Skip to content

2026-05-24 — New Search: DB-first stepper + per-DB query examples

Motivation

On the New Search workspace the stepper ran Program → Query → Database. The default example queries (MPXV, P. falciparum 18S, SARS-CoV-2) only produce hits against core_nt / nt, so users picking 16S_ribosomal_RNA, ITS_RefSeq_Fungi, pdbnt, swissprot, nr, or refseq_protein would load an example, run a search, and see zero hits. Worse, the recommendedDb text on the P. falciparum entries mentioned 18S_fungal_sequences (not in our catalogue), which was actively misleading.

User-facing change

  1. Stepper reorder — Step 2 is now Search set, step 3 is Query sequence. The Query section is greyed out with a notice until a database is picked, so the example picker can always scope to a real DB. Switching to a category with downloaded databases now auto-selects the first database in that category (for example, Standard databases → core_nt, rRNA/ITS databases → 16S_ribosomal_RNA) so Query unlocks immediately.
  2. Per-DB example filtering — opening "Load example" now shows only the templates that produce hits against the currently selected database. A subtitle reports the count, e.g. "Showing examples that hit 16S_ribosomal_RNA (5/30)." Custom or uncovered databases show an empty curated-example state instead of falling back to unrelated examples.
  3. Expanded authoritative examples (NCBI / UniProt / RCSB): every built-in DB now has at least five matching examples.
  4. E. coli 16S rRNA — NR_024570.1 (1,450 nt) → 16S_ribosomal_RNA, core_nt, nt
  5. Additional 16S rRNA examples: B. subtilis NR_102783.2, S. aureus NR_037007.2, P. aeruginosa NR_117678.1, M. tuberculosis NR_044826.2
  6. S. cerevisiae ITS — NR_111007.1 (752 nt) → ITS_RefSeq_Fungi, core_nt, nt
  7. Additional fungal ITS examples: C. albicans XR_002086439.1, A. fumigatus PZ411633.1, C. neoformans PZ094085.1, P. chrysogenum PZ394474.1
  8. Yeast tRNA-Phe — PDB 1EHZ chain A (76 nt, U→T) → pdbnt, core_nt, nt
  9. Additional PDB nucleotide examples: P4-P6 ribozyme PDB 1GID, TPP riboswitch PDB 2GDI, A. thaliana TPP riboswitch PDB 3D2V, adenine riboswitch PDB 5SWE
  10. Human insulin — UniProt P01308 (110 aa, blastp) → swissprot, nr, refseq_protein
  11. Human p53 — UniProt P04637 (393 aa, blastp) → swissprot, nr, refseq_protein
  12. SARS-CoV-2 Spike — UniProt P0DTC2 (1,273 aa, blastp) → swissprot, nr, refseq_protein
  13. Additional protein examples: human hemoglobin beta P68871 and human cytochrome c P99999
  14. Program auto-switch — loading a protein example sets program = blastp; loading a nucleotide example keeps blastn. Previously every example forced blastn.
  15. Fixed misleading recommendedDb on the five P. falciparum 18S entries (18S_fungal_sequences or core_nt → core_nt or nt).
  16. Stale subrange hardening — loading a new example clears any previous query_from / query_to values so an old subrange cannot silently clip the newly loaded sequence.

API / IaC diff summary

Frontend only. No backend, no infra, no Bicep, no Container App template changes. No new dependencies.

Touched files: - web/src/pages/blastSubmit/queryExamples.ts — added matchingDbs: string[] and blastProgram: BlastProgram fields on the template type; populated for all 30 templates (10 existing migrated, 20 new added across NCBI / UniProt / RCSB sources). - web/src/pages/blastSubmit/queryExamples.test.ts — new assertions that every template tags a known program + at least one known DB, that every key in DB_DESCRIPTIONS is covered by at least one template, and that nucleotide templates only point at nucleotide DBs (and likewise for protein). Also pins the selected-DB-only filter counts for 16S_ribosomal_RNA, ITS_RefSeq_Fungi, pdbnt, swissprot, unknown custom DBs, and empty DB selection, and enforces at least five examples for every built-in DB. - web/src/pages/blastSubmit/DatabaseSection.test.ts — pins category auto-selection paths for Standard, rRNA/ITS, Genomic, and Custom database buckets, including empty-category behaviour. - web/src/pages/blastSubmit/SubmitStepper.tsx — swapped STEPS[1] and STEPS[2], updated stepDone() case ⅔ mapping. - web/src/pages/blastSubmit/DatabaseSection.tsx — SectionHeader step={3} → step={2}; category tab changes now auto-select the first downloaded database in the selected category unless the current selection already belongs to that category. - web/src/pages/blastSubmit/QuerySection.tsx — SectionHeader step={2} → step={3}, gated textarea + upload + Load example + Reverse complement + Deduplicate + Clear behind dbSelected, disabled query subrange inputs until a DB is picked, added a one-line notice when no DB is picked, loadExample now respects example.blastProgram and resets stale subranges, and QueryExampleDialog filters by selectedDbName with a count subtitle. - web/src/pages/BlastSubmit.tsx — swapped the JSX render order so DatabaseSection sits at sectionRefs[2] and QuerySection at sectionRefs[3].

Validation evidence

  • npx vitest run src/pages/blastSubmit/ — 10 files, 134 tests passed (7 assertions in queryExamples.test.ts, including selected-DB-only filter coverage, minimum-five-per-DB coverage, plus database category auto-selection coverage).
  • npm run build — clean, no TS / ESLint errors (only the pre-existing

    500 kB chunk warning on blast-submit-*.js).

  • Tier 2a host-mode loop only; no Container App redeploy required per charter §13 (frontend-only change, no sidecar layout / terminal/Dockerfile* / Bicep touched).

Sources

  • NCBI Entrez E-utilities (efetch.fcgi): NR_024570.1, NR_102783.2, NR_037007.2, NR_117678.1, NR_044826.2, NR_111007.1, XR_002086439.1, PZ411633.1, PZ094085.1, PZ394474.1.
  • UniProtKB REST (/uniprotkb/{accession}.fasta): P01308, P04637, P0DTC2, P68871, P99999.
  • RCSB PDB (/fasta/entry/{id}): 1EHZ, 1GID, 2GDI, 3D2V, 5SWE (RNA U bases converted to DNA T letters where needed).