ABCG5 manual readiness notes

Status: not initialized

The normal just fetch-gene human ABCG5 attempt failed before downloading UniProt/GOA or generating a review YAML. The failure was NameResolutionError for rest.uniprot.org with [Errno 8] nodename nor servname provided, or not known; the command exited 1. A normal direct curl to HGNC REST also failed to resolve its hostname (curl exit 6). No network settings, sandbox settings, outside terminal or alternate execution environment were changed.

These are manually authored readiness notes, not a completed annotation review or a provider research report. No ABCG5 source cache, GOA, UniProt record, review YAML or Falcon report has been fabricated. Falcon and gene-publication caching have not been launched because valid seeding did not complete. No annotation counts or actions are assigned.

Identity and conflict checks

NCBI Gene 64240 gives the HGNC-provided official symbol ABCG5, name ATP binding cassette subfamily G member 5, primary source HGNC:13886, human taxon, aliases STSL and STSL2, and Ensembl ENSG00000138075. Its RefSeq/Swiss-Prot cross-references include Q9H222. The MGI orthology record MGI:1351659 independently links human ABCG5, HGNC:13886, NCBI Gene 64240 and UniProt Q9H222. The raw human UniProt source remains to be fetched normally before seeding.

The local genes/human/ABCG5 directory contained no review/source files. The GitHub main contents endpoint for that path returned 404. Searching all PR states for ABCG5 returned only the shared campaign progress PR #3137, not a dedicated gene review. A separate STSL PR search was also checked. No human ABCG5/Q9H222 or mouse Q99PE8/MGI:1351659 entry was found in the currently cached GO-CAM index. This is an index observation, not evidence that no external GO-CAM exists. No shared project or queue files were edited.

Primary evidence to cache and inspect after seeding

The public literature-search tool exposed the primary abstracts and selected indexed article passages below despite shell DNS failures. Several direct PMC page opens returned a browser challenge, so this readiness pass does not claim exhaustive full-text review. The cited papers are not yet in the local publication cache.

• PMID:12208867 — Coexpression of ATP-binding cassette proteins ABCG5 and ABCG8 permits their transport to the apical surface. Recombinant epitope-tagged mouse proteins in cultured cells associate and undergo glycan maturation when coexpressed. In polarized WIF-B cells, ABCG5 reaches the apical/canalicular membrane with ABCG8; expression alone retains immature protein in the ER. Indexed discussion distinguishes the mature heterodimer from ER-retained homo-oligomeric material. Trafficking and assembly evidence should not be called proof of functional ABCG5 homodimer transport.
• PMID:16867993 — Sterol transfer by ABCG5 and ABCG8: in vitro assay and reconstitution. Recombinant mouse G5/G8 in Sf9 vesicles and reconstituted preparations support ATP-dependent, vanadate-sensitive transfer of cholesterol and sitosterol. The abstract reports minimal corresponding transfer of cholesteryl oleate, phosphatidylcholine or enantiomeric cholesterol. Free-sterol transport should not be broadened into efficient sterol-ester or general phospholipid transport on the strength of this assay.
• PMID:18402465 — Purification and reconstitution of sterol transfer by native mouse ABCG5 and ABCG8. Purified endogenous mouse liver G5/G8 reconstitutes selective, stereospecific sterol transfer. Native electrophoresis, sedimentation and crosslinking support a functional heterodimer, with no higher oligomer detected across the tested catalytic states. G5 palmitoylation was not essential for the tested dimerization, trafficking or biliary secretion endpoints.
• PMID:21209088 — Sequences in the nonconsensus nucleotide-binding domain of ABCG5/ABCG8 required for sterol transport. Mutagenesis and mouse biliary-sterol assays distinguish the two composite nucleotide-binding sites. A catalytically nonessential site can still be structurally necessary, and selected mutations discriminate cholesterol versus plant-sterol transport. The future review must map the paper's NBD/site nomenclature to each protein before assigning residue- or subunit-specific ATPase claims.
• PMID:27144356 — Crystal structure of the human sterol transporter ABCG5/ABCG8. A nucleotide-free human heterodimer structure establishes the architecture and asymmetric coupling of nucleotide-binding and membrane domains. Structural interpretation should be separated from direct flux measurements.
• PMID:34404721 — Molecular basis of cholesterol efflux via ABCG subfamily transporters. Human G5/G8 sterol-bound structures identify a cytosolic-leaflet-accessible site and a second transmembrane site. The paper also examines ABCG1; its macrophage-to-HDL pathway must not be assigned to G5/G8 merely because both proteins occur in the article. G5/G8 supports secretion toward bile and gut lumen.

