# Lys-C allele peptide specificity — executed reference-proteome screen

14 September 2026. **COMPUTATIONAL SEQUENCE SCREEN, NOT AN ASSAY VALIDATION.** This adds new evidence to the earlier [design packet](PROTEOMICS_DESIGN_PACKET.md), whose statement that proteome-wide specificity was untested was correct at its checkpoint.

## Result

In the downloaded UniProt human reference-proteome release **2026_03**, including its additional isoform file, the full WT 26-residue peptide has **14 sequence-record matches, all labeled SCN5A**. Treating I and L as indistinguishable adds no full-peptide match. The R104Q counterpart has no match in this unmodified reference database. This is consistent with a useful sequence-discrimination candidate; it does not establish absence of all possible biological or spectral interference.

**A useful design warning emerged:** the WT eight-residue prefix appears SCN5A-specific under literal sequence matching but also matches SCN8A when I/L are collapsed. The checked WT 12-residue prefix does not have that additional gene match. Thus short diagnostic fragments must not inherit the parent peptide's apparent specificity.

## Scope and provenance

The existing nonrestricted scan inventory was inspected for protein FASTA/proteome inputs; it provided selected reference/candidate sequences but no identified complete protein-proteome file. This is a filename/inventory finding, not proof that no archive contains one. No new broad filesystem or private-record search was performed.

Official source: [UniProt human reference-proteome directory](https://ftp.uniprot.org/pub/databases/uniprot/current_release/knowledgebase/reference_proteomes/Eukaryota/UP000005640/). The official README identifies the base FASTA and additional isoform/variant set. See [UniProt proteome documentation](https://www.uniprot.org/help/proteome) and [canonical/isoform explanation](https://www.uniprot.org/help/canonical_and_isoforms).

The streaming API timed out without returning data. The precomputed official files succeeded. Before downloading, HEAD responses established sizes of 47,819,245 and 5,423,103 bytes; 264 GiB local free space was observed. Downloads were capped at 50 MB and 6 MB respectively. No paid resource or GPU was used.

The saved [official metalink](UP000005640_RELEASE.metalink) identifies release 2026_03 and exact sizes/MD5s; both checks passed before sequence analysis. This authenticates the downloaded snapshot against the retrieved publisher manifest, not against an independently signed manifest. Dynamic URLs can change; retained files and hashes define this experiment.

| Input | Records scanned | Residues scanned | SHA-256 |
|---|---:|---:|---|
| UP000005640_9606_2026_03.fasta.gz | 147,503 | 69,576,816 | 3c2e371b769543fac552a5ff0bd9f8a4d8eb136aab53b08e623d1ba417632fbd |
| UP000005640_9606_additional_2026_03.fasta.gz | 22,131 | 13,097,415 | 55a80407a7c76be203c87f72b374daaec045b8ce7b75a74954bbc8d32db0232f |

Total: **169,634 sequence records and 82,674,231 residues**, not that many genes. Reference canonical/reviewed/unreviewed and isoform records must not be mistaken for independent biological observations.

## Query results

Queries are all contiguous substrings, unrestricted by protease boundaries. Prefix/suffix names indicate sequence segments, **not observed fragment ions or recommended transitions**. R104/Q104 is position 4 in the full candidate, so every segment below contains the differing residue. Counts are occurrences across records, not distinct genes or predicted mass-spectrometry peaks.

| Segment | WT exact / I-L-equivalent matches | Mutant exact / I-L-equivalent matches | Interpretation |
|---|---:|---:|---|
| Prefix 4: TIFR / TIFQ | 270 / 1,145 | 350 / 1,164 | Many background sequences; inadequate alone for sequence discrimination |
| Prefix 8 | 14 / 23 | 0 / 0 | WT exact matches SCN5A; I/L adds SCN8A |
| Prefix 12 | 14 / 14 | 0 / 0 | WT matches carry only SCN5A gene labels |
| Prefix 16 | 14 / 14 | 0 / 0 | Same gene-label scope |
| Suffix 23, beginning at variant residue | 14 / 14 | 0 / 0 | Same gene-label scope |
| Full 26 | 14 / 14 | 0 / 0 | Same gene-label scope |

The full WT matches include Q14524 and five named isoforms plus other SCN5A-labeled records. Hence the peptide does **not** distinguish full-length versus fragment/isoform sources containing this sequence. The zero mutant reference count is expected for an introduced variant and is not evidence that mutant protein cannot exist or be detected.

## Verification and reproducibility

[proteomics_specificity_check.py](proteomics_specificity_check.py) parses both compressed inputs through EOF, checks publisher size/MD5, records SHA-256, validates sequence characters and searches exact and I/L-collapsed substrings. For hits, regex positions are cross-checked against direct position-wise matching; absence is checked by substring membership. [PROTEOMICS_SPECIFICITY_RESULTS.json](PROTEOMICS_SPECIFICITY_RESULTS.json) retains all matched headers, positions, input files, counts and gene labels. Measured local runtime: **20.86 seconds**. Output creation is exclusive to preserve this version.

## What remains unmeasured or outside the database

- No spectral-fragment intensity, retention time, peptide abundance, digestion yield, calibration, recovery, detection or quantitation limit was predicted.
- This is a versioned reference set, not every human allele, unannotated translation, nonhuman contaminant, cell-line mutation or sample-specific protein.
- I/L collapse addresses one ambiguity, not every isobaric composition, modification, charge/adduct, coisolation or mass-tolerance interference.
- A substring match does not guarantee that a background protein generates that exact peptide after digestion; conversely, no exact match does not exclude an interfering spectrum.
- SCN5A-only gene labels are database annotations, not experimental attribution. Actual sample sequence and matrix still matter.

**Next decision:** retain the Lys-C pair as a sequence-qualified candidate within this reference scope, and carry the SCN8A short-fragment caution into empirical transition selection. Proceed with the core-facing feasibility/validation checklist in the design packet; do not launch repeated background scans absent a new database or a specific interfering candidate.