Author: Codex (/root/bloc_followup, GPT-6). This is an assessment of the imported normal source record, pending a distinct whole-science peer before canonical application. It is not external maintainer approval.
The imported human record is reviewed UniProt Q96G97 (398 amino acids), HGNC:15832, NCBI Gene 26580. The 30 seeded GO assertions, their evidence/reference/partner fields, all 17 reference identities and titles, and all three alternative products are retained. The product IDs are Q96G97-2, Q96G97-3 and Q96G97-4; a Q96G97-1 product is not invented. There are no seeded NOT or isoform-specific annotation flags. The normal source and publication caches are immutable.
The ordinary research command used the installed exact deep-research client 0.2.7rc1 with falcon and the normal perplexity-lite fallback. Both provider calls failed at DNS resolution; no provider output was created. The ordinary PMID caching command passed with all 12 PMIDs already cached. These notes are manual work, not provider-generated research. The reference set comprises 12 PMIDs and five GO_REF records.
Seipin is an oligomeric ER membrane protein that contributes physical structure to lipid-droplet assembly and growth. Human structural work reports an anionic-phospholipid-binding lumenal domain and an undecameric assembly (PMID:30293840(https://pubmed.ncbi.nlm.nih.gov/30293840/), complete cached abstract). Human live-cell work places the seipin–LDAF1 machinery at droplet initiation sites and shows that recruited machinery can redirect those sites (PMID:31708432(https://pubmed.ncbi.nlm.nih.gov/31708432/), relevant Results/Methods read). Seipin remains ER-associated while LDAF1 can move onto the growing droplet. Human knockout, rescue, endogenous tagging and cargo-exchange experiments connect seipin to functional ER–LD contacts (PMID:27879284(https://pubmed.ncbi.nlm.nih.gov/27879284/)). These are participation data: seipin supplies part of the organizing apparatus rather than merely being a substrate consumed by another process.
The proposed single core combines phospholipid binding with lipid-droplet formation. GO:0140042 is an is_a child of lipid droplet organization and part_of lipid storage in the inspected official PomBase ontology view. The broader processes therefore do not need duplicate core units. The study's language about facilitating triglyceride phase separation does not establish triglyceride-synthesis chemistry by seipin. AGPAT2 catalysis, downstream desaturation changes, and adipocyte lipase activities are not transferred to BSCL2.
The broad GO:0005811 definition allows lipid-droplet-associated proteins. Thus seipin's ER-embedded contact-site position does not make the old term intrinsically false. For the direct PMID:27879284 localization row, GO:0160259, endoplasmic reticulum membrane-lipid droplet contact site, is more precise. The official definition names the contact between the two structures; its links to ER membrane and lipid droplet are overlaps, not is_a. The candidate changes only this row's proposed review action/term, leaving its source term intact. The original PMID:30970241 localization experiment is not exposed in the abstract and remains UNDECIDED. The broad UniProt location mapping is retained with the contact-site interpretation.
A disclosed annotation-reviewer consultation by /root/bloc_followup/brca1_late_binding checked the eight original protein-binding rows against their actual source and partner fields. PMID:30901948 supports cellular seipin–promethin/LDAF1 co-isolation; it includes tagged human forms and separate mammalian-cell contexts. PMID:31708432 adds endogenous human association and recombinant complex isolation. These two supported assertions are KEEP_AS_NON_CORE. Cellular coassociation is not described as a purified binary affinity measurement.
The specific TMEM19/SMIM3 edges from PMID:16189514 and PMID:25416956 are UNDECIDED because their target-level screen records were not inspected. The canonical PMID:25416956 cache has full_text_available: true but only the abstract and Introduction/Discussion excerpt; that flag is not a claim that its Results or interaction table was read. Its fuller fetched variant remains separate and was not installed or used to invent target results. The PMID:30970241 abstract states a DFCP1–Rab18/ZW10 interaction and a BSCL2 perturbation phenotype, but does not show the two seipin-specific binding assays; those assertions also remain UNDECIDED. No paper is called wrong-gene merely because its title emphasizes DFCP1.
