2026-10-01 — Initial manual review proposal by science_followups, pending independent ROOT science review. The normal seed contains 30 source annotations from 31 raw GOA rows and has no alternative_products slot. All source fields, source order, raw files and existing caches are preserved; no products are invented. No NEW annotations are proposed.
BLOS3 is a conserved component of BLOC-1. The core synthesis describes the assembled complex supporting endosome-to-melanosome cargo delivery and platelet dense-granule organization, with cytosolic localization. It leaves a BLOS3-specific molecular activity unresolved. No claim of protein translocation through a membrane, autonomous motor activity, isolated adaptor activity or BORC membership is made.
The four original normal PMID caches (15102850,16385460,17182842,22203680) are abstract-only. Every complete abstract and the Reactome429815 summary were read. Both newly created caches 16385460/17182842 were imported exactly from authenticated Seed54; different existing 15102850/22203680 caches were preserved. Auxiliary apply 46cbaa/post 393b1b receipts are in tmp/seed54-auxiliary-import-*.json. The normal UniProt record and all 30 seeded annotations were read in full.
Primary evidence beyond the normal abstracts was inspected within these limits:
Official AmiGO definitions/parents were checked for transmembrane protein transporter activity0008320, protein transmembrane transport0071806, endosome-to-melanosome transport0035646, AP-3 adaptor complex binding0035651, BLOC-1 complex0031083 and melanosome transport0032402. The two transmembrane annotations are rejected because the demonstrated mechanism is endosomal cargo sorting rather than crossing the lipid bilayer. The historical GONUTS mouse record explicitly marks its16837549/0008320 CACAO annotation incorrect, strengthening the source-side interpretation. That record also connects neuronal donor assertions to21998198 and19546860. It is an indexed historical snapshot, not a reconstructed current QuickGO/PAINT graph. Direct donor/API attempts failed; no claim of current donor deletion is made. PAINT PTN001592525/527 node placement was not independently reconstructed. Target-self evidence is legitimate; no circularity claim is made. The local PTHR31974 PAINT file was absent and gocams/index.tsv had no target hit.
The proposed decisions are 13 ACCEPT,13 KEEP_AS_NON_CORE,2 REMOVE (only electronic transmembrane assertions),2 UNDECIDED (the exact experimental localization/whole-organelle movement assertions). One core function, seven PMID references. Exact cached quotation counts are15102850=6 words,17182842=8,22203680=11; the latter plus the previously used BLOC1S1 eight-word excerpt remains19. Other supports are source references with explicit prose limits.
One normal Falcon wrapper invocation with configured perplexity-lite fallback was attempted after auxiliary import. Writable UV_CACHE_DIR, UV_TOOL_DIR and UV_TOOL_BIN_DIR were supplied, with workspace .venv first on PATH. Both starts failed at offline dependency resolution for deep-research-client[cyberian]==0.2.7rc1 before either provider ran (851d24, exit1). No provider files appeared; 11 source/seed preimages stayed identical (bdfb38). No retry or provider-named manual report was created. Exact invocation and terminal output are in tmp/BLOC1S3-science-preread/provider-*.json.
Preparation stop 3b245e caught a missing trailing space in a six-word cached quote before any proposal output was written. The helper preimage and honest stop summary are retained; the cached text was not changed. Corrected build f86ab9 passed all exact quote and source-projection checks. Focused TMP validation passed (start bbb896, terminal 71df22) with one intentional generic-binding advisory under the user ActionEnum. Exact quotation, title and authored term checks passed. Source-object equality was checked separately because --no-goa was used for the proposal outside the canonical gene path. Distinct science peer and canonical adoption remain pending.
2026-10-01 — ROOT supplemental primary-literature reading, after initial independent science PASS822a87.
The later BLOC1S5 fetch recovered longer HTML extractions of PMID17182842 and PMID22203680. The canonical publication caches above remain unchanged. ROOT read the recovered Methods, Results and Discussion for both papers; initial displays clipped and the missing ranges were separately read. These extracts omit figure images and most figure captions, and supplements were not inspected. This paragraph updates the earlier reading limits without rewriting their history.