Review priorities once authentic sources are available

ABCG5 participates in the ABCG5–ABCG8 sterol-export complex. Preserve any seeded contributes_to qualifier and distinguish an obligate complex activity from an independently sufficient monomer. Inspect localization by construct, maturation state, partner coexpression and cell polarity. ER residence during assembly is compatible with apical-membrane function and should not automatically become the main functional location.

Trace human-versus-mouse evidence explicitly, including a possible Q99PE8 donor chain, instead of treating rodent experiments as directly performed on human protein. Retain the source fields and verify all GO definitions before refining broad transport/binding terms. Do not infer a novel process from sitosterolemia necessity evidence alone; directly performed sterol export is the central activity, whereas disease-associated vascular or metabolic consequences require separate process review. Any NEW process would still require the repository's participation, comparator, parent-term and GO-CAM checks.

Resume with normal fetch-gene; then launch genuine default Falcon (--fallback perplexity-lite, 1,200-second timeout) and gene-publication caching concurrently, followed by annotation-reviewer consultation, all seeded annotation decisions, core-function synthesis, history, validation and rendering. Until then, no initialized/completed review or provider output should be reported.

2026-09-27 substantive review after authentic source recovery

This entry supersedes the readiness/status statements above. The historical notes remain unchanged as a record of the earlier failed initialization. ABCG5 is now authentically initialized and all 53 seeded annotations have been reviewed. The source3 normal-fetch artifact was independently validated and imported by the parent coordinator, including the human Q9H222 UniProt record and GOA; no machine source was manufactured or edited. The current review is DRAFT because eleven required notes-inclusive publication caches remain missing.

Identity, conflict and source integrity

The authoritative HGNC snapshot lists Approved ABCG5, HGNC:13886, alias STSL; the NCBI identity record also gives STSL2. UniProt Q9H222 is human ABCG5. The parent verified that canonical/alias directories and open gene PRs did not overlap at main 87539cd615ffd315519c8bd9004c0d8bb9948cd8; the initial local notes were preserved. Source baseline is /tmp/ABCG5-root-baseline.json, with primary import receipt tmp/source3-canonical-import-receipt.json and subsequent auxiliary import receipt tmp/source3-auxiliary-import-receipt.json. Both alternative-product entries (Q9H222-1 and Q9H222-2) remain unchanged. All 53 annotation source objects, including qualifiers, supporting entities and original reference IDs, and all 17 seeded reference identities are preserved.

I applied the review, annotation-reviewer and core-function-synthesizer instructions. A fresh genuine Falcon attempt with a 1,200-second timeout and perplexity-lite fallback ran concurrently with normal GOA publication caching. Both provider attempts failed before provider contact because the uvx dependency could not resolve PyPI. /tmp/ABCG5-fresh-research.log records the actual failure; no provider report was created or relabeled. The GOA fetch log is /tmp/ABCG5-fetch-goa.log; the coordinated source3 import supplied the seven genuine GOA publications while that job was running. No cache refresh was attempted. A separate standard fetch for the eleven missing historical/donor records terminated 0/11 with DNS errors, as recorded in /tmp/ABCG5-notes-donors-fetch.log. All process handles have been reaped.

