ShinyQC Inspector

ShinyQC Inspector explores expression-PCA structure alongside technical QC metrics. It provides interactive PCA panels, ranks samples by QC/PCA atypicality, and can export HTML reports. Batch utilities generate the same report format for multiple cohorts, such as TCGA.

Quick start

Use R 4.5.1 (the version recorded in renv.lock) and restore the locked project environment:

renv::restore()
shiny::runApp()

The local renv/library directory is intentionally ignored. A new checkout therefore needs outbound package access for renv::restore() before the app is run.

The app accepts a normalized expression matrix with a Gene column, sample metadata, and QC metadata. The selected sample-ID columns must match the count matrix sample names after the app's underscore-to-hyphen normalization.

Batch reports

Create a manifest, then render per-cohort reports:

Rscript scripts/create_tcga_manifest.R /path/to/tcga_root /path/to/tcga_manifest.tsv
Rscript scripts/batch_qc_reports.R /path/to/tcga_manifest.tsv /path/to/report_output
Rscript scripts/summarize_tcga_qc_pc_axes_from_html.R /path/to/report_output /path/to/summary_output

batch_qc_reports.R --help describes the required manifest columns. The detailed analysis design and its provisional thresholds are in docs/tcga_qc_bias_scoring_plan.md.

Validation

Run the repository smoke test from the project root:

Rscript tests/run_smoke_tests.R

It generates a temporary synthetic cohort, renders a report, summarizes the resulting report, and checks that the Shiny app and helper functions load. This does not replace review with real cohort data.