Containerized pipeline to amend a previously-run qc database with some new tests, and return performance and discrimination stats and plots. Run as an interactive environment:
docker container run -it -v $(pwd)/new-tests:/AutoQC/dev thisimage bash
where $(pwd)/new-tests/qctests
has the new qc tests to be integrated, and $(pwd)/new-tests/iquod.db has the database with the canonical tests pre-evaluated. All analysis artefacts will be dumped to $(pwd)/new-tests
.
User scripts:
experimental-qc.sh
: Evaluate AutoQC for the new qc tests, and append results to iquod.dboptimize-classifier.sh
: Run catchall.py, and append a column to iquod.db indicating resulting [T,F]x[P,N] classification for each profile. IfCUSTOM_PERF=1
, skip catchall, and instead copy/AutoQC/dev/custom_perf.json
in incatchall.json
's placegenerate-plots.sh
: Generate plots for each testing profile, categorized by [T,F]x[P,N].perf-uncertainty.sh
: Generate an uncertainty estimate on TPR and FPR by runningoptimize-classifier.sh
a bunch of times