The human DLG4 review contains 388 source assertions from PMID:36115835(https://pubmed.ncbi.nlm.nih.gov/36115835/). This analysis reads the published SD1 workbook, 41467_2022_33018_MOESM4_ESM.xlsx, from the Europe PMC supplement archive. The workbook SHA-256 is ada912921164f0a79630c6a714b9ff264a7217ad3e646153ef498eb456206795.
There are 1,344 original DLG4 domain measurements: 448 each for PDZ1, PDZ2 and PDZ3. Matching the review's partner accessions creates 1,197 candidate crosswalk rows. Of these, 1,179 match a natural sequence exactly; 18 do not match the natural-sequence field. Three of the latter are AXIN1 internal ligands whose explicitly capped peptide core matches the saved protein sequence. Thus 1,182 measurements are selected by sequence support. Candidate mapping is not equivalent to exact matching, positive binding or a native interaction.
The selected measurements give nonzero composite values for 248 source pairs. All selected composite values are zero for the other 140 pairs. The workbook's zero is not zero affinity and does not establish noninteraction. The source IPI assertions remain intact; unresolved measurement/identity evidence is distinguished from negative biological evidence. Every row's original supporting entity, including isoform identifiers, is retained.
The table preserves COMP_neglogKd, its source method, independent FP/DAPF/CALIP/SAPF fields, peptide sequences, modifications and worksheet row numbers. Values are not averaged into an invented affinity, and no missing or zero value is replaced with a fabricated number. Composite values quoted in review reasons are source values rather than newly measured constants. The phage/peptide and isolated-domain assays do not establish native full-length binding, cellular colocalization, physiological relevance, or equivalent behavior of all DLG4 splice products.
-INTERNAL- placeholder is not an amino-acid sequence; selection instead uses the explicit capped peptide core.The molecular function of the full protein is not determined by the physical form of an assay reagent. The positive peptide/domain interfaces are retained as evidence without a blanket conversion to peptide-binding function. More informative partner-class refinements are made where independent biology supports them: glutamate receptors, potassium transporters, Preso, BCR, ACTN2, ERBB4 and DLGAP4. Other positive fragment interfaces remain non-core binding observations; zero-only cases without independent positive evidence remain uncertain.
The integrated DLG4 core is synaptic receptor-adaptor activity. Receptor recognition, phosphorylation-sensitive SH3-GK recognition and structural organization are cooperating parts of that scaffold, not automatically separate core functions. See the gene source guide for the independent originals and construct limits.
reproduce.py parses the original workbook and the exact review partner identities using identity-sequences.fasta and identity-provenance.tsv. It regenerates measurements.tsv, source-row-crosswalk.tsv and run-summary.json. --check compares generated bytes with those files without changing them. It fails if the workbook or sequence digests differ. The sequence provenance records public UniProt URLs; the frozen FASTA permits replay without assuming that a future official release has identical sequences.