Oncovirus antigen database for on-target ligand classification (#249 follow-up)

[iskandr]

Context

Follow-up to #249's ligand-source-tagging taxonomy. The viral slot of on_target ∈ {mutant, viral, CTA} covers ligands derived from oncoviral antigens — the legitimate on-target case for therapeutic cancer vaccines against virus-driven malignancies.

Goal

Curate (or fetch via tsarina / pirlygenes) a minimal oncovirus antigen set so vaxrank can:

1. Tag ligands matching viral antigen sequences as on_target_viral.
2. Distinguish them from incidental host-genome viral-fragment integrations or processed pseudogene artifacts.

Initial set to consider

The four oncoviruses with established therapeutic-vaccine relevance:

• HPV — E6, E7 (cervical, head & neck, anal cancer)
• EBV — LMP1, LMP2, EBNA1 (nasopharyngeal carcinoma, post-transplant lymphoma)
• HBV — preS, S, X, core (HBV-driven hepatocellular carcinoma)
• MCPyV — large T, small T (Merkel cell carcinoma)

KSHV (HHV-8) and HTLV-1 are secondary; HCV is treated upstream of vaccines today (DAAs).

Sourcing decision

• Bundle a minimal default — small (a few KB), version-pinned, no network dependency. Concrete, low-friction.
• User-supplied FASTA — for trial-specific virus strains. Add --viral-antigens-fasta for override.
• tsarina / pirlygenes route — preferred if these tools already curate the set; otherwise bundle.

Likely answer: bundle the four canonical oncovirus reference sequences from UniProt/RefSeq, allow override via FASTA.

Acceptance

• Bundled FASTA (or documented upstream-tool fetch) covering the four oncoviruses with primary sources cited.
• ligand_source(peptide) returns 'viral' (with virus + protein name) when the source matches.
• Override flag for trial-specific viral genomes.
• Tests against HPV E7 (well-characterized: HLA-A*02:01 epitope YMLDLQPETT etc.).

Related

• Boundary-processing-aware SLP windowing #249 — ligand-source taxonomy and pepsickle credibility tagging