[hyphaltip]

This implements the v0.9.0 checklist on #195 and #196, following the curator ruling on the design note (#321) and the grilling of 2026-11-09 (decisions Q1–Q9 below).

What

• datasets/reads/sra_survey/pilot/causes.tsv (curated). One row per species: species, cause, build_cause, basis, with the evidence for each.
  • homothallic: both MAT genes in every genome, at one locus or unlinked loci → NOT_TYPABLE:homothallic.
  • silent_cassettes → NOT_TYPABLE:silent_cassettes.
  • zygosity: joins ZYGOSITY, so both is MTL a/alpha.
  • cause_unchecked → UNVALIDATED:cause_unchecked plus the usual soft reasons.
• wave_driver.py:
  • loads the causes at import;
  • the cause overrides the numbers in gate, regate and curate, and also for species held at the panel stage;
  • new command apply-causes;
  • the ≥50%-both shortcut no longer gives PROVISIONAL_HOMOTHALLIC: a species with no cause gets UNVALIDATED:cause_unchecked (Q3);
  • submit-full does not queue NOT_TYPABLE species.
• cause_evidence.py (new) → cause_evidence.tsv (42 species). For each species it gives:
  • the assembly evidence: genomes with one idiomorph, with both in one locus, or with both at separate loci (from release v0.8.1 loci.tsv);
  • the reads gate numbers;
  • for failed builds, the panel lineages, each with its idiomorph counts.
• Committed gate tables: 36 rows change against main, each listed in wave*/cause_changes.tsv. 18 were a HOLD, 12 PROVISIONAL_HOMOTHALLIC and 6 UNVALIDATED.
• Site:
  • new confidence label "not typable by presence", showing the reason and the evidence line;
  • allele counts hidden on the species page and in the index;
  • the index explains the label;
  • the empty "provisional (homothallic)" filter is removed.
• reads_availability.py: new action not_typable.
• Docs: CONTRACT.md has a new causes section.

Decisions that change (all listed in cause_changes.tsv)

Now	Species
NOT_TYPABLE:homothallic	A. nidulans, A. latus, A. spinulosporus (both genes in every assembly, unlinked); S. sclerotiorum, D. hansenii, C. fimbriata (one locus)
NOT_TYPABLE:silent_cassettes	S. paradoxus, S. eubayanus, S. pastorianus, S. mikatae, S. kudriavzevii, S. uvarum, S. arboricola, K. marxianus, K. lactis, K. phaffii, O. haglerorum, L. thermotolerans, L. fermentati, L. kluyveri, Zygosaccharomyces ×8, S. pombe, S. japonicus
PASS_ZYGOSITY / PROVISIONAL_ZYGOSITY	C. tropicalis (87 of 115 assemblies a+alpha) / L. metapsilosis
UNVALIDATED:cause_unchecked	N. tetraspora, Lasallia pustulata, H. anomala, M. pulcherrima (were "provisional (homothallic)"; their assemblies do not support it)

C. dubliniensis and C. viswanathii get the cause zygosity, but stay HOLD:panel_build_failed, a technical reason: their panel loci are all a. L. orthopsilosis and Mucor hiemalis have a build cause only.

Why the yeast panels failed (#196). In every failed build, the panel's assembly loci cover one idiomorph only, so there is nothing to build from:

• all alpha in Zygosaccharomyces;
• all a in C. dubliniensis, C. viswanathii and L. orthopsilosis;
• D. hansenii cut no loci at all, because its a and alpha genes sit in one locus.

Curator decisions (grilling, 2026-11-09)

• Q1: one cause named homothallic; the locus layout goes in basis.
• Q2: unchecked species stay visible as UNVALIDATED.
• Q3: the ≥50%-both rule no longer claims homothallism.
• Q4: S. uvarum is silent_cassettes; its 263 single calls are noted on Reads: copy-number ratio test for active MAT in cassette yeasts and MTL zygosity #322.
• Q5: C. fimbriata's 7 unsearched assemblies were searched first.
• Q9: C. fimbriata is homothallic.
• Q6–Q8 concern the A. fumigatus sketch-store follow-up, a separate PR.

C. fimbriata. The 7 assemblies v0.8.1 skipped (taxid 5158 has no NCBI lineage) were searched on the taxid route (1035309). Each carries unmodelled hits of both MAT1-1-1 and MAT1-2-1, and none carries one idiomorph only. A --phylum attempt and a control on the called assembly showed that --phylum leaves every gene unmodelled. Both detector problems are filed as #324.

Checks

• New and updated tests: tests/test_wave_driver_gate.py (cause overrides, apply-causes, the causes file uses only ruled causes), tests/test_cause_evidence.py, and a not-typable test in tests/test_build_reads_site_script.py.
• Full suite on the HPCC at 96d4bcb: 1967 passed, 6 skipped. The last commit, 9110a44, removes one filter option from the index page and is covered by no test.
• Site build: 247 species, 30 of them "not typable by presence".
• Headless browser:
  • K. phaffii shows the cassette reason and evidence, with counts hidden;
  • Lasallia is unvalidated, with counts visible;
  • the index shows 30 "not an allele frequency" cells;
  • no console errors.

Closes the v0.9.0 checklists of #195 and #196. The issues stay open for v0.10.0 (#322, #323).

🤖 Generated with Claude Code