Main primary evidence and species/assay scope

• PMID:11099417, cached abstract: human sitosterolemia mutations include one ABCG5 mutation and six ABCG8 mutations in the reported patients. This is positive ABCG5 disease evidence, not a gene-misattribution issue. Direct transport mechanism is supplied by subsequent studies, so the phenotype is not used alone to manufacture unrelated downstream processes.
• PMID:12208868, full normal cache: a human G5/G8 genomic transgene was expressed in mice. Biliary cholesterol increased more than fivefold, fractional absorption fell about 50%, fecal neutral sterols increased and hepatic synthesis compensated. The Discussion explicitly leaves direct intestinal efflux versus greater biliary sterol delivery unresolved. Its reduced-absorption result supports the negative-regulation direction but does not isolate a human-enterocyte flux experiment. Primary article.
• PMID:14504269, cached abstract plus original full article recovered through an author-uploaded copy and independently read by the ABCG8 reviewer: Figure 1 detects native mouse G5 at apical membranes; Figure 2 detects human G8 in liver membranes from human-G5/G8 transgenic mice. Cultured-cell experiments include human G5/G8 and interspecies partners. The approximately tenfold biliary-cholesterol rescue in Figure 5 uses the mouse pair in double-knockout mice. Original full article.
• PMID:16870176, cached abstract plus publisher-indexed original Results/Figure 7 recovered by the independent ABCG8 reviewer: EGFP-G5/EGFP-G8 and HA-tagged partners are genuine HeLa co-immunoprecipitation controls. The ABCG1-focused title does not imply a false G5/G8 citation. Direct publisher opening failed, so the access claim is specifically the indexed original Results and Figure 7. Publisher article.
• PMID:16893193, cached abstract plus author-uploaded original full article checked in the independent consultation: full-length human proteins were expressed in Pichia, copurified and assayed for MgATPase. Figure 9 uses nucleotide trapping/competition/crosslinking on the pair. The assays support nucleotide handling and catalysis of the heterodimer, not independently resolved native monomer rates. The existing contributes_to ATP-binding and ATP-hydrolysis qualifiers are preserved. Original full article.
• PMID:17474147, cached abstract: a human SH3-domain peptide-array screen and selected PLCgamma1 full-protein follow-up. The exact ABCG5–NCK1 hit was not recovered. Generic protein binding is removed for lack of specific functional information, without asserting that the curator-reported interaction is false or that the screen failed. No adaptor function is invented from that interaction.
• PMID:27144356, full normal cache including Methods/Extended Data: human G5/G8 was expressed in Pichia, purified and assayed/crystallized; human wild-type and mutant constructs were then tested in deficient mice. Human wild-type rescue increases biliary cholesterol about thirtyfold, whereas G5 A540F and Y432A impair export despite maturation. This separates direct function from hypotheses about density/sterol binding and simulations. NBS denotes the composite site: active NBS2 combines G5 Walker motifs and the G8 signature, whereas the opposite site is noncanonical. This does not imply equal ATP competence at both sites. Primary Nature article.

The independent ABCG8 reviewer also recovered full PMID:16867993 (PMC4527585): mouse G5/G8 in Sf9, inside-out vesicles and purified proteoliposomes support ATP-dependent sterol mass transfer and substrate controls. Its Discussion leaves flippase versus other membrane-transfer mechanism unresolved; it does not demonstrate a soluble aqueous-pocket carrier. The historical PMID:12208867, PMID:18402465, PMID:21209088 and PMID:34404721 source leads remain scoped as above. Current PubMed abstract/figure captions for PMID:21209088 were rechecked. No normal cache was hand-edited to incorporate externally read text.

Ontology and annotation decisions

All 53 rows now have decisions: 34 ACCEPT, 12 MODIFY, 6 UNDECIDED and 1 REMOVE. There are no NEW assertions.

1. Broad transporter molecular functions are refined to GO:0034041 ABC-type sterol transporter activity. Its formal reaction couples ATP hydrolysis to sterol movement across a membrane. Existing ATP-binding/hydrolysis contributions are preserved and the integrated core uses contributes_to for the full transporter activity.
2. GO:0120020 cholesterol transfer activity currently describes an aqueous-phase carrier with a hydrophobic pocket between membranes or particles. The ABCG5/8 experiments establish a membrane exporter. The three existing assertions are refined to GO:0034041 while preserving contributes_to. This is a term-scope issue, not loss of cholesterol transport, a false ancestry claim, or an inference that “transfer” in an article title establishes the modern carrier mechanism.
3. GO:0043235 signaling receptor complex requires signal-ligand recognition that initiates a cellular response. The positively established assembly is GO:0043190 ABC transporter complex, whose four core domains may be distributed over multiple polypeptides. The two receptor-complex rows are refined accordingly. This does not dispute the actual G5/G8 assembly experiments.
4. GO:0030299 intestinal cholesterol absorption describes uptake from the small intestine into blood. The IC annotation attached to the reduced-absorption transgene result is refined to GO:0045796 negative regulation of intestinal cholesterol absorption. The latter has a negatively_regulates relation to the absorption process, not an is_a or part_of relationship that makes the exporter a positive uptake step.
5. Broad transmembrane transport is refined to GO:0035382 sterol transmembrane transport, a verified child of both sterol transport and transmembrane transport. The existing broad sterol process remains an accurate core assertion; no redundant NEW descendant is added.
6. Accurate broad membrane/plasma-membrane/apical-part annotations are retained at their source resolution. The presence of more specific apical evidence is not a reason to reject broad localization. Mutant Reactome events receive their own caveats; those caveats are not pasted onto unrelated sources. The expression event's regulator is NR1H2/NR1H3, whereas ABCG5 is the membrane-transporter product.
7. The G5/G8 generic binding and dimerization rows are refined to heterodimerization. The NCK1 generic-binding row is removed as uninformative without denying the interaction. One integrated core combines this assembly function, contribution to ATP-powered sterol transport, apical location, ABC-transporter-complex membership, efflux and negative absorption/homeostasis roles.

