PRJNA694496: selected-versus-control reanalysis

Generated 2026-09-25T18:09:49.510479+00:00.

Exploratory study-level results. QC and leave-one-out sensitivity have been reviewed; biological limitations still constrain interpretation and cross-study pooling.

Design and methods

Three selected and three control whole-larva libraries; positive effects indicate higher constitutive expression in the selected population. Reference: GCF_963853765.1, RS_2024_06. Salmon 1.10.3 with decoys, sequence and GC bias correction; tximport gene-level summarization; DESeq2 ~ condition. Genes require at least 10 counts in at least three libraries. BH-adjusted p < 0.05 defines significance. Unshrunk effects and standard errors are retained; apeglm effects are for ranking only.

Results

  • Genes retained after count filtering: 19,816.
  • Genes with nonmissing raw p-values: 19,814.
  • Genes with nonmissing adjusted p-values: 19,814.
  • Significant genes: 8,802; 4,353 higher and 4,449 lower in selected larvae.
  • Mapping rates: 42.13%–63.94%.
  • PC1 explains 90.6% of variance and separates all selected libraries from all controls.
  • Control correlations are 0.996–0.996; selected B/C correlation is 0.991.
  • Selected A is displaced on PC2 and has 6 genes above the Cook’s-distance threshold; it is retained in the primary model.

Sample PCA

MA plot

Selected A sensitivity analysis

A leave-one-out model excludes selected A while retaining the same filtered gene set, leaving two selected and three control libraries. This is a sensitivity analysis; the six-library model remains primary.

  • Effect correlation with the primary model: 0.962.
  • Primary significant genes retaining the same direction: 8,802 of 8,802.
  • Primary significant genes remaining significant at FDR < 0.05: 8,639.
  • Significant genes in the sensitivity model: 11,994.

Functional annotation and evidence release

All 19,816 tested genes have exact GeneID mappings and versioned descriptions/biotypes from GCF_963853765.1 / RS_2024_06. RefSeq-characterized descriptions are available for 2,703 selected-higher genes and 3,271 selected-lower genes.

Local gene-biotype over-representation found selected-higher enrichment for lncRNA (1.28-fold; BH-adjusted p=2.89e-07) and rRNA (4.55-fold; adjusted p=7.88e-10), and selected-lower enrichment for protein-coding genes (1.03-fold; adjusted p=2.65e-20). These broad classes do not substitute for formal GO/pathway analysis.

The production harmonized evidence release contains every tested effect and standard error, with exact current-reference mapping.

Formal version-matched g:Profiler analysis mapped all direction-specific RefSeq symbols without failure or ambiguity. Selected-higher genes produced 35 significant GO terms led by RNA binding, RNA processing, and ribonucleoprotein-complex annotations. Selected-lower genes produced 108 terms led by catalytic activity, small-molecule metabolism, cytoskeleton, and phosphorus/phosphate metabolism. The oyster build supplied GO only; KEGG, Reactome, and WikiPathways results were unavailable.

Descriptive comparison with the earlier heat-response study

Of 60 mapped, published significant heat-response genes in PRJNA516762, 52 occur among the genes retained here; 33 are significant here. The overlap table retains both effect directions and the earlier study’s inferred mapping confidence.

No pooled effect or enrichment test is calculated. This historical overlap uses the publication-derived, significant-only PRJNA516762 list, which lacks standard errors. A later raw-data reanalysis of PRJNA516762 is now available in the harmonized release; it is not the input to this historical overlap. Acute heat response in juvenile gill differs from constitutive expression in selected larvae. Direction agreement is descriptive and does not establish a common protective mechanism.

Interpretation limits

Selection is confounded with population background. Survival was assessed in related later-stage cohorts, not the sequenced larval pools. These results are associations, not validated predictors of individual survival. Mapping-rate variation, sample correlation/PCA, and Cook’s distances require joint review; no samples were removed automatically. Missing p-values and adjusted p-values remain missing.

Files

  • gene_results_annotated.tsv: all retained genes, effects, uncertainty, transcript provenance, and ranking-only shrunken effects.
  • top_25_genes.tsv: significant genes ordered by absolute shrunken effect.
  • cross_study_descriptive_overlap.tsv: mapped overlap, without statistical pooling.
  • exclude_selected_A_summary.tsv and exclude_selected_A_comparison.tsv: leave-one-out robustness results.
  • refseq_annotation_report.md, refseq_biotype_summary.tsv, and refseq_biotype_enrichment.tsv: local versioned functional annotations.
  • enrichment/functional_enrichment.tsv, functional_enrichment_report.md, and enrichment_qc.json: direction-specific g:Profiler GO results and conversion QC.
  • QC tables, figures, software session information, and input SHA-256 checksums are alongside this report.