Keeping the two supported generic-binding assertions follows the user's explicit ClinGen task instruction, documented in the campaign curation policy. It is a task-specific deviation from the skill's generic uninformative-binding removal suggestion, not fabricated repository-wide consensus or maintainer sign-off. No unsupported specific binding function is invented to avoid that policy distinction.
The GOA IBA node is PANTHER:PTN000476621. The official FlyBase seipin record independently exposes that exact node for ER membrane, lipid storage and lipid-droplet organization. The underlying PAINT IBD, complete phylogeny and MSA were not recovered; no new family or ancestral placement is asserted. The candidate accepts those three functions from direct human biology, while recording the ancestral source as unresolved for this independent inspection. Neither the number of donors nor human BSCL2's presence among them is treated as a defect. The local GO-CAM index had no Q96G97/BSCL2 match in the inspected snapshot; this absence is not evidence for a missing annotation.
For the ISS source Q9Z2E9 (mouse Bscl2), the official MGI GO view exposes experimental fat-cell differentiation and negative lipid-catabolism regulation. MGI J:183701 resolves to PMID:22269949(https://pubmed.ncbi.nlm.nih.gov/22269949/). The full indexed primary abstract was read: differentiating mouse cells show excessive lipolysis and impaired terminal adipocyte differentiation, with rescue by lipolysis inhibitors. Full Methods/Results access failed, and no canonical cache or YAML PMID reference was manufactured. J:198139 is another MGI differentiation reference whose identity was not resolved. The exact historical donor selected for the human GO_REF transfer is therefore not claimed to have been reconstructed.
These differentiation/lipolysis transfers remain non-core and context dependent. The cached PMID:27748422 mouse and brown-preadipocyte experiments were read in detail. They show largely normal brown-cell differentiation and lipolysis alongside impaired cold-acclimated thermogenic capacity in the organism. The paper favors a substantial systemic contribution, including substrate supply and insulin resistance; compensatory shivering and maintained fed core temperature do not erase the measured thermogenesis deficit. This justifies retaining the thermogenesis transfer as a bounded non-core physiological annotation, without asserting a direct heat-producing activity or a direct human experiment.
The candidate reviews all 30 source assertions: 17 ACCEPT, five KEEP_AS_NON_CORE, one MODIFY and seven UNDECIDED. It proposes one coherent core and zero NEW annotations. The three core anchors use seven, six and seven words from PMID:30293840, PMID:27879284 and PMID:31708432 respectively; no other source quotations are added here. All caches, raw source files, original references and product descriptors remain unchanged. Status is DRAFT because source-specific uncertainties and advisory warnings remain; campaign completion, if later reached, would not resolve those biological limitations.
The next useful evidence would be target-level screen/interaction records, the original DFCP1-paper seipin localization/interaction experiments, and the complete PAINT/donor chain. These gaps are stated as questions rather than prompting speculative source fetches or fabricated results.
Normal candidate schema, term and best-practice validation passed on 2026-10-03 with three warnings. Two are the documented generic-binding policy deviation. The third reports different actions for GO:0005811; those differences are deliberate source-specific judgments: a directly resolved contact-site localization is refined, a source experiment not exposed in its abstract remains unresolved, and the broad UniProt-associated location mapping remains acceptable. DRAFT is retained. No global validation, canonical application or history creation is claimed at this stage.
The normal hosted publication fetch and authenticated artifact supplied the genuine cache for PMID:22269949(https://pubmed.ncbi.nlm.nih.gov/22269949/), previously inspected only through the indexed primary abstract and official MGI mapping. The cache contains the complete abstract; normal PMC XML, HTML and PDF access did not recover full text. No full Methods/Results reading is claimed. Its mouse MEF/SVC terminal-differentiation and lipolysis findings corroborate the existing contextual non-core judgments in source rows 23 and 24 (zero-based); the exact historical ISS donor selection and unresolved J:198139 remain unchanged.
The original 17 references remain intact, with this one verified publication added: 18 total, comprising 13 PMIDs and five GO_REFs. Reference-only support was added to the two existing judgments and their access wording was corrected. All 30 source objects, all actions, three alternative products, one core function and existing quotations remain unchanged. No NEW annotation or additional quotation was introduced. The earlier dated access account records the prior state and is preserved.
