The decision table retains each complete imported source object, its index and the current direct review. The product-context table separates natural splice records from fragments, rescue constructs and disease-associated bands. No publication quotations or new annotation rows were added.
PMID:19234460(https://pmc.ncbi.nlm.nih.gov/articles/PMC2857384/) provides human platelet/leukocyte measurements and KINDLIN3 rescue/knockdown. Read Figures 1C,D and 2F–H for activation-dependent antibody/ligand binding with preserved receptor abundance; Figures 2B,C for leukocyte aggregation; and Figures 1B and 4C–F for substrate contacts and spreading. Main Methods, “Cell lines and kindlin rescue experiments,” identifies IMAGE clone 6709978 as the rescue source. The actual main Results and Methods and Figures 1–4 captions were read; supplementary panels and movies were not independently inspected. P-selectin induction and several upstream signaling responses remain present, so the defect is not loss of all signaling. A patient 56-kDa band and proposed downstream initiation do not define a normal splice product. The osteopetrosis mechanism is exploratory.
PMID:19234463(https://pmc.ncbi.nlm.nih.gov/articles/PMC2680140/) independently identifies human FERMT3 lesions, reduced protein expression and defective adhesive contacts and migration. Figure 4A measures contact with immobilized ICAM-1; Figures 4B,C measure migration. The Methods explicitly calls the rescuing construct mouse EGFP-KINDLIN3, expressed in human patient B cells. Human genetics, recipient-cell behavior and cross-species complementation are distinct evidence. The human 663-residue diagram in Figure 2 is not an exact sequence join to the mouse rescue construct or proof of a current human splice-product-specific effect. Main Results/Methods and all four figure captions were read; supplementary panels and movies were not.
Integrin activation is retained as the directly resolved process. Leukocyte adhesion, migration, regulation of integrin-mediated cell-cell adhesion, signaling-associated spreading and platelet behavior remain positive contextual annotations, without inventing separate enzymatic activities. For cell-cell adhesion, the living-cell aggregation results in PMID:19234460 complement the ICAM-1-coated substrate assays in PMID:19234463. Inside-out activation alone is not treated as proof of every outside-in signaling event.
PMID:27101375(https://pmc.ncbi.nlm.nih.gov/articles/PMC4839668/) directly tests the recombinant Leu142–Gln224 F1 loop. Methods describes the fragment, vesicles and NMR measurements; Results and Figure 3 report interaction with POPS and POPG, with much weaker POPC perturbation. These substrates support GO:0005543 phospholipid binding, a verified descendant of lipid binding. Selecting only phosphatidylserine binding would omit the tested phosphatidylglycerol interaction. This paper does not measure phosphoinositide-specific binding by the PH domain.
Main Methods, Results, Discussion and all six captions were read. Figure 6 is explicitly hypothetical, and the text does not establish its full-length membrane-orientation model experimentally. The loop endpoints agree with the current human sequence, but Methods does not state its donor organism. The printed EK21 peptide contains 22 residues despite its stated 21-residue E146–K166 description; it has an extra initial E compared with current human residues 146–166. That discrepancy is preserved in the product-context table, not converted into an inferred natural isoform, a negative binding result or a source-defect assertion. Neither fragment is a newly authored natural product.
The normal cache for PMID:23382103(https://pmc.ncbi.nlm.nih.gov/articles/PMC3650338/) is marked full text but contains mainly abstract and Discussion. The complete original PMC HTML was separately recovered. Main Methods, Results, Table I and Figures 1–8 captions were read. Table I, spot 6, identifies URP2_HUMAN/Kindlin-3, reduced 2.07-fold in MDS platelets. Protein identification methods specify human database searches and discuss ambiguous family/isoform reporting; supplemental peptide coverage was not inspected.
Figures 5–8 measure talin–integrin proximity, receptor activation, spreading and aggregation. The FRET assay is talin-specific, and the study changes several platelet proteins without selectively perturbing FERMT3. Thus the platelet annotation is retained as non-core contextual support, reinforced by the human FERMT3-deficiency data in PMID:19234460 Figure 1; the MDS phenotype is not attributed uniquely to kindlin-3. Failed XML/BioC/PDF attempts do not negate the successful original HTML recovery.
The final article and supplement for PMID:19946888(https://doi.org/10.1002/jms.1696) were not recovered. Related primary evidence is available in the first author’s institutional 2008 thesis, A Proteomic and Physicochemical Approach in Defining Membrane Proteins. Its human YTS membrane preparation and Chapter 3, section 7.2.8.1, were read. Table S7, printed page 304/PDF page 323, and Table S8, printed page 378/PDF page 397, contain ENSP00000279227, named Kindlin-3/Unc-112-related protein 2; both target rows were visually inspected. The thesis reports 2278 combined identifications, whereas the published abstract reports 1843, so the thesis tables are not silently substituted for the final published supplement.
The related positive membrane-fraction detections, human contact localization and phospholipid interactions support retaining the broad gene-level membrane annotation with curator deference. The exact published target join remains unverified. Neither the fractionation nor the current Ensembl crossreference proves integral membrane topology, exclusive plasma-membrane residence, historical sequence identity or isoform-unique peptides.
For PMID:20458337(https://doi.org/10.1038/icb.2010.64), the exact title/authors/DOI and operative Results/Methods were read from the coauthor-posted original article. “Identification of exosomal proteins” and the Methods describe human RN B-cell exosomes, 539 identified proteins, and Supplementary Table 1. The exact FERMT3 entry in that table remains inaccessible, so the exosome annotation stays UNDECIDED. Main Table 1 describes a distinct detergent-isolated MHC-II-associated subset. Neither broad membrane association nor hypothetical cargo biology establishes the missing exosome target join, MHC-II binding or a role in vesicle biogenesis. No wrong-gene or negative-result claim is made.
All six immutable publication caches are retained, including their genuine availability flags. Every source annotation, unowned database review, reference identity/finding and both official splice-product objects remain unchanged outside the assigned review fields and complete manual reference assessments. The two official splice products differ at residues 360–363; fragment assays and engineered rescue constructs are not automatically assigned to either natural product.
The genuine Falcon attempt and its configured fallback failed during intake. No provider report was fabricated. This authored guide records the actual evidence and limitations; it is not an automated provider output. Shared database scope, core synthesis and final source adoption are handled separately during integration.