Reviewed all 26 source annotation rows, the cited publication abstracts, relevant available full text, UniProt and the existing Falcon synthesis. Preserved all GOA-sourced fields. QuickGO confirmed the definitions of GO:0090628 (plant epidermal cell fate specification), GO:0000976 (transcription cis-regulatory region binding), and GO:0010154 (fruit development). Electronic corrections use explicit biological rationale and cited primary evidence without a blanket snippet requirement.
GL1 is the DNA-binding MYB component of the nuclear trichome regulatory complex, with direct regulatory-promoter occupancy demonstrated in PMID:18434419. Functional GFP-tagged localization is direct evidence even though the previous Falcon synthesis did not retrieve it: PMID:14561633. Core-function support now cites these experiments rather than a generated report or a paper title.
Replaced interchangeable interaction summaries with the actual experiments and partners: GL3 binding (PMID:11063707), EGL3 binding (PMID:12917293), gl3-sst/TRY competition (PMID:14561633), At1g01380 inhibition of GL1-GL3 (PMID:15584952), JAZ interactions (PMID:21551388), DELLA interactions (PMID:24659329), and purified GL1/GL3/PAP1 affinity measurements (PMID:32844153). Direct GL1-TTG1 binary interaction must not be inferred from complex membership: PMID:11063707. The PAP1 interaction assay does not establish an anthocyanin biosynthetic role for GL1.
The protein-localization annotation specifically concerns GL3 distribution: PMID:18434419. Gibberellin and jasmonate contexts are now consistently retained as non-core transcriptional signaling outputs. GA evidence is more than mere correlation: PMID:9625690, complemented by direct DELLA interaction evidence in PMID:24659329. Refreshed PMID:21551388 full text directly confirms disruption of JA-induced trichome initiation in gl1 mutants.
The fruit-development citation remains wrong, but the decision is changed from REMOVE to UNDECIDED: a broken citation does not establish absence of the biological phenotype. The cached abstract explicitly concerns Campylobacter plasmids, and a live Europe PMC identifier lookup confirms PMID:3793867 is the 1987 bacteriology paper. The reference is marked WRONG_IDENTIFIER/NONE; no replacement PMID is guessed. Intended plant evidence requires recovery from the source curator.
Broad epidermis development is a true parent of the core trichome role, not automatically peripheral. The overly broad system-development inference is narrowed to the already experimentally supported trichome differentiation term.
Validation passes with an advisory warning that annotation rows do not cite the deep-research report. Primary publications provide the quotations; adding a generated-report paraphrase solely to suppress that warning would reduce evidence quality.
The required fresh research attempt failed: Falcon timed out after 600 seconds, then the requested perplexity-lite fallback returned HTTP 401 with insufficient_quota. The pre-existing Falcon report remains unchanged; no manual text was substituted into a provider-named report. Re-review proceeded from existing synthesis and primary publications. Concurrent publication caching succeeded; pipeline refresh obtained full text for PMID:21551388, while PMID:12244266 remains abstract-only. The provider failure does not change the source-supported core conclusions, but no successful new provider synthesis is claimed.
Applied the annotation-reviewer skill to all seven remaining generic protein-binding MARK_AS_OVER_ANNOTATED rows, changing them to REMOVE as uninformative functional labels. Their source-specific summaries, experimental qualifiers, and verified interaction quotes are preserved. These are interactions with transcriptional partners or regulators; this action neither rejects those observations nor invents an additional molecular activity from binding alone. For example, [PMID:11063707, "Yeast two-hybrid assays indicate that GL3 participates in physical interactions with GL1, TTG1, and itself, but that GL1 and TTG1 do not interact."] remains mechanistic evidence. Existing unresolved evidence judgments remain unchanged. All source identities were preserved. No new research job was needed for this bounded policy application.
reference_review.replacement (reason WRONG_IDENTIFIER) on PMID:3793867; original_reference_id kept.