[study] GEORGE2023_HEATWAVE_PLOIDY: rerun on main's single-end code

[sr320]

Follow-up to #18, which merged on 2026-10-15 with outputs from the branch's pre-merge single-end code. In that code fastp auto-detected the single-end adapter. main took a different single-end implementation (#25) that trims an explicit adapter (--se_adapter, default TruSeq read 1), so #18's committed results could not be reproduced from main.

What this does

• Reruns all four comparisons on current main: Nextflow 26.04.7, native, on Hete. FastQC and the Salmon index came from cache; fastp, Salmon quant and DESeq2 were rerun.
• Replaces the four standardized tables, workflow manifests and HTML reports.
• Removes process_versions.yml. Since Build manifest software_versions from what actually ran #19 the manifests record the tools that actually ran, and they now also record se_adapter.
• Updates the YAML provenance with the rerun and its effect.
• Adds a test that every manifest records main's adapter, so the gap can't silently return.

How much changed: almost nothing

Comparison	DE now	DE in #18	Shared DE genes	Fold-change r
diploid 30 °C (primary)	3,076	3,081	3,074	0.9994
triploid 30 °C	2,258	2,254	2,252	0.9994
diploid 30 °C + 44 °C	1,696	1,688	1,688	0.9991
triploid 30 °C + 44 °C	1,621	1,614	1,614	0.9981

• Per library, a handful of reads differ. In SRR22817776, for example, 2 of 1.67 M mapped reads and 91 transcripts change. For 3′ tags the poly-A trimming, which was already on, dominates.
• Mapping is unchanged: 20.8–50.1%, median 38.4%.
• Replicate dispersion is unchanged: 0.028–0.052.
• Agreement with the authors' DE lists is unchanged: 88.4 / 96.4 / 96.9 / 97.3% same direction.
• Harmonized: 100,245 evidence records (previously 100,241).

Checklist

• aree validate-study passes
• Curation tests: 13 (full suite 593)
• check-study-scope passes; only study-owned files change
• No raw reads, work/ or results/ committed

🤖 Generated with Claude Code