Case data and provenance

Download the inputs

The complete project supplies the helpers and all files. Individual inputs are also available below; preserve the matching metadata and manifests when reusing them.

Case Measurements Metadata
Airway Gene counts Paired samples
Proteomics LFQ values and flags Enrichment design
Cattle RNA panel, metabolites Animal covariates, RNA totals

Teaching-input checksums · Source checksums · Cattle identity audit · Cattle coding audit.

Licence texts: DEP Artistic 2.0, airway LGPL 2.1, LGPL 3 and incorporated GPL 3.

About these teaching extracts

These files contain actual public measurements. They are separate from the hypothetical fixtures described in data/README.md. Prepared on 8 October 2026. manifest.csv records MD5 checksums of the teaching inputs; R/case_helpers.R verifies them before reading. MD5 is used for accidental-change detection, not security authentication.

Airway RNA-seq

Source: airway 1.26.0, Bioconductor 3.20, Michael Love; experiment by Himes et al. (2014), DOI Himes et al. study, GEO GSE52778. Package: airway 1.26.0 package .

The package’s complete assay is exported as airway_counts.csv.gz (63,677 genes × 8 libraries). airway_samples.csv retains sample IDs, cell line and dexamethasone condition. CSV is a changed representation, not a reanalysis of sequencing reads. Gene IDs and integer counts are retained without filtering. The four untreated/treated pairs represent cultured primary smooth-muscle cell lines. See the package vignette for its annotation and counting procedure.

The package declares LGPL. Accompanying airway-LGPL-2.1.txt, airway-LGPL-3.txt and airway-GPL-3.txt retain the LGPL texts and the GPL text incorporated by LGPL-3. Original package and source: airway source repository . This CSV export is a modified format of the package data; no package authorship is claimed. The original study article is CC BY. Credit Himes et al. and the airway package when reusing.

Ubiquitin interaction proteomics

Source: DEP by Arne H. Smits and Wolfgang Huber, pinned repository revision b425d8d0db67b15df4b8bcf87729ef0bf5800256: pinned DEP source . Files data/UbiLength.rda and data/UbiLength_ExpDesign.rda. Source reference: Zhang et al. (2017), An Interaction Landscape of Ubiquitin Signaling, DOI Zhang et al. study .

ubiquitin_lfq.csv.gz is a modified column/format extract of 3,006 protein groups. It retains all twelve LFQ measurements, protein IDs, gene names and three analytical flags, and adds stable row labels PG0001 etc. ubiquitin_samples.csv retains the original label, condition and replicate design. It does not contain raw spectra, and no observations are imputed. Zeros remain in the extract and are recoded as unquantified only in the analysis. A protein group is not necessarily a uniquely identified protein or gene. These are affinity-enrichment replicates, not animal cohorts.

DEP declares Artistic-2.0. The accompanying DEP-Artistic-2.0.txt supplies its licence. Preserve the original attribution, identify this modified teaching extract, and retain access to the original standard data files at the pinned source above. This redistribution does not include publisher artwork or paper text.

Feedlot cattle disease study

Source: Li et al. (2022), Applying multi-omics data to study the genetic background of bovine respiratory disease infection in feedlot crossbred cattle, DOI Li et al. study . Public dataset: Borealis dataset , version 1.0, released 24 October 2022. Borealis declares CC0 1.0 (CC0 1.0 terms). RNA counts: GEO GSE217317, GEO GSE217317 . The study article is CC BY. Retain attribution and study context even where the dataset imposes no attribution requirement.

Inputs: Borealis files 411516 (phenotype) and 411515 (metabolites), plus GEO GSE217317_143_BRD_CON_readcounts.csv.gz and its two platform series matrices. source_checksums.csv records acquired source bytes. Original Borealis file checksums are verified against its version 1.0 metadata; the original files are XLSX bytes despite the archive’s .tab labels. For R import we separately use the repository’s ingested TSV exports. No original large files are committed here.

cattle_samples.csv retains the 138 phenotype records, with status, feedlot, sex, RNA batch, days on feed and three genomic breed-composition covariates. cattle_metabolites.csv retains all 139 source records and all 55 HMDB columns. The paper describes 54 metabolites; the teaching extract deliberately preserves the supplied columns and flags this discrepancy. Missing/zero measurements are preserved. No unsupported concentration unit or compound name is added.

RNA names lose _BRD/_NB, four-digit names receive the R prefix, and the explicit P4647 → P4687 correction follows the documented public preparation script: public identity/preparation source . No other fuzzy name matching is performed. All 143 processed count titles are matched to the two GEO platform sample records. cattle_identity_audit.csv preserves source title, canonical ID, GEO status, phenotype status, status-conflict flag and assay/metadata availability. Four records (U6623, U6450, U6325, Y3551) have GEO BRD characteristics but Control phenotype/metabolite labels and _NB count titles. The primary teaching analysis uses the canonical phenotype table and explicitly repeats the adjusted comparison after excluding these four controls (131 animals). This is a documented working choice, not verification of the correct biological labels. Phenotype sex 0 and metabolite sex 2 are both male; code 1 is female. Shared status/feedlot/recoded sex are checked. cattle_coding_audit.csv retains both source sex codes, including unavailable records.

cattle_rna_samples.csv preserves 143 RNA identities, original column titles and full-gene library totals. cattle_rna_panel.csv.gz contains the 300 highest mean-CPM genes across those 143 libraries, with ties resolved by gene ID. This label-independent selection is for compact exploration, not transcriptome-wide testing. CPM denominators always use full-gene totals.

The primary matched intersection has 135 animals (78 BRD, 57 controls). No missing phenotype is reconstructed. PCA uses 36 metabolite columns positive/observed in all 135 animals; association examples use feature-specific positive observations with sample counts and a minimum of 20 per disease group. No imputation is performed. The study used pen-matched controls, but pen/matched-set IDs are absent from these tables. Independent-animal working-model p-values therefore do not account for the original matching. The lesson does not reproduce the paper’s GWAS, differential-expression screen or integrated discovery counts.

Rebuild explicitly

Website rendering never downloads these sources. To rebuild, create a separate source directory and acquire these public files yourself:

  • Install airway 1.26.0 in a compatible Bioconductor 3.20 R environment.
  • Download UbiLength.rda and UbiLength_ExpDesign.rda from the pinned DEP revision above.
  • Download https://borealisdata.ca/api/access/datafile/411516 as cattle_metadata.tsv and 411515 as cattle_metabolites.tsv.
  • Download the same two endpoints with ?format=original as cattle_metadata_original.xlsx and cattle_metabolites_original.xlsx.
  • From https://ftp.ncbi.nlm.nih.gov/geo/series/GSE217nnn/GSE217317/, download suppl/GSE217317_143_BRD_CON_readcounts.csv.gz as cattle_counts.csv.gz, and the matrix/GSE217317-GPL23295_series_matrix.txt.gz and GPL26012 counterpart as cattle_series1.txt.gz and cattle_series2.txt.gz.
  • Verify downloaded bytes against the existing source_checksums.csv before rebuilding. Sources can change; a changed checksum requires review, not automatic acceptance.

From the project root, run Rscript scripts/prepare_case_data.R <source-directory>. The script verifies original Borealis checksums and all existing source-manifest checksums before exporting; it checks GEO identity/status differences, writes audits and regenerates manifests in a temporary staging directory, then copies validated outputs into the project. It stops if run outside the course root. Review any manifest differences; compressed outputs may differ with compression-library versions. Keep downloads and installation libraries outside the repository.