Live term definitions and parents were read through the linked primary GO pages. A review of gocams/index.tsv found no matching human Q9H222, mouse Q99PE8 or rat Q99PE7 entry in the available index; this does not assert absence from all external GO-CAMs. No proposed process depends on a supposed gap in another gene's annotations, so no new comparator-based assertion is made.

Propagation provenance and unresolved donor context

The MGI comparative Abcg5 GO graph identifies mouse Q99PE8/ENSMUSP00000069495 and rat Q99PE7/ENSRNOP00000007174 separately. The graph was generated 2023-03-10; its historical annotation snapshot is not presented as a fresh donor GAF.

These six human transfers remain UNDECIDED. The donor expression changes are positive findings and IEP is not rejected categorically. Full source context and human protein-level functional transfer remain unresolved. Rat cholesterol-homeostasis transfer is accepted because independent human disease/transgene/transport evidence establishes the target function; its rat IEP source is still described accurately.

All IBA propagation reviews use only their PTN ancestral nodes as source_entities, not the extant member list. The original IBD placement/alignment was not recovered, so source status records that limit. Human ABCG5 among descendant evidence is valid experimental grounding, not circularity. InterPro and ARBA predicates not inspected are explicitly UNRESOLVED rather than claimed verified. The imported PANTHER family data are useful identity context; unreviewed machine-generated family descriptions are not biological evidence.

Reference/cache and verification boundary

All 17 original reference identities remain unchanged and all 29 current references have manual assessments. VERIFIED denotes the recorded identifier/source and bounded supporting content actually checked, not a claim that every possible experiment or every transferred term has been verified. The seven GOA publication caches include full normal records for PMID:12208868 and PMID:27144356; other full_text_unavailable flags remain true because those local caches are abstract-only even where an external original article was read.

The exact notes-inclusive missing normal publication records are PMID:12208867, PMID:16867993, PMID:18402465, PMID:21209088, PMID:34404721, PMID:15710224, PMID:17109865, PMID:17132608, PMID:23117815, PMID:25263431 and PMID:25612518. Normal fetch failed 0/11 with DNS errors; none is a fabricated provider source. The four cited Reactome records are present from verified source3 recovery. These eleven publication records keep the review/PR in draft until standard retrieval and exact quote/title validation can run against them.

Supporting snippets in the YAML are exact cached excerpts, checked case-sensitively after whitespace normalization. Source preservation, history validation, targeted validation and rendering are recorded in the final manifest. The initial notes and protected machine-source bytes remain preserved.

Independent sibling review of all 53 decisions, description and integrated core found no biological blocker. Its scope refinement was adopted: unresolved human response mechanisms are bounded to the source studies inspected, rather than framed as an exhaustive absence from the literature.

The parent independent read of all 53 reasons and the integrated core found no biological blocker. Validation exposed an object-shape error in the two knowledge-gap entries; these were corrected to schema-defined gap_statement/boundary objects, with an exact existing source quote for the absorption boundary. Final schema, ontology, GOA and best-practice checks pass. The complete targeted reference/markdown validation subsequently exited 0 with all validations passed and one grouped warning for the eleven documented missing publication caches. History validation and rendering also passed; the review remains DRAFT.

2026-09-27 — source8 closure and current-head review follow-up

This entry supersedes the earlier cache-gate and action totals without rewriting their historical record. The starting canonical review, notes, HTML, GOA, UniProt and published history were byte-for-byte identical to PR #3285 head dbe5fffaf0a84e03c2f72d6b4d0f74604afd4df8. Exact copies and hashes are in tmp/ABCG-pair-followup-baseline/. The complete formal review and issue comment were read at that head. No source annotation, qualifier, supporting entity, alternative product or reference ID/title is changed.

