CEP104 is a ciliary-tip microtubule regulator. The 2025 full-length primary reconstitution supports microtubule stabilizing activity (GO:0140778), while earlier TOG-domain assays establish tubulin binding and bulk polymerization behavior under their own construct conditions. These results need different descriptions, rather than assigning a general polymerase mechanism to the whole protein. The activity core is placed at the cilium; the genuine centriolar, centrosomal and mitotic-pole localizations are retained as non-core context.
All 17 original source assertions and all three product records are preserved. The review has three ACCEPT, thirteen KEEP_AS_NON_CORE and one UNDECIDED decisions, plus one directly supported NEW molecular function. The uncertain row is the weak ARMC9 association, reported in 2020 but not reproduced in 2025. No NEW process is asserted and no partner’s kinase, motor or splicing activity is transferred to CEP104. Specific functional differences between the three products remain unresolved.
Generic associations follow the published ClinGen project instruction: a supported, biologically correct generic binding assertion is retained as KEEP_AS_NON_CORE when no justified narrower activity has been established. This is the standing user instruction for this project. It does not turn a BioID proximity observation into a direct-binding experiment. The three 2015 proximity pairs have later source-specific co-IP corroboration; SF1 remains a fragment-screen association with unknown physiological role.
The PAINT cilium inference is evaluated as ancestral functional inheritance. Human O60308 appearing among its own descendant evidence is legitimate. No donor-count criticism or uninspected ancestral placement is introduced. Exact machine-fetched PANTHER metadata is preserved; its generated family description is not treated as primary scientific evidence.
Primary evidence and precise access limits and literal interaction records accompany the review. Cache availability and actual external reading are distinguished: 32453716, 23970417, 27402853 and 32820051 retain authentic abstract-only caches despite successful external selected full-text access. The final 36606424 body was not recovered; its primary preprint is explicitly distinguished, and independent full papers corroborate the relevant associations/localization. No inaccessible supplementary figure is claimed inspected.
The genuine normal Falcon attempt timed out after 90 seconds. Its genuine Perplexity-lite fallback returned a quota/authorization error. No provider report was fabricated. Research proceeded from authentic primary papers, curated interaction records and current ontology definitions. A bounded independent core peer checked human/ortholog and full-length/domain assay scope. Only one short primary-source quotation is authored in the YAML; it is not repeated here.
This entry supersedes the earlier no-process and single-core statements without rewriting that record. The functional synthesis now separately represents tubulin binding (GO:0015631) and microtubule stabilization (GO:0140778), both at the specifically supported ciliary tip (GO:0097542). Human TOG-domain SEC-MALS directly supports the binding component PMID:28017521. Soluble-tubulin binding, full-length end engagement and isolated-domain bulk polymerization remain distinct assays. Stabilizing activity does not subsume binding in GO.
No NEW binding row is added: current GO ancestry puts GO:0015631 and GO:0051010 below the existing GO:0005515, and the standing user instruction rejects a NEW ancestor or descendant of an existing term. This is the direction of the hierarchy intended by the earlier reference to an existing ancestor. The original partner-specific source assertions cannot be changed to tubulin binding merely because independent papers demonstrate it; their WITH partners and assays remain SF1, CCDC66, CSPP1, TOGARAM1, ARMC9 or CEP104 itself. The core records the independently supported molecular function without manufacturing that source transfer.
The 2025 purified-protein reconstitutions strengthen the CCDC66/CSPP1 interaction evidence: each recruits CEP104 onto microtubules and potentiates end blocking/stabilization, with recruitment lost on deletion of CEP104's jelly-roll region. These experiments are distinguished from the earlier cellular co-IPs and BioID proximity measurements. An atomic contact interface is not claimed. The ARMC9 annotation remains UNDECIDED: the original weak positive co-IP is preserved, and a narrowly DISPUTED finding records later non-replication under the tested conditions. It does not label the earlier assay false or establish universal absence of interaction.
The human centrosome screen separately reports proteins that PMID:21399614. This short anchor supports the broad centrosomal context; the precise distal-centriolar/ciliary-tip localization comes from the independently inspected human experiments in PMID:23970417, not from an uninspected target panel in the screen. The YAML also anchors the SF1 screen in the full PMID:26420826 cache. Primary quotations are distributed across these claims and the ARMC9 finding without duplicating long passages.
The process gate supports one NEW GO:1902017 regulation of cilium assembly, specifically ciliary elongation. Human RPE1 depletion reduces length, but the decisive link is the full-length WT versus TOG tubulin-contact-mutant rescue in PMID:32820051 Figure 2D–F, together with direct tubulin interaction and end regulation by CEP104 itself. The Methods do not explicitly identify the rescue cDNA species/accession; the endogenous human depletion and independently defined human biochemical constructs are recorded separately. Neither a full-length polymerase mechanism nor a role in the initiating CP110-removal step follows. This is an inferred connection between the mapped cellular phenotype and established molecular work, not a claim that in-vitro stabilization alone reconstructs the whole cilium.
The bounded current QuickGO comparator check returned 11 assembly/descendant annotations: CCDC66 has experimental GO:0060271 (PMIDs:28235840,36606424), TOGARAM1 has experimental GO:0060271 (PMID:32453716), and ARMC9 has inferred assembly annotations. The query returned none for CEP104 or CSPP1 and no exact GO:1902017 hit in this five-protein comparison. This positive same-role assembly convention is not proof that every regulator carries the identical term. GO:1902017's regulatory parents match modulation of assembly extent; it has no ancestor/descendant overlap with the 17 machine assertions or the existing stabilization NEW. The exact PR-head GO-CAM index had no O60308/CEP104 entry. One broad regulatory process is therefore added; neither a second assembly term nor a directional child is proposed. Detailed query provenance appears in the adjacent manual-evidence record.
The unchanged raw GOA has 18 records, while normal seeding produced 17 machine assertions. TSV lines 14 and 18 encode the same O60308/GO:0005515/IPI/PMID:32453716/Q9Y4F4 assertion and differ only in assigned source (IntAct versus UniProt) and date (20260725 versus 20210803). That provenance-level duplicate explains the count; no biological assertion or product was lost. All 17 source objects, three products, and the prior stabilization NEW are preserved. There are now 19 reviewed entries: three ACCEPT, thirteen KEEP_AS_NON_CORE, one UNDECIDED and two NEW. The literal interaction JSON remains unchanged because it is a machine-field extraction, not a new analytical model; changing its format would not improve the source linkage.