The published workbook contains 1,344 DLG3 domain–motif measurements, 448 each for DLG3_1, DLG3_2 and DLG3_3. The 396 imported PMID:36115835 annotation rows yield 1,221 accession-candidate joins to those measurements. Of these, 1,200 match the supplied current product sequence through the workbook's natural_10mer field; 21 do not. Three AXIN1 internal-peptide rows additionally match through the actual peptide sequence, giving 1,203 sequence-supported scientific selections. A selection is an identification/mapping decision, not a positive-binding verdict.
The source is Gogl et al., 2022, PMID:36115835, Supplementary Data 1, sheet PDZ-PBM interactome. The public supplementary archive contains 41467_2022_33018_MOESM4_ESM.xlsx; that workbook's SHA-256 is ada912921164f0a79630c6a714b9ff264a7217ad3e646153ef498eb456206795.
reproduce.py downloads that archive or reads a local workbook, verifies the workbook version, and extracts the original rows without imputing affinities. It reads imported partner accessions from the adjacent review and checks their explicit sequence inputs against identity provenance. The provenance records official UniProt URLs, record and sequence digests, product lengths, taxa and alternative-sequence features. The frozen sequences are data inputs, not predicted sequences or inferred activity. README.md gives runnable commands.
| Output | Contents |
|---|---|
| measurements.tsv | All 1,344 DLG3 rows, original one-based sheet row numbers, and all 39 source columns. The single unnamed source column has an explicit generated header. |
| source-row-crosswalk.tsv | All 1,221 candidate joins, imported zero-based annotation index and literal accession, original peptide, separate sequence-mapping flags, and selection flag. |
| run-summary.json | Counts and source/input digests computed by the parser. |
All 53,760 extracted cells were compared with the independently saved original-workbook projection. All 1,221 candidate mappings and their natural-sequence positions match the prior source crosswalk; the 1,203 selected row identifiers match the individually reviewed source projection. The same parser was also run against DLG4 and its own identity inputs: it extracted 1,344 measurements, 388 source rows and 1,197 candidate joins. That second-input check establishes parser generality, not a DLG4 scientific verdict.
The study tests recombinant PDZ domains against synthetic motifs. LYSC uses bacterial lysates; SAPF and DAPF use single- and double-affinity-purified domain preparations. The stated binding-intensity thresholds are 0.1 for LYSC/SAPF and 0.05 for DAPF. The workbook's COMP field selects the preferred assay estimate; it is not an average of discordant assays. Each original assay field is retained separately.
A zero-coded composite indicates that the selected experiment did not supply a quantifiable affinity under its detection conditions. It is not a measured zero affinity, an affinity of 1 M, or proof of biological noninteraction. The authors explicitly discuss possible binding below the detection threshold. Conversely, a non-negative interaction-database flag is not an independent positive affinity measurement.
Positive motif binding supports molecular recognition at the actual fragment, modification and construct scope. It does not by itself establish a native full-length complex, colocalization, receptor activation, partner regulation, or activity of every DLG3 splice product. The review retains supported generic associations as non-core unless independent target-specific context supports a more informative partner-class function. GluN2 receptor recognition, FRMPD4/Preso binding and ERBB4-tail binding have such additional context, described in the evidence guide.
| Imported row / partner | Source measurement and scope |
|---|---|
| 28 / GAGE12J | SD1 rows 9021, 55785 and 64021 test PEEGKKQSQC; row 9021 has composite pKd 4.03522737. The supplied current A6NER3 product has PEEGEKQSQC. These remain unmatched candidate rows; they are not counted among the 1,203 sequence-supported selections. A positive mismatched peptide does not settle the literal current-product assertion. |
| 47 / AXIN1 | The natural_10mer field is an internal-motif placeholder. Rows 8454, 34568 and 47894 specify the actual amidated internal ligand Biotin-GSGSGS-TLSLTDSSVDGIPPYR-CONH2; its amino-acid core matches AXIN1. These are the three additional selected rows. Row 8454 is positive; selection of the other two does not make their zero composites positive. |
| 93 / SARS-CoV-2 E | Row 21885 has LYSC/CALIP pKd 4.09810075 but selected DAPF/COMP zero. The discrepancy is retained rather than relabeled as universally undetected or proven native binding. |
| 166–167 / PDLIM4 products | The paper's isoform names differ from current UniProt names. Candidate accession joins are broader than exact sequence matches; literal product sequences determine the matched tails. Both mapped native tails are below quantification. |
| 226 / RPS6KA1 | Composite-positive measurements are confined to the minus-one-phosphorylated motif. The minus-three-phosphorylated motif also has a discordant LYSC/CALIP estimate at row 21824, while its selected composite is zero. No unmodified-tail affinity is inferred. |
| 383 / WHRN | The internal synthetic ligand ends in Y where the corresponding current sequence ends in F. The native C-terminal motif is below quantification. The internal positive is therefore retained among unmatched candidate observations, not exact native-product support. |
| 253 / ARHGAP17 and 408 / DLGAP4 | Tail measurements are below quantification, but independently curated AP-MS associations support non-core retention. Those captures do not establish the isolated-tail mechanism. |
The 120 unresolved source assertions comprise 117 below-quantification cases and the GAGE12J, SARS-CoV-2 E and WHRN identity/assay discrepancies. The evidence was inspected; the uncertainty is biological or measurement-specific rather than a claim that the paper was inaccessible. Imported supporting entities remain source metadata and are not rewritten to peptide identifiers.