the other way, and single-subunit recovery is not 'the predicted outcome'. Net: E3 alone no longer carries the decline; the verdict still stands on spectral count, nuclease treatment and abundance (the other evidence items), and those, not E3, are what to cite for it. Mol Cell Biol 2003

Nuclease digestion is not necessarily complete - 15 min at 4 C with chromatinised template leaves short protected fragments, and RPA-coated ssDNA is itself nuclease-resistant to a degree. So 'no DNA-bridged interaction survived' is the intended effect of the protocol rather than a verified outcome, and the authors report no gel or fragment-size check.

• The nuclease treatment argument cuts both ways and the other direction is the more interesting one: because template-bridged associations were removed, any RPA2 that DID co-purify must have done so through a protein-protein contact. If the six spectra are real, they are evidence of a direct or complex-mediated contact rather than of shared chromatin - which would make the row MORE interesting, not less. The entry declines it on the number of spectra and the missing subunits, not on the nuclease. ProteomeXchange / PRIDE, EMBL-EBI 2018

Bounded by titles and abstracts (rule 80) - an RPA2 blot in a supplementary figure of a KMT2A paper is invisible, and for a repair protein that is exactly where such a measurement would live. Name list tried: RPA2 only. NOT tried: RPA32, RPA34, REPA2, RFA2, 'replication protein A', or the complex-name form rule 64 requires ('RPA complex'), and RPA is a complex whose literature is overwhelmingly written at complex level. The zeros are therefore bounded by a weak name list and should be read as 'no paper names RPA2 and KMT2A together in a title or abstract', nothing wider.

• Rule 64 applies directly and this evidence item does not satisfy it: 'RPA' as a complex name was not searched, and a paper reporting an MLL-fusion interaction with the RPA complex would frame it that way and be unreachable from the subunit symbol. This corpus has already lost one absence claim to exactly that failure (gene-ssrp1 and the FACT complex). The zero above should be treated as provisional until the complex form is run and title-checked. This repository, docs/QUEEN_BRIEF.md 2026