[hyphaltip]

Follow-up to #327. This queues up the 17 thin-evidence species that have more runs to type, so the top-up can start as soon as the HPCC shutdown ends. Nothing was run on the HPCC: the code and runbook were written and tested on the laptop.

What

• expand_runs.py --species-file (optional) limits a top-up to the species it lists. Comments are skipped, and a listed species with no untyped run is reported on stderr. Without the option, nothing changes.
• append_stage.py (new) appends a stage's runs to a committed wave.
  • Existing rows are left untouched, including their rescue and screen columns. merge_screen.merge rewrites every row's screen columns from the stage, which wiped them when I first used it for Reads gate: type thin-evidence species (one strain or 1-3 runs) as UNVALIDATED instead of holding them #327 (caught before commit).
  • The species that gained runs are decided again through merge_screen.regate_species, so their curator rulings and thin-evidence tokens are kept. Changed rows go to append_changes.tsv.
• orchestration/thin_topup_2026-11-10/:
  • species.txt: the 17 species, with eligible and typed counts from the 2026-11-06 snapshot.
  • run.sh refresh | lists | submit | check.
  • README.md.
• CONTRACT.md: when to use append_stage.py and when to use merge_screen.py.

Expected size

About 340 new runs, going by the 2026-11-06 snapshot; the refresh decides the real numbers.

species	eligible isolates	typed now
F. sambucinum	159 (one study)	8
P. fijiensis	78	9
Madurella	58	20
C. amylolentus	45	11
F. andiyazi	36	9
Hymenoscyphus	59	41

The few_runs species, Pneumocystis among them, have only 4 to 9 eligible isolates.

Choices to confirm

• Cap 700: the same as round 3.
• No per-study limit (--per-study 700), as in expand300b. This matters for F. sambucinum, whose isolates all come from one study; a 25-per-study limit would keep it at about 33 runs.

Checks

• Full suite, run on the laptop because the HPCC is down: 1965 passed, 20 skipped. The HPCC skips 8; the extra 12 skips here are tests whose external tools the laptop lacks.
• New tests:
  • tests/test_append_stage.py: existing rows untouched; a topped-up thin species stays capped; a species with no new runs is not decided again.
  • tests/test_expand_runs_script.py: the species filter.
• run.sh passes bash -n. It has not been run.

After the shutdown

cd <frozen worktree of main>
bash datasets/reads/sra_survey/pilot/orchestration/thin_topup_2026-11-10/run.sh refresh
bash datasets/reads/sra_survey/pilot/orchestration/thin_topup_2026-11-10/run.sh lists
bash datasets/reads/sra_survey/pilot/orchestration/thin_topup_2026-11-10/run.sh submit

Run lists and read expand_summary.tsv before submit.

🤖 Generated with Claude Code