The normal fetch supplied 456 annotations, including 423 literal protein-binding pairs. All source dictionaries, accession/isoform fields, qualifiers and references remain intact. The completed review contains 295 MODIFY, 33 KEEP_AS_NON_CORE, 122 UNDECIDED and 6 ACCEPT decisions, with one integrated postsynaptic scaffold core and no NEW annotations. Three official human products were reconstructed from sequence features and independently matched to official FASTA sequences.
The public evidence guide gives original-source URLs, assay/construct limitations, the fragmentomics crosswalk, exact AP-MS edges and propagation provenance. The project binding policy is applied to supported associations. Generic binding is not removed merely for being generic. Different literal partners can have different actions; no disagreement is based solely on evidence code.
The full SD1 workbook was reconstructed independently. Zero/below-threshold peptide measurements do not establish biological noninteraction. Positive measurements support fragment-specific peptide binding, including modification and sequence-mapping limits. The RPS6KA1 -1 result is qualified as composite-positive; the discordant -3 LYSC value is retained. ARHGAP17 and DLGAP4 remain non-core associations because independent co-complex evidence survives the undetected-tail result. AP-MS-only KCNA4 and ARHGEF26 are likewise retained without unsupported direct-recognition refinements.
The sole core integrates structural contribution to the postsynaptic density and receptor localization to synapse. Actual original human/rodent constructs, rodent splice experiments, and the current short human isoforms are distinguished. Epithelial polarity, kinase/phosphatase associations and developmental outcomes remain contextual. The 1997 proliferation original was not recovered; the publisher response was an abstract-only HTML preview despite the attempted PDF filename. The tear-proteomics target-specific identification also remains unresolved. Neither uncertainty is a removal or a claim of curator error.
Configured provider attempts and their failures are retained in the ordinary bootstrap provenance. No provider-named research content was invented. The protected main publication caches remain unchanged; separately recovered originals are used only with explicit access descriptions. The existing repository and legacy workspace content were untouched by this TMP review.
The blanket peptide-binding replacements are superseded. The positive synthetic domain–motif measurements remain supported observations, but the generic reagent-binding term does not add their biological role. Keep these associations as non-core except where independent target-specific context supports the existing GluN2, FRMPD4/Preso or ERBB4 partner-class interpretation. The assay and product limits remain explicit. Imported parent-protein identifiers describe the source of the tested fragments and remain unchanged.
The 120 unresolved fragmentomics pairs retain distinct measurement or identity uncertainty. UNDECIDED is permitted for an unclear assertion, even when the paper is accessible. A zero selected composite value is neither a measured zero affinity nor proof of noninteraction. The new reproducible evidence artifact provides the original workbook rows, every assay value, sequence inputs and parser. Its counts are 1,344 raw DLG3 measurements; 1,221 accession-candidate joins, including 1,200 natural-ten-mer sequence matches and 21 unmatched joins; and 1,203 sequence-supported scientific selections after matching three AXIN1 internal-peptide rows. Selection is not a positive-binding decision.
The earlier tear-fluid access limitation is now resolved at the protein-identification-list level. PMID:22664934, Table II explicitly includes DLG3. Schirmer-strip collection, detergent extraction of pooled samples, gel separation and mass-spectrometric identification support contextual tear-fluid detection, now retained as non-core. The table does not supply a DLG3 peptide sequence or score, and neither a secretion route nor an extracellular biological function is inferred. Absence of a classical signal peptide alone does not invalidate an extracellular detection.
PDZ-mediated NMDA-receptor-tail recognition is described within the single postsynaptic scaffold core, with a direct receptor-recognition source anchor added there. It is not counted as a separate biological unit. Natural short-product descriptions now contain sequence/domain facts; their functional scope remains unresolved, and they are not equated with the historical rat splice constructs.
The vitronectin and EIF4H-2 HuRI rows are now retained as positive non-core associations. The array, prey-pooling and validated-two-hybrid records represent stages of the same study, not three independent physiological replications. The original study distinguishes measured binary interactions from their incompletely resolved cellular functions. EIF4H’s cytoplasmic localization is compatible with DLG3 and does not establish a compartment mismatch; neither pair is rejected. These decisions supersede the earlier partner-class refinements for these two HuRI rows. See HuRI, Generation and characterization of HuRI and Discussion.
The Preso/FRMPD4 refinement remains cytoskeletal regulatory protein binding. The original Figure 1 and Results explicitly include SAP102 in motif-controlled interaction experiments, cellular complexes and rat-brain complexes. Preso’s spine and filamentous-actin phenotypes independently establish the recognized partner class. The term does not identify the physical binding surface as a FERM domain or transfer PSD-95-specific localization effects to DLG3.
The tear-proteome identification comes from the separately recovered publisher Table II. That table is absent from the normal publication cache, whose original full-text flag remains unchanged. The extracellular-region reason now makes this access distinction explicit. The contextual detection decision and all measurement-artifact files are unchanged.