The 2026-10-10 audit retains the published Q9UPV0 assertion objects and both alternative-product records. Fresh normal intake independently reproduced the same 48 assertion objects and two products; its 51 raw GOA records seed 48 assertions. Existing GOA, UniProt, publication/Reactome caches and the genuine older Falcon report are preserved. A new bounded Falcon attempt timed out and the explicit Perplexity-lite fallback returned a quota/authentication error; no provider report was fabricated. The scientific assessment below is manual.
CEP164-source-evidence.yaml is the durable bounded field extract. It includes exact source tuples, query and response hashes, record URLs, participants, host species, methods, features and negative flags; HPA cell-line/antibody fields; the tear-table target row; and all 21 official Reactome event identities/versions/compartments. It does not reproduce whole publication prose or stand in for the machine GOA source.
Selected full target sections of PMID:17954613 establish human RPE1 ciliation phenotypes and distal-appendage immunogold placement. PMID:29789620 human RPE1 dSTORM maps the CEP164-bearing appendage blades. PMID:21399614 supplies the mature-centriole appendage context. PMID:24421332 Fig. 3A–E directly images CEP164 in human RPE1; PMID:26337392 Fig. 2A supplies human RPE1 co-staining separately from mouse tracheal experiments. The genuine fresh full PMID:27623382 text was read without replacing the old abstract-only cache: Figs. 5F/6A/S6E use anti-CEP164 as a human RPE1 basal-body marker. PMID:23348840 was inspected through its authentic abstract and indexed primary Results/figure text; the detailed membrane-docking perturbation targets CEP83, not CEP164 alone. Core assembly participation is independently grounded by CEP164-specific recruitment and rescue.
All four IBA source objects retain their PAINT node and supporting entities. Target-containing evidence lists are legitimate, not circular. The historical node reconstruction was not independently repeated. Direct human recruitment work supports is_active_in at the centrosome/centriole, beyond merely seeing the protein there. Current GO:0097539 is a protein-containing complex, making core in_complex appropriate. No new annotation is needed for the existing assembly coverage.
The 21 Reactome event records were independently retrieved at their original identifiers. Their exact versions, titles, cytosol compartment and participant summaries are extracted. This authenticates event-level location context; the complete recursive component tree and every underlying primary assay were not reconstructed. The location annotations do not confer AURKA/PLK1/RAB3IP enzymatic activities on CEP164.
The 18283122 primary Fig. 1/Results and 26966185 official full publisher selected Results/Discussion/Methods were inspected through indexed full access. The positive HeLa nuclear assays and later negative human-cell findings remain distinguishable. HPA037606 and HPA061504 also differ in U-251MG. The U decisions preserve this conflict without diagnosing proven cross-reactivity or universal absence. Canonical abstract caches remain unchanged; lack of a local full download is not absence of external primary access.
PMID:22664934 full XML Table II explicitly lists CE164_HUMAN / CEP164. The original HDA extracellular observation concerns pooled human Schirmer-strip tear proteomics. The sample route, extracellular activity and secretion are unresolved; absence of the row from the text-only cache does not refute detection. The target row and XML hash are retained in the source extract.
No assay-specific UniProt isoform tags were invented. The original two products remain unchanged. Old notes are an exact prefix, including their dated, superseded interpretations and inherited quotations. Repeated YAML quotations are consolidated and new excerpts use only each source's remaining budget; no new quote is added from a source whose inherited notes already exceed 25 words. This is an explicit legacy-preservation exception, not a claim that the entire inherited notes file passes the current aggregate cap. The new dated notes supersede the earlier generic-binding removals and extracellular over-annotation judgment.
The later PMID:42288497(https://doi.org/10.1038/s41467-026-73638-4) full normal cache adds a bounded Fig. 3/Methods check. Source-specific interpretation is recorded in its reference assessment and the narrow 23253480 finding review. Existing reaction caches remain unchanged.
The earlier quotation-handling paragraph records the first audit's choices. The current follow-up restores short publication anchors at the mechanistic claims while preserving that history. The 25-word-per-source discipline is an authoring constraint, not a repository validator requirement. Current YAML and the new notes append remain within that limit including repetitions; unchanged inherited notes are explicitly excluded from newly authored quotation counts. No whole-file historical compliance claim is made.
Exact unchanged cache passages now anchor TTBK2 complex formation/recruitment (PMID:24982133), isolated-domain binding and the negative autophosphorylation result (PMID:34499853), ciliation loss (PMID:17954613), appendage structure (PMID:29789620), negative nuclear localization in the tested models (PMID:26966185), and purified Rabin8 interaction (PMID:23253480). These excerpts do not alter assay, species or construct scope, nor the narrow later docking-model dispute. The cytosol lists retain their authentic event references and omit the unrelated nuclear-study support. This ledger append adds no literature quotation.