ROOT completed the distinct whole-science peer of all 30 decisions, 17 original references, three products, notes and the single core; a distinct focused peer covered the later normal donor-cache/reference delta (18 references total). The exact reviewed YAML and notes were applied after checking source preimages. Normal canonical validation passed with three advisories: two task-policy generic-binding warnings and the documented source-specific localization action difference. A standard codex/gpt-6 CREATE history was scaffolded and validated. Status remains DRAFT with seven UNDECIDED assertions; uncertainty was not converted into evidence of absence. The earlier candidate-stage statements above record the prior stage. Raw source records, existing publication variants and previous histories were preserved. No provider output, remote publication or external maintainer approval is claimed by this application.
The current-head review of PR #3902 (e835ac3972579669a3d6c809f50099527ad5d270) identified two scientific issues. This follow-up keeps every one of the 30 original source objects, all 18 references and three alternative products. The sole original action change is the thermogenesis ISS row from KEEP_AS_NON_CORE to MARK_AS_OVER_ANNOTATED; seven source-specific UNDECIDED judgments remain.
The official GO:0120162 definition and parents permit organismal regulation, so non-cell-autonomy alone cannot invalidate this annotation. Re-reading the unique Methods/Results and Discussion of cached PMID:27748422 shows impaired maximal thermogenic capacity in global knockout mice, alongside largely preserved cultured brown-adipocyte respiration, lipolysis and insulin signaling. Some thermogenic transcripts and UCP1 protein do decrease: the culture findings are not uniformly negative. The authors favor systemic white-fat substrate/adipokine failure and secondary insulin resistance. The narrower judgment is that these experiments do not separate a specific thermogenic regulatory step from the distal lipodystrophy phenotype. The donor identity remains sound; the revised propagation assessment concerns term scope.
The official GO:0170007 definition names ER–lipid droplet tether activity and places it under GO:0043495 protein-membrane adaptor activity. It is not a phospholipid-binding descendant. One NEW molecular-function annotation records seipin’s structural contact role, supported by the human knockout/rescue, endogenous imaging, ultrastructure and cargo-exchange experiments in cached PMID:27879284, with independent contact manipulation and lipid-binding evidence. The knockout phenotype includes missing contacts in a subset of droplets and heterogeneous residual contacts; it is not summarized as loss of every contact. This supplies physical participation, beyond a necessity-only inference, without adding a biological process or an adaptor ancestor. No LDAF1–seipin complex identifier is invented.
Canonical PMID:31178403 remains abstract-only. Relevant human A431 tagging/relocalization/degron and cellular-droplet Results/Methods were additionally read in the institutional author manuscript. The reconstitution uses isolated cellular droplets, not purified human seipin alone. No full supplementary-image inspection or all-isoform equivalence is claimed. Prior fly-specific experiments remain attributed as before. The cache is unchanged, and reference-only support avoids quoting uncached text.
Two cooperating cores now distinguish seipin’s own anionic-phospholipid recognition from its own ER–droplet contact activity. They share the existing formation process; neither is expressed as contributes_to another protein’s molecular function. GO:0140042 remains the sole structured process per core, with the previously verified organization/storage coverage unchanged. The three existing core quotations are redistributed intact, with zero additional quotation words. The broad IEA droplet association is explicitly retained at the source mapping’s resolution, while direct contact-site imaging supports the finer IDA replacement. Partner names aid readability without changing source accessions, and unrelated catalytic negations were removed from selected reasons.
Normal candidate schema, term and best-practice validation passed with the same three explained advisories (two retained-binding policy warnings and the source-specific LD localization difference). A final line-wrap normalization preserves the exact validated parsed document. This TMP proposal awaits ROOT’s distinct focused science peer. No canonical source, cache, previous journal entry or history record is altered by its preparation.
After ROOT’s distinct focused science PASS, the exact proposed YAML and appended notes were applied. Normal canonical validation passed with the same three explained advisories. Targeted HTML rendering and a newly scaffolded codex/gpt-6 EDIT history passed their checks. All 30 original source objects, 18 references, three alternative products, prior history and source/cache bytes remain intact. The current 31 decisions include one NEW tether activity and seven UNDECIDED assertions; DRAFT is retained. Remote publication is a separate ROOT-owned step.