Skip to content

Trust and verification signals on the Sequence Detail page

Motivation

The NCBI parity work made the Sequence Detail page a good browse surface, but a molecular-diagnostics researcher still had to pivot back to NCBI to verify three things the page could not express:

  1. Version currency — is this accession the current .version, or has it been replaced/suppressed/withdrawn?
  2. Truncation completeness — long fields (translation, product, comment, definition, taxonomy lineage) are clipped server-side, but the page gave no signal that the visible value was partial.
  3. Related-link precision — symbol/organism deep links looked as authoritative as stable-id links, so a researcher could not tell an exact match from a text search that may return several records.

This change turns the page from browse-only into a verification surface without adding any new /api call (the related links remain external NCBI URLs).

User-facing change

  • Record-status trust badges in the header:
  • replaced / suppressed / withdrawn / dead / unverified records show a warning pill; a replaced record links straight to the superseding accession when NCBI reports replacedby.
  • A live record shows a calm "Live record" pill.
  • A MANE Select keyword (parsed from the GenBank KEYWORDS block) shows a trust pill so the canonical transcript is visible at a glance.
  • Truncation "view full on NCBI" markers wherever the backend clipped a value: per-feature qualifiers (translation/product), the Comment card, and the GenBank record card (definition / taxonomy lineage). Each marker is an external link to the full NCBI record.
  • Related-resource confidence tags — links built from a stable id (GeneID, taxid) are treated as exact; links built from a gene symbol or organism string are tagged search with a tooltip noting the query may return several records.

API / IaC diff summary

Additive only; the nuccore route returns the service dict directly, so no route or Pydantic model changed.

  • api/services/ncbi/nuccore.py
  • esummary parse now emits status and replaced_by.
  • New _truncate_flagged(...) -> (clipped, truncated) helper.
  • GenBank parse now emits a truncated_fields: list[str] record-level field and a per-qualifier truncated: bool.
  • web/src/api/ncbi.ts — NuccoreSummary.status, NuccoreSummary.replaced_by, optional NuccoreQualifier.truncated, optional NuccoreGenBank.truncated_fields.
  • web/src/pages/sequence/SequenceDetail.tsx — trust badges, truncation markers, related-link confidence tags.

No IaC change.

Validation evidence

  • uv run pytest -q api/tests/test_ncbi_nuccore.py → 54 passed (adds test_fetch_nuccore_genbank_flags_truncated_fields and test_fetch_nuccore_summary_flags_replaced_record, plus new asserts on the untruncated fixture).
  • cd web && npm run build → clean type-check + build.
  • uv run ruff check api/services/ncbi/nuccore.py api/tests/test_ncbi_nuccore.py → all checks passed.