CDHR1 is a photoreceptor membrane cadherin whose best-supported physiological role is outer-segment organization and maintenance. The molecular core is cadherin binding, inferred from a directly tested ortholog interaction and conserved human localization. This is not an assignment of phototransduction activity, calcium transport, protease activity, or a native human intercellular junction.
All 14 machine assertions, their original evidence codes, references, qualifiers and supporting entities, and both alternative products are preserved. Seven annotations are ACCEPT, four KEEP_AS_NON_CORE, two MODIFY, and one MARK_AS_OVER_ANNOTATED; no NEW row is proposed.
The broad inherited cell/cell-cell adhesion processes remain non-core because the ortholog mixed-cell assay establishes trans-adhesive capability. The principal proposed retinal function instead connects membrane regions of one cell. These are different scopes, not contradictory observations. The homophilic term specifically requires identical molecules on adjacent cells; a general cadherin domain plus the observed heterotypic assay does not establish that specificity. The over-annotation decision is not a universal claim that every human product lacks homophilic binding.
Cell-adhesion-molecule binding is refined to GO:0045296 cadherin binding. GO:0098641 would additionally constrain the activity to a cell-cell process, which would misdescribe the proposed physiological intraphotoreceptor connection. The broad morphogenesis term is refined to the already represented outer-segment-organization process, retaining the human evidence and acknowledging the independent early fish phenotype. No new developmental process or redundant binding assertion is added.
Calcium binding remains a plausible domain-supported structural property, retained as non-core. The selected papers do not measure a CDHR1 calcium affinity. Plasma membrane, membrane and outer-segment membrane describe the canonical product's functional setting. For the plasma-membrane IBA, surface-mediated activity is supported by the ortholog assays in addition to localization; no historical PAINT node was reconstructed. The original PTN identifiers are retained, and the structured source status for the binding refinement is UNRESOLVED at the historical-node level.
The bounded source extract records the exact source assertions, relevant current QuickGO term metadata, and experimental mouse-donor annotation fields with retrieval URLs and hashes. Mouse Q8VHP6 has IDA plasma membrane from PMID:11738025 and IDA outer-segment membrane from PMID:18654668. The current donor records corroborate the preserved transfer; they do not reconstruct the full historical Ensembl/PAINT state.
Q96JP9-1 and Q96JP9-2 remain exactly as seeded. The canonical product has a transmembrane span at 701–721; the preserved alternative sequence changes the corresponding region and removes the distal canonical tail. The cited assays do not establish equivalent localization or activity for both products. Their equivalence is left as a biological question rather than inferred from a shared gene symbol.
The normal gene and publication commands were used in an isolated workspace with tooling pinned to main c488afbe915c1b6b58ebf39c5820bb383f581910. Existing publication bytes were reused exactly where present. The genuine Falcon attempt timed out at its configured 90-second boundary; the explicit Perplexity fallback returned an insufficient-quota error. A surviving task-specific child was terminated after the wrapper had finished. No provider report was produced or imitated; these notes contain the manual synthesis. Independent annotation/core consultation checked the adhesion geometry, construct species and morphogenesis distinction.
This review follows the project's binding instruction at the pinned source commit. There is no generic GO:0005515 assertion in this seed, so no generic-binding override was needed. No new process term was proposed: the existing organization and maintenance assertions already cover the structural role, and being cleaved does not make CDHR1 the enzyme that performs cleavage.
The unresolved scientific questions are the native human partner/stoichiometry and geometry, calcium dependence, alternative-product behavior, and the mechanism regulating ectodomain shedding. They are preserved in the structured questions. Authentic external full-text access is distinguished from an abstract-only cache: the latter alone does not require labeling an actually inspected full paper inaccessible.
Verbatim evidence anchors appear in the YAML and one short geometry anchor above, with aggregate authored quotations kept below 25 words per PMID across YAML and notes. Other literature descriptions are paraphrased; no quotation is repeated.
The calcium-binding row now cites only its original GO_REF:0000002 domain inference. PMID:41995082 remains discussed as an access boundary in the reason, but is no longer listed as calcium-binding support; its aggregation assay did not measure calcium binding or dependence. The non-core action is unchanged.
The human IMP morphogenesis row now states the source-specific basis for its existing replacement explicitly: adult SD-OCT/AOSLO structural findings support outer-segment organization, whereas the original morphogenesis term describes acquisition of specialized photoreceptor features. GO:0035845 is a lateral semantic replacement, not a descendant of GO:0008594. The early zebrafish structural phenotype remains positive context and a caveat against excluding developmental roles; it is not the rationale for narrowing the human source. No action, source object, replacement term or core function changed.
A structured question now asks whether the PROM1 association is direct and how it relates to PCDH15. The existing mouse PCDH21 co-immunoprecipitation evidence does not establish a purified binary interface or the native human complex. This adds no binding annotation. Existing evidence records, source caches and the prior notes journal are preserved.