GOA carries GO:0004089 carbonate dehydratase activity twice, and I checked the
evidence codes myself in CA8-goa.tsv:
| row | evidence | reference | assigned by |
|---|---|---|---|
enables GO:0004089 |
IEA (ECO:0000256) |
GO_REF:0000002, InterPro IPR018338/IPR023561 |
InterPro |
enables GO:0004089 |
TAS (ECO:0000304) |
PMID:8977131 | PINC |
enables GO:0008270 zinc ion binding |
IEA (ECO:0000256) |
GO_REF:0000002, same InterPro signatures |
InterPro |
There is no IDA, IMP or EXP row for either term anywhere in the file. So the catalytic
assignment rests on (a) a fold/domain match and (b) a single traceable author statement.
PMID:8977131 is Sjöblom et al., FEBS Lett 1996;398(2-3):322-5,
"Two point mutations convert a catalytically inactive carbonic anhydrase-related protein
(CARP) to an active enzyme", DOI 10.1016/s0014-5793(96)01263-x — verified against PubMed,
abstract-only in cache (full_text_available: false).
The title alone says it, and the abstract is unambiguous:
PMID:8977131
The whole point of the paper is that you have to engineer two substitutions
(Arg117→His, Glu115→Gln) to get activity:
PMID:8977131
And the same sentence that establishes zinc binding establishes it for the mutant, by
contrast with the wild-type:
PMID:8977131
So the PINC TAS for GO:0004089 is a misannotation of the very paper it cites: the source
reports the absence of the activity in the wild-type protein. That also disposes of the
GO:0008270 zinc-ion-binding IEA — unmodified CARP was not isolated as a zinc complex, and
the fold-based signature that produced the IEA is exactly what the experiment refutes.
(The paper works on murine CARP, the CA8 ortholog. That is not a problem for the
argument: the human protein has the identical active-site substitution, UniProt flags it,
and PINC used this paper for the human entry in the first place.)
FUNCTION: Does not have a carbonic anhydrase catalytic activity. andCAUTION: Although it belongs to the alpha-carbonic anhydrase family, Arg-116 is present
instead of the conserved His which is a zinc-binding residue. It is therefore expected
that this protein lacks carbonic anhydrase activity.Decision: GO:0004089 (both the IEA and the TAS rows — same term must carry the same
action) and GO:0008270 → REMOVE. This is not second-guessing an experimental
annotation: there is no experimental annotation to second-guess, and the one cited paper
says the opposite of what the TAS asserts.
The fold is retained and repurposed as a protein-interaction surface. The primary finding
is Hirota et al., Biochem J 2003 (PMID:12611586, verified; abstract-only in cache):
Mechanism, restated by the structural work:
PMID:32316137
The interaction is reciprocally validated from the receptor side by Ando et al., PNAS
2018 (PMID:30429331):
PMID:30429331 and
PMID:30429331
This is the strongest part of the case: mutations on either side of the interface
converge on the same functional readout.
Disease: recessive CA8 mutations cause cerebellar ataxia (SCAR34 / CAMRQ3).
PMID:19461874
Directionality: CA8 is a negative regulator —
PMID:42268453, and it acts
on the receptor, not on the ligand:
PMID:42268453
IP3R1 (ITPR1) is a ligand-gated intracellular calcium-release channel. CA8 binds it and
reduces its activity. I looked up candidates rather than guessing:
| candidate | verdict |
|---|---|
GO:0019855 calcium channel inhibitor activity (MF) — "Binds to and stops, prevents, or reduces the activity of a calcium channel." |
chosen; the definition is almost a paraphrase of the CA8 result |
GO:0005246 calcium channel regulator activity (MF) |
correct but less specific — CA8's effect is unidirectional inhibition |
GO:0044325 transmembrane transporter binding (MF) |
records the binding but drops the functional consequence |
GO:0070679 inositol 1,4,5 trisphosphate binding (MF) |
wrong — CA8 binds the receptor's modulatory domain, not IP3 (PMID:42268453 above) |
GO:0051280 negative regulation of release of sequestered calcium ion into cytosol (BP) |
chosen as the process |
There is no GO term for "IP3 receptor binding" specifically; GO:0019855 is the most
informative MF available and it is in the molecular_function branch (checked via QuickGO),
so it passes the strict branch validation on core_functions.
