Skip to content

Varcode, Isovar, and Vaxrank

Vaxrank evaluates protein/peptide candidates. It should preserve the evidence behind a sequence, not reinterpret an RNA junction or select a transcript merely because its effect class ranks first.

Library Responsibility
Varcode Generate structural/transcript hypotheses and predict their coding consequences.
Isovar Reconstruct RNA-supported sequences, compare them with those hypotheses, and preserve unresolved alternatives.
Vaxrank Evaluate the resulting protein/peptide candidates, retaining their evidence.

This is the shared responsibility split; it does not imply that all candidate paths are already connected to the default pipeline.

What Vaxrank receives

A sequence candidate needs its nucleotide/transcript source, compatible transcript IDs, reference/annotation identity, completeness, frame assumptions, and sample/library/read evidence where available. Keep RNA-supported and DNA-only predicted sequences distinguishable.

Vaxrank uses these candidates for peptide generation, prediction, filtering, ranking, and vaccine design. Binding or manufacturability scores do not resolve which RNA isoform is expressed. RNA support does not itself prove translation, antigen presentation, tumor specificity, or immunogenicity.

If candidates encode the same protein, grouping them must retain all source structures and provenance. Different proteins remain distinct candidates. Missing RNA is not proof of no expression; insufficient evidence should remain visible rather than becoming an unsupported confidence claim.

Available today and remaining work

  • Ordinary RNA path: from_isovar_result uses IsovarResult.top_protein_sequence. Isovar can retain more alternatives in sorted_protein_sequences; the ordinary Vaxrank path does not evaluate all of them automatically.
  • Supplied-fusion path: fusion_antigens_from_isovar retains coding hypotheses and reports unresolved/ambiguous status. It is separate from the small-variant pipeline; see fusion inputs.
  • Opt-in DNA fallback: consequences come from Varcode, not RNA reconstruction. It currently selects an effect before checking all fusion protein candidates (#482). The fix must evaluate usable alternatives before selection. A set-level protein-change flag does not guarantee its first protein is changed.
  • RNA discovery/reconciliation: Isovar's supplied-fusion translator is not an automatic soft-clip assembly workflow. That upstream connection is Isovar #305.

The intended contract is to retain supported alternatives and their uncertainty through evaluation. A consumer that requires one result must make the selection policy explicit; effect severity, RNA support, and peptide ranking answer different questions.

Other library guides