Exclude terminal lysine solubility tags from vaccine scoring

[iskandr]

Terminal solubility additions were counted as vaccine mismatches, so CD109's reconstructed 20-aa core followed by FMV failed against the recorded peptide ending in KKK. Apply the requested benchmark convention to every vaccine peptide: exclude complete terminal runs of K, KK, KKK or KKKK at either end from containment, mismatch and missing-coverage scoring. This includes all vaccine labels and minimal epitopes; internal lysines and terminal runs longer than four remain required.

Keep full recorded sequences visible and in FASTA, with excluded tags purple. Export scored regions and excluded positions in JSON, retain exact RNA allele/target overlap requirements, and correctly align N-terminal tags with core-start protein coordinates. Reject stale reports with an older scoring policy. Empty cores never match.

Re-scored all 29 archived methods from run 37664740743, with identical artifact hashes, proteins and RNA candidates. Recovery changes from 23/36 targets and 52/78 peptide entries to 25/36 and 57/78. Added targets: CD109 and PTH1R. Added peptide entries also include DYNC1H1, CDC40 and EXOC4. The 2+ RNA subset becomes 24/26; top-window disagreements drop from 10 targets to 6. The full-panel amino-acid result remains 31/33. This is an explicit evaluation convention, not a vendor-confirmed synthesis annotation.

Validation: 173 Python tests pass (4 opt-in integration tests skipped locally), 48 JavaScript tests pass, lint and site build pass. Browser checks cover CD109's C-terminal tags, EXOC4's N-terminal tags, containment shading, read counts and top-window flags. Archived-output audit confirms every prior match is preserved after coordinate adjustment and each new match contains the full scored core.