In17182842, human BLOS3 was expressed in mouse melan-rp cells. Figure6 and the associated Results support reduced Tyrp1 surface missorting after rescue. The Results also place reconstituted BLOC-1 in MuHA cells on tubulovesicular structures near the Golgi, melanosomes and endosomal vacuoles (SupplementalFigureS3b/c). That complex-level observation does not identify the exact BLOS3 transport-vesicle experiment, so row9 remains UNDECIDED. The work follows cargo delivery to melanosomes, not whole-melanosome movement; row18 remains UNDECIDED pending the precise experimental basis and omitted supplementary evidence. The2017 correction has been considered; the original transferrin-recycling result is not used as decisive support.
In22203680, recombinant human BLOS3 was present in the purified hetero-octamer. Whole-complex AP-3 pull-down does not isolate a BLOS3 interface. BLOS3 and Muted were shed more readily than the two stable trimers under perturbation; proposed connections to non-core portions of dysbindin remained an interpretation to test. This strengthens BLOC-1 membership without supplying a new BLOS3-specific molecular activity.
Actions remain13ACCEPT,13KEEP_AS_NON_CORE,2REMOVE(electronic transmembrane assertions),2UNDECIDED; one core and noNEW. Only the two unresolved reasons and the two reference reading notes are updated. Existing quotations are unchanged. The earlier quotation note counted the prior BLOC1S1 excerpt conservatively as eight words; its actual count was seven, so the combined22203680 total is18 words. Supplemental independent review precedes canonical application.
2026-10-01 — Completed review checks. Initial whole-gene independent science PASS822a87 and supplemental distinct science PASS91ad99 preceded canonical application3f984b. Focused schema, source GOA, reference and term validation passed (ee5f25) with one intentional generic-binding advisory under the supplied ActionEnum. HTML renderd05699 and history validation4bd2bb passed. The rendered page embeds the exact reviewed YAML. All30 source objects, absent products and raw31-row GOA are preserved. Ten protected source/cache files are unchanged. No successful provider report, figure-image/supplement reading or global validation is claimed.
The current PR3758 review was checked against the existing source assertions and primary evidence. This section supersedes the earlier unqualified retention of axon cytoplasm and the reliance on a historical wiki warning; the prior dated notes remain intact.
Axon cytoplasm (GO:1904115) is a particular neuronal compartment. The complete Discussion paragraph in the canonical PMID:21998198(https://pmc.ncbi.nlm.nih.gov/articles/PMC3237628/) cache was reread. Besides the cell-body cargo-sorting results, it describes earlier detection of BLOC-1/AP-3 subunits in axons and dendrites and a nerve-terminal supracomplex. These observations prevent a claim that upstream sorting proves exclusion from axons. They do not identify BLOS3 in that compartment. The source electronic row therefore becomes UNDECIDED, with neither a new exclusion claim nor an invented imaging experiment. Its source qualifier and transport-derived supporting entities remain unchanged.
The GO:0035651 AP-3 adaptor complex binding definition describes association with AP-3. Selected PMID:22203680(https://pmc.ncbi.nlm.nih.gov/articles/PMC3285357/) expression/purification Methods, Figure 1 textual legend and AP-3 pull-down Results were reread, together with the subcomplex and peripheral-subunit Discussion. The recombinant human octamer binds AP-3 from human and mouse cell lysates with two negative controls; this is positive whole-complex evidence. The inspected assays do not test the effect of selectively omitting BLOS3 on AP-3 binding. Current GO gene-product relation guidance restricts contributes_to to subunits required for that molecular activity. Accordingly, the existing contextual AP-3 row remains and the core knowledge gap now states the precise missing requirement test. No automatic core MF or duplicate NEW annotation is added. No images, supplements or raw assay data were newly inspected.
The BLOS2 association is independently explicit in the canonical UniProt record and remains a supported non-core interaction. Breadth alone does not make the existing experimental assertion false. A more specific activity for that pair has not been established, and its partner must not be exchanged for AP-3 to make the term appear more informative. The generic-binding advisory is retained under the supplied ActionEnum.
