Deep research: falcon failed (Edison API 429, 2026-10-05).
This dated audit supersedes conflicting earlier shorthand. All 73 original annotation payloads, including qualifiers and inherited NEW rows, were retained exactly. Every row was reconsidered against the target biology and cited sources. Actual Falcon and fallback provider attempts failed; the orchestration logs are preserved in .cache/module-audit-logs/auxin-polar-fetch/provider-publication-results.json. Existing provider reports, where present, served only as discovery aids. Manual primary-source inspection and normally fetched caches support the decisions below. No provider report was fabricated.
Direct IAA-export evidence is retained, but auxin-only specificity and unqualified ATPase/IAA coupling are not asserted. Full PMID:38513023 plus supplementary text was recovered normally from the Ghent author manuscript at publications/url_https___biblio.ugent.be_publication_01HTMBK3P3PPNB73YFCMFXP3ZZ_file_01HTMC3Z7XJ6XPXCM124Z4RD35.pdf.md. It provides direct brassinosteroid transport and BR-stimulated ATPase activity; its purified binding/ATPase/proteoliposome IAA-negative results coexist with a positive ABCB19+TWD1 HEK293F cellular IAA loading/export assay (Fig. S3F). A negative ATPase stimulation result is not by itself a transport assay. PMID:39497419 mainly studies ABCB1 but contains ABCB19 comparator assays; those source-specific distinctions replace the old assumption that any ABCB19 result was merely likely.
The 28887381 cytosol claim is MARK_AS_OVER_ANNOTATED after all seven actual XLSX supplements were inspected, not because a 12-TM protein cannot have a cytoplasmic pool. ABCB19 S1 rows 2145/2804 are True membrane, with blank dual-localization flags; the material described as soluble is cholate-solubilized microsomes, not cytosol. Imported SUBAcon labels are not new localization measurements, and a blank dual-localization flag does not prove absence of any minor pool. Reproducible download/extraction, exact hits and SHA-256 provenance are in ABCB19-bioinformatics/. Nucleus prediction is UNDECIDED, superseding the old membrane-exclusivity REMOVE.
Full PMID:35512423 was inspected at publications/url_https___oup.silverchair-cdn.com_article-minimal_6581225.md. Native ABCB19–GFP occupies filament epidermis/provasculature at stage 12; abcb1 abcb19 has shorter filament cells, reduced pollen number/viability and abnormal nuclei. These support contextual morphogenesis and gametophyte contributions, not exclusive ABCB19 necessity. The failed cell-plate/callose aggregation, stamen-number, nectary and early-dehiscence effects are TWD1-specific. GO:0090627 requires acquisition of epidermal identity; measured growth of existing epidermal cells does not establish it, so both corresponding rows are overannotated. The independent leaf study PMID:32855213 concerns petiole angle/flattening rather than primordium phyllotactic patterning.
Both GO:0043481 UV-anthocyanin rows are UNDECIDED. Live GO requires UV stimulus; the accessible 15908594 abstract/publisher preview explicitly supports pigment changes and far-red/red/blue responses, but cannot establish UV treatment. PMC was browser-challenged and Europe PMC returned 403. This is not a claim that the inaccessible full study lacks a relevant experiment. Broad membrane/ATPase/transmembrane IBA refinements preserve the genuine ancestral properties; actual PTHR43394 PTN002517083/PTN000657189 entries include ABCB19 among legitimate experimental seeds, not circular support.
Final actions: 22 ACCEPT, 37 KEEP_AS_NON_CORE, 4 MARK_AS_OVER_ANNOTATED, 6 MODIFY, 1 REMOVE, 3 UNDECIDED. This panel contains no negated source rows; there are no manufactured NOT assertions. Exact-row payload proof and focused validation are recorded in .cache/module-audit-logs/auxin-polar-six-*. Reciprocal scientific consultation with the parallel annotation reviewer confirmed source-specific localization, integral binding and substrate-assay distinctions.
Full15908594 recovered after the earlier access-limited assessment. Results/Fig.3C and Methods were independently read: anthocyanin assays use continuous far-red/blue illumination; UV-A is introductory photoreceptor context. Existing curated function supports light-dependent pigments without UV specificity, and a targeted ABCB19/MDR1 UV literature search found no independent target UV-pigment assay. Both UV rows remain UNDECIDED because the biological claim is neither established nor contradicted. This replaces the obsolete inaccessible-full-text rationale; it does not convert non-UV experiments into negative UV evidence. Normal full source: publications/url_https___pmc.ncbi.nlm.nih.gov_articles_PMC1150410_.md. No action or annotation payload changed.
The live QuickGO definition of GO:1900459 is positive regulation of brassinosteroid-mediated signaling, not homeostasis. The normally cached PMID:38513023 author manuscript (Ghent PDF, Results/Fig. 4 and Fig. S13) reports reduced dephosphorylated BES1 in abcb19 and double mutants and increased readout in ABCB1/ABCB19 overexpression plants. Full PMID:39497419 adds ABCB1 ATP-dependent BL transport and protoplast CSA export. Both paralogs retain three supported GO:1900459 rows as KEEP_AS_NON_CORE; direct BR transport remains a core MF, without a core signaling BP. This is a core-function scope decision, not a claim that upstream ligand supply cannot positively regulate signaling. No native ABCB1 gain-of-function experiment is invented: the source includes overexpression lines and gain-of-function BES1 downstream rescue.
The AtSubP GO_REF:0000122 ISM location prediction is now MARK_AS_OVER_ANNOTATED. It was assessed as a computational compartment assignment against positive target localization and carrier architecture, rather than deferred under the rule for inaccessible experimental papers. Native imaging and the multipass carrier architecture support plasma-membrane transport; the classifier provides no independently established nuclear targeting, nuclear-envelope transport role or cleaved nuclear fragment. Nuclear membranes can contain integral membrane proteins, so topology is not claimed to make every nuclear pool impossible. This revises the earlier unresolved judgment without inventing an exclusion assay or asserting loss of a phylogenetically inherited location. All original row payloads, flags, evidence codes and raw source files are preserved; no new annotation is proposed.