roc

Visualizations Cookbook

Receiver Operating Characteristic Curves

A Single ROC Curve with Bootstrapped 95% Confidence Intervals

Code only work for Python 3.10 and above.

python -m wanshi.visualizations.roc \
    --outpath=/project/path/single_roc.svg \
    --target-label=IDH \
    --true-label=MUT \
    --n-bootstrap-samples=1000 \
    /path/to/patient-preds.csv

ROC Curves for a Cross-Validation Experiment

python -m wanshi.visualizations.roc \
    --outpath=/project/path/crossval_rocs.svg \
    --target-label=IDH \
    --true-label=MUT \
    /path/to/fold-*/patient-preds.csv

ROC Curves for Different Subgroups

python -m wanshi.visualizations.roc \
    --outpath=/project/path/roc_subgroups.svg \
    --target-label=IDH1 \
    --true-label=MUT \
    --n-bootstrap-samples=1000 \
    --clini-table=/path/to/clini.xlsx \
    --subgroup-label=Subtypes_2021 \
    --figure-width=6 \
    /path/to/patient-preds.csv

--figure-width=6 is used to increase the figure width to 10 inches, in order to make space for the longer text in this type of plot.

A Single ROC Curve Colored by Threshold

python -m wanshi.visualizations.roc \
    --outpath=/project/path/threshold_roc.svg \
    --target-label=IDH \
    --true-label=MUT \
    --threshold-cmap=viridis \
    /path/to/patient-preds.csv