Official definitions and displayed parentage confirm that melanosome transport concerns movement of the organelle, whereas endosome to melanosome transport concerns delivery of substances to it. They are not an ancestor/descendant duplicate. The experimental whole-organelle row remains UNDECIDED. The transport-vesicle row also remains UNDECIDED: the previously inspected MuHA complex-localization text is positive context, but its omitted figure/supplement and exact BLOS3 assignment are unresolved. This follow-up does not claim new access to them.
The two electronic transmembrane assignments remain REMOVE because sorting membrane cargo between compartments is not movement of protein across a lipid bilayer. Their current reasons now stand on that mechanism, not the old GONUTS flag. The donor's current status is explicitly unresolved and the first propagation classification is a term-scoping problem. The description spacing is corrected to Hermansky-Pudlak syndrome type 8.
All 30 source objects, their order, absent products/alternative products, reference objects and existing quotations are unchanged. The proposed tally is 13 ACCEPT, 12 KEEP_AS_NON_CORE, 2 REMOVE and 3 UNDECIDED, with one core synthesis and zero NEW annotations. No cached source, raw record, canonical review, existing note or history was overwritten. This TMP proposal awaits independent science review and canonical adoption.
Independent scientific review approved the six rationale refinements and axon-cytoplasm decision. Focused canonical validation passed with one explained generic-binding advisory, rendering passed, and the new history record validated. All 30 source assertions, prior reference quotes and raw sources remain unchanged.
The current BLOC1S3 review contains 13 ACCEPT, 12 KEEP_AS_NON_CORE, 2 REMOVE and 3 UNDECIDED decisions across 30 source assertions, with one core synthesis and no NEW annotations. These are the current totals. Earlier dated entries with 13 non-core and 2 undecided decisions describe the initial review and are superseded by the location follow-up. The two electronic transmembrane removals now rest on the mechanism: sorting membrane cargo between compartments does not translocate a protein across a lipid bilayer. The earlier GONUTS observation remains a historical note and is not the present justification or a claim about the donor's current database status.
Only one generic GO:0005515 protein binding assertion is retained as non-core: BLOS3 association with BLOS2, identified as UniProt Q6QNY1. The normal BLOS3 UniProt record explicitly records this interaction and three experiments. The canonical abstract of PMID:15102850(https://pubmed.ncbi.nlm.nih.gov/15102850/) places BLOS3 in endogenous BLOC-1 co-isolation and a binary-interaction network; the exact original pair experiment was not reconstructed from that abstract. Independent human BLOC-1 reconstitution in PMID:22203680(https://pubmed.ncbi.nlm.nih.gov/22203680/) supports the containing assembly. It does not establish a specific molecular activity of the isolated BLOS3–BLOS2 pair.
The annotation-reviewer guidance and generic-binding validator normally favor a supported, more informative molecular-function replacement or removal for lack of functional information. They explicitly distinguish that removal from declaring an interaction false. This is a new review, so the legacy allowance does not apply. Its one binding advisory is a real departure from that default, rather than evidence that the action complies with it.
The task instruction for this review is to preserve supported generic interactions outside the core when no justified specific replacement is available, and not to remove them merely because the term is uninformative. The BLOS2 association has positive support and no established specific replacement activity. It therefore remains KEEP_AS_NON_CORE under that instruction. Reclassifying it as UNDECIDED solely to obtain a permitted action would obscure the positive assessment already made. This gene-specific explanation does not change the general skill or validator.
The retained row records a source-specific association with BLOS2; the accepted BLOC-1 membership rows describe participation in an assembly. AP-3 association is a separate assertion involving another partner, so it cannot replace the BLOS2 row. Whole-octamer AP-3 assays do not isolate a requirement for BLOS3 in that molecular activity. The existing knowledge gap, three unresolved location/transport assertions and two mechanistically supported electronic removals remain unchanged. No new pair validation, full-paper reading, figure-image or supplement inspection is claimed by this documentation follow-up.