GO:0005515 protein binding × 5 IPI rows (PMIDs 25416956, 25910212, 27107012, 31515488,MARK_AS_OVER_ANNOTATED per project guidance (same action on every row of the sameGO:0005737 cytoplasm (IEA, GO_REF:0000107, Ensembl Compara from mouse P28651) —references: titles filled with fill_refs.py.supporting_text verified as normalised verbatim substrings of the cachedpublications/PMID_*.md files.core_functions (GO:0019855, GO:0051280, GO:0005737) and the CL idCL:0000121 Purkinje cell were each looked up (QuickGO / OLS), not written from memory.The September review and its history are preserved above as historical evidence. This reassessment supersedes its generic-binding removals, its implication that the former catalytic cavity is an experimentally established ITPR1 interface, and any transfer of mouse construct numbering to the 290-residue human protein.
The existing normal GOA file contains 26 assertions. The earlier YAML collapsed these into nine source/reference groups, omitting individual partners. Running the normal seeder in an isolated temporary output restored 17 additional partner records and backfilled source metadata in eight retained rows. All nine older source groups remain represented, and each of the 26 raw assertions maps to exactly one restored source object. This is restoration of machine source assertions, not 17 scientifically proposed NEW annotations. The two pre-existing authored NEW annotations are retained after independent mechanistic and ontology review; no additional NEW is proposed. Raw UniProt, GOA, publication caches and the earlier history are unchanged. There is no alternative-products slot to manufacture or infer.
The final direction is three REMOVE decisions for the unsupported wild-type catalytic/zinc-site assignments, 22 KEEP_AS_NON_CORE interaction assertions, one ACCEPT for cytoplasm, and the two retained NEW proposals for channel inhibition and direct negative regulation of calcium release. Supported generic binding is kept under the standing project instruction; limited informativeness alone is not a biological reason to remove it. The distinct binding consultation read all five source abstracts and available assay-framework passages, and corroborated all 22 literal partners against the human UniProt interaction list. It did not inspect individual CA8 supplemental target cells or claim that every pair received orthogonal or native-context validation. In particular, KLHL8 remains the literal UniProtKB:Q9P2G9-2 product. None of these partners is presumed to mediate ITPR1 inhibition.
For PMID:25416956 and PMID:25910212, the HTML caches contain Abstract and Introduction/Discussion, without the target Results/table. PMID:27107012 provides the barcode-fusion yeast-two-hybrid framework; PMID:31515488 describes Y2H screening and selected human-293T PCA validation; PMID:32296183 describes HuRI screening and retesting. Those framework results support qualified curator deference, not a claim that the CA8-specific supplemental results were personally verified. The consultant's reference-only recommendations were joined to the exact raw records; no new quotations were assigned to these rows.
The one integrated core links CA8's own calcium-channel inhibitor activity to negative regulation of release from intracellular stores. CA8 supplies regulatory work itself; it is not the transported calcium, pore, or a merely necessary cargo. The existing process proposal was checked against its regulation/transport parents. A local comparator check found the more specific GO:0060315 child on the noncatalytic channel regulator CALM1 and broader calcium-regulation annotations on BCL2. This is not a claim of exhaustive comparator coverage or systematic absence across MODs. No CA8 activity was found in the local GO-CAM index. No additional developmental, disease or parent/child process annotation is proposed.
The provider attempt used the installed Falcon client and its configured perplexity-lite fallback once. Both failed DNS resolution and produced no research output. Manual primary-source reading and the independently attributed binding consultation therefore supply the evidence account; no provider-named research file was invented. Further questions retain the context-dependent negative interaction result, the unresolved contact surface and the possibility that some neuronal effects are not exclusively mediated through ITPR1.
The two additional references were retrieved through the ordinary reference fetcher in isolated hosted storage, authenticated and imported as new normal caches without changing existing files. PMID:26399641 is abstract-only and explicitly reports no detected direct binding in its Purkinje-culture study. PMID:19360879 is bibliographic-only despite its Abstract heading; its full structural paper was not read. The primary human PDB 2W2J identity was separately inspected, but is not presented as a substitute for reading every paper-specific structural claim. Both references are listed with these access limits.
All 26 source assertions and the two previously authored proposals have been assessed. Full normal candidate validation, including references, GOA and ontology terms, passed with 22 generic-binding policy advisories and no errors. The status is DRAFT because those advisories persist under the standing binding instruction; it does not mean these records were left unreviewed. The nine short YAML quote occurrences match normal canonical caches exactly, with aggregate reuse totals of 20, 18 and 12 words in the three quoted sources.
