Antecedent oxygen history predicts gas-bubble-trauma risk during total dissolved gas supersaturation after reoxygenation
Description
This dataset contains the experimental data and frozen analysis resources supporting the manuscript “Antecedent oxygen history predicts gas-bubble-trauma risk during total dissolved gas supersaturation after reoxygenation.” Juvenile grass carp experienced 42 d at dissolved-oxygen targets of 3.5, 5.0, or 8.0 mg L−1, followed by 24 or 336 h of reoxygenation and a 48-h total dissolved gas (TDG) challenge at approximately 0, 8, 14, or 20 kPa effective residual pressure. The design comprised 24 history parent tanks, 96 TDG challenge daughter tanks, 960 registered fish, and 576 fish in the primary external gas-bubble-trauma endpoint analysis. The deposit includes: (i) the unchanged authoritative raw-data archive; (ii) curated analysis datasets, QA and metadata corrections, frozen model specifications, numerical model outputs, and analysis code; (iii) source data for manuscript Figures 1–5 and Supplementary Figure S1; (iv) Supplementary Tables S1–S8 and a source map for the two main-text tables; and (v) provenance and reproducibility documentation. First external gas-bubble trauma was analysed using actual-time interval censoring in a discrete-time complementary log-log model. Primary M1 uncertainty used all 2^8 = 256 Rademacher wild-cluster sign patterns across eight challenge batches. RP50,48 is a model-derived risk benchmark and not a universal safety threshold. Because prior dissolved oxygen was collinear with parent rack column and recovery duration was aligned with separate challenge-batch sets, antecedent-state contrasts are interpreted as predictive associations rather than independently randomized causal treatment effects. The original raw-data archive is preserved unchanged. Its legacy analysis directory is retained for provenance only; authoritative final analysis inputs and manuscript numerical results are provided separately in the frozen analysis package.
Files
Steps to reproduce
1. Download all repository files and verify file integrity against SHA256_MANIFEST.csv. 2. Confirm that 01_RAW_DATA_LOCKED.zip has SHA-256: 82bffd1d8f9d2885983e0718220c226305ec9ddd165e6d35a8956339c13ea4f6. 3. Extract 02_ANALYSIS_MODEL_CODE.zip. ANALYSIS_DATA_V1.0 contains the authoritative final analysis inputs, and RESULTS_FREEZE_V1.0 is the authoritative source for manuscript coefficients, model comparisons, RP50 values, and cross-validation metrics. 4. Read CODE_FREEZE_V1.0/README_CODE.md before running the analysis scripts. 5. STAGE4_5_RECONSTRUCTION_RECORD.py documents reconstruction of the frozen analytical dataset and model structure. Earlier ordinary block-resampling outputs were superseded before manuscript use and are retained only for provenance. 6. STAGE5_WILD_CLUSTER_FINAL.py is the authoritative final uncertainty script for M1 coefficient and RP50 intervals. It enumerates all 256 Rademacher sign patterns across the eight challenge-batch clusters. 7. The frozen software environment used Python 3.13.5, NumPy 2.3.5, SciPy 1.17.0, statsmodels 0.14.6, and patsy 1.0.2. 8. Figure source data are provided in 03_FIGURE_SOURCE_DATA.zip. Current Fig. 5 contains panels a–e only; Supplementary Fig. S1 corresponds to the frozen predictive-performance figure source. 9. Supplementary Tables S1–S8 are provided in 04_TABLE_SOURCE_DATA.zip. Main-text table values map directly to the authoritative frozen result files listed in MAIN_TABLE_SOURCE_MAP.csv. 10. The design limitations documented in the provenance files must be retained when interpreting antecedent-state effects.
Institutions
- China Three Gorges UniversityHubei, Yichang