Python implementation of SigClust, as described by Huang et al.

To install in development mode, use conda develop . after setting up the conda environment. If you accidentally run this more than once, conda will recognize that and not do anything. You can "uninstall" it project (i.e. un-link it) using conda develop -u ..

To run tests, use python -m unittest discover -s tests/

Use the .fit method, where data= is a pandas DataFrame or numpy matrix. Rows should be observations and columns should be features. Pass a list (or array) of cluster labels to labels=. Access the p-value from the procedure with sc.p_value.

from sigclust import SigClust
import matplotlib.pyplot as plt

sc = SigClust()
sc.fit(data=dataframe, labels=[1, 1, 2, 2])

print(sc.p_value)
print(sc.z_score)

# Plot null distribution and test statistic
plt.hist(sc.simulated_cluster_indices)
plt.axvline(sc.sample_cluster_index)

cd tests/
python -m unittest discover

I borrow from Arthur Tilley's Python version of SigClust.