No falcon deep-research file was present during review.
Reviewed every original row (21) and core function using review/annotation-reviewer instructions, current cached primary publications and the existing research. Preserved all source annotation fields and all pre-existing proposed rows. Reused the existing genuine Falcon provider report and its artifacts. These plant targets do not require bacterial FEBA retrieval.
Native redundancy withTGG1 is directly supported by16640593; loss of both changes damage-induced hydrolysis and some generalist herbivore outcomes. The19703694 purified isoforms areTGG1/4/5, so TGG2 secondary O-glucosidase remains inherited/comparative, not claimed directly assayed there. Refined O-hydrolase to the already-present beta-glucosidase term. Unchecked chloroplast/plastid/ribosome data remain undecided. No new seed-turnover process is inferred:31850033 explicitly does not findTGG1/2 responsible for the germination decline.
Broad and substrate-specific terms were assessed consistently across evidence codes. No NEW rows were added. Previously proposed rows were evaluated for direct participation rather than filled from pathway adjacency. Experimental cohort calls whose protein-level data were not verified were left UNDECIDED; full-text availability alone was not treated as proof of a target-specific assay. Original sources and provider reports were not hand-edited.
Kept location claims consistent with retained experimental compartments: a supported apoplast assignment refines a general extracellular prediction, while a main cytosolic or vacuolar location does not by itself disprove a secondary predicted pool. Conditional Arabidopsis ESP activity is distinguished from a universal coding knockout. Original annotation fields remain unchanged.
Revisited review findings 1–3 against the actual publication cache and the primary sources. Removed generic source-unresolved boilerplate where the assay design is adjudicable. Restored relevant verbatim biochemical and localization evidence; did not restore label-only CHAIN features, unrelated introductory statements, or quotations that no longer support the revised claim. Existing source annotation fields were preserved.
Primary supplemental check for PMID:18431481: annotated-protein workbook s005, worksheet supplTable5, and abundance workbook s003, worksheet supplTable1, explicitly classify AT5G25980 as "not plastid". Internal workbook titles and article table numbering differ, so file identifier plus locus are the stable join. These are direct reads of publisher supplements, not an inference from absent targeting signals. Working downloads are under .cache/module-audit-logs/18431481-s003 and 18431481-s005; source caches were not manually modified.
Read PMID:28887381 full text, especially "Identification of True Membrane-associated Proteins" and Fig. 7: washed microsomes and sucrose gradients distinguish membrane association but do not separate organelles; named compartments are imported from SUBAcon/GFP knowledge. The paper states "no organelle fell in a single cluster". Mark the specific compartment claim as over-annotated while preserving proteomic detection. This applies projects/ORGANELLES.md at the actual resolution of the assay.