concordex can be installed via pip

pip install git+https://github.com/pachterlab/concordex.git

After installing, concordex can be run by simply typing concordex on the command line

usage: concordex [-h] <CMD> ...

concordex 0.0.1: Assess clustering

positional arguments:
  <CMD>
    map
    stat

optional arguments:
  -h, --help  show this help message and exit

The concordex map command generates the cluster x cluster matrix where each entry is the fraction of cluster i's neighbors which are in cluster j (i, j are row, column indices into the matrix). The matrix can be generated by running

concordex map -a assignments.txt -o map.csv mtx_file

This matrix can be used to visualize the concordex measure (in leiu of producing a non-linear 2-dimensional embedding).

The concordex stat command generates the concordex statistic as described in this manuscript. The stat can be generated by running

concordex stat -a assignments.txt -o map.csv mtx_file

and with python (in a jupyter notebook)

from concordex.concordex_stat import concordex_stat
import anndata

adata = anndata.read_h5ad("adata.h5ad")

mtx = adata.X # or adata.layers["my matrix"], should be scipy sparse matrix
assignments = adata.obs["assignments"].values

trace, random_trace, corrected_trace = concordex_stat(mtx, assignments)