The independently reviewed candidate was applied after ROOT science approval. Full normal canonical validation passed with the 22 documented binding advisories, the status report retained DRAFT, and the standard EDIT history passed validation. The 26 restored source assertions and 2 retained prior NEW annotations are unchanged from the reviewed candidate. Existing GOA, UniProt, all 13 publication caches and the earlier history remain unchanged. Targeted HTML rendering was completed; remote publication is recorded separately.
The follow-up preserves all 26 normal source assertions, both prior NEW objects, every partner identifier and all 16 reference identities. It refines the broad cytoplasm decision to cytosol and uses that location in the existing single core. The resulting decisions are 3 REMOVE, 22 KEEP_AS_NON_CORE, 1 MODIFY and 2 retained NEW; there is no additional NEW annotation.
The mechanism anchors now retain the complete affinity-reduction clause in PMID:12611586(https://pubmed.ncbi.nlm.nih.gov/12611586/) and the receptor-side V1538M binding/inhibition result in PMID:30429331(https://pubmed.ncbi.nlm.nih.gov/30429331/). The selected Introduction passage in PMID:32316137(https://pubmed.ncbi.nlm.nih.gov/32316137/) supplies concise allosteric context, and the Discussion synthesis in PMID:42268453(https://pubmed.ncbi.nlm.nih.gov/42268453/) distinguishes receptor interaction from binding the IP3 ligand. These latter passages summarize established work; they are not represented as new direct assays. No solved CA8 contact surface or experimentally proven repurposing of the former active-site cavity is inferred.
The alternative-term assessment is explicit. GO:0019855 calcium channel inhibitor activity captures binding plus inhibition. Its GO:0005246 regulator parent does not specify direction; GO:0044325 transmembrane transporter binding omits the functional consequence. GO:0070679 denotes IP3 ligand binding, which is not established as CA8's activity. The preserved process proposal describes CA8's own inhibitory contribution to calcium release and retains its prior participation/comparator justification.
GO:0005829 cytosol is the non-organelle portion of the cytoplasm and is related to cytoplasm by part_of. The explicit cytosolic description in the complete cached PMID:30429331 Discussion supports the more precise location, alongside the previously reviewed human-fragment interaction work in PMID:19461874. This is cross-source refinement of an existing location assertion, not a newly claimed localization experiment or integral ER-membrane assignment. The original Ensembl source fields remain intact.
PMID:19360879 remains in the bibliography with its bibliographic-only access caveat and is removed from all positive supported_by lists. No paper-specific conclusion is inferred from its title. PMID:26399641 remains visible as countervailing evidence in the annotation reasons, core description, reference assessment and an assay-context question, but is removed from positive support for inhibition. Its negative cultured-Purkinje result does not invalidate every positive recombinant or receptor-defined HeLa experiment, and the positive experiments do not erase the negative result. No unread Methods or image analysis is claimed.
The substantive questions about targeted ITPR1 curation, CA10/CA11 propagation, term specificity and possible residual chemistry have been restored. They are questions rather than assumptions of a missed annotation, a shared CARP interface or an undiscovered catalytic activity. Structural characterization of the CA8–ITPR1 complex and controlled residual-activity screening are again concrete experimental suggestions. The 22 interaction reasons retain each exact partner and curator-deference basis in shorter prose; individual target-table cells and pair-specific orthogonal validation remain uninspected, as detailed above.
Ten verbatim YAML quotation entries have aggregate reuse totals of 20 words from PMID:8977131, 20 from PMID:32316137, 23 from PMID:12611586, 22 from PMID:30429331 and 15 from PMID:42268453. The mechanism-bearing clauses are retained without restoring the previous duplicated long quotations. The old history is append-only and is not rewritten for spacing corrections; any follow-up provenance is recorded in a new standard history entry.
Full normal candidate validation, including reference quotes, GOA consistency and authored term checks, passed with 22 expected standing-policy binding advisories and no errors. DRAFT is retained.
The independently approved candidate was applied and passed full normal validation, a status dry run, targeted rendering and standard EDIT history validation. DRAFT retains 22 binding-policy advisories. All 26 source assertions, two prior NEW assertions, 16 reference identities, products, caches and earlier history remain preserved. The sole action change refines cytoplasm to cytosol. See the follow-up history.