The previously missing 11 records were produced by the normal fetcher in source8, not source9. Root independently validated and imported the artifact without overwriting existing caches. The import receipt tmp/source8-canonical-import-receipt.json has SHA-256 1ebe816a5f570279c1a6fe45f87dbb54841b3827e7ce090a5c53d41df3ba5bce and source run 36301782511. Every included publication byte matches that receipt and each added cache path is absent from this exact PR base. No new provider report, authored cache or repeat fetch was used.

Actual recovered source scope

PMID:34404721 now has a genuine full main-article cache. Human G5/G8 cDNAs were expressed in HEK293 cells and Pichia; human WT and mutant adenoviruses were tested in deficient mouse hosts. Figures 1–3 and their Methods support pair ATPase, inhibitory G8-binding Fab2C7, sterol-site mutations, and loss of biliary rescue despite comparable protein maturation. The G8 Walker-containing NBS1 is non-equivalent to the active NBS2, which combines G5 Walker elements with the G8 signature. Its ABCG1 ATP-bound structures, macrophage/HDL context and ABCG1 ATPase kinetics are separate experiments. The cached XML repeats some sections; neither that metadata flag nor this read certifies every supplement or original figure pixel. The corresponding full_text_unavailable flag is now false. Primary content was also checked through PMC8403869.

Every other recovered record remains abstract-only locally. PMID:12208867, where cited, explicitly studies epitope-tagged mouse proteins in cultured cells; PMID:16867993 uses recombinant mouse proteins in Sf9 membranes, whereas PMID:18402465, where cited, purifies native mouse liver transporter. None is relabeled as a human purified assay. Prior external full-text reads remain explicitly separate access routes. The recovered abstracts corroborate the recorded assay/organism boundaries, rather than supplying unseen full experiments.

The rat response sources remain abstract-only. Their positive diet, drug, exercise and radiation expression findings are retained. Cache recovery does not resolve the complete source-specific experiments or conserved effector role in humans, so the existing UNDECIDED judgments remain. An IEP annotation is not rejected merely because it is expression-based, and lack of a human assay alone does not refute orthology transfer. Resolving these judgments requires the actual full experiments and a justified transfer of that particular response context. The established human sterol-export and homeostasis judgments remain supported independently.

Feedback adjudication

The generic GO:0005515 IPI row on PMID:16870176 is now REMOVE because the same paper and the exact same partner are already represented by the retained GO:0046982 heterodimerization row. This is removal of an uninformative term for this verified interaction, not rejection of the interaction or a rule that duplicate GO IDs across independent sources are invalid. The source row and its evidence are preserved. The informative assembly, conserved broad compartment assertions and specific transporter refinements remain unchanged. Broad sterol transport is true at the source's resolved level even where another source supplies a narrower membrane mechanism.

The reviewer proposed changing activity judgments or recommending qualifier edits on the basis of ordinary enables. The annotation-reviewer skill explicitly treats ordinary relationship qualifiers as inert in this review workflow: they are preserved, not added, edited or used as the reason for an action. Existing contributes_to assertions retain their weaker, biologically meaningful scope, and the integrated core retains contribution to the heterodimer's transport activity. Thus no artificial qualifier-change recommendation is added.

The final action totals are 34 ACCEPT, 11 MODIFY, 6 UNDECIDED and 2 REMOVE, with zero PENDING and zero NEW. One integrated core remains. The recursive authored/provider census contains 18 distinct PMIDs and four Reactome records, all locally present. All DOI and PMC identifiers in authored notes resolve to those same primary cache identities; no separate DOI-only or preprint gate was found. No provider, raw provider HTML or provider PDF was generated. Immutable UniProt bibliography alone is not expanded into new citations. Final validation, source-object and case-sensitive quote checks, append-only history and rendering are recorded in the follow-up manifest. Local cache closure does not imply that the unresolved rat assays are now full-text verified.

Full targeted validation exited 0 with no review warnings after source recovery. All 92 ordinary supporting excerpts match canonical text case-sensitively after whitespace normalization; no quote requires case folding. The local citation census has zero missing PMID/Reactome records. Status is therefore COMPLETE; the 6 deliberately unresolved response judgments remain transparent and do not become positive biological assertions. The final status change is revalidated below the same scoped workflow; PR publication and review/CI decisions remain with the coordinator.