Metadata-Version: 2.1
Name: ClustAssessPy
Version: 1.0
Summary: Python package for systematic assessment of clustering results stability.
Author: Rafael Kollyfas
Author-email: rk720@cam.ac.uk
License: MIT
Keywords: clustering,stability,assessment,machine learning,graph,network,community,detection
Requires-Python: >=3.7
License-File: LICENSE.txt
Requires-Dist: numpy
Requires-Dist: pandas
Requires-Dist: scanpy
Requires-Dist: umap-learn
Requires-Dist: seaborn
Requires-Dist: matplotlib
Requires-Dist: scipy
Requires-Dist: networkx
Requires-Dist: plotnine
Requires-Dist: pynndescent
Requires-Dist: leidenalg
Requires-Dist: louvain
Requires-Dist: igraph

# ClustAssessPy

ClustAssessPy is a lighter Python adaptation of ClustAssess (R). This Python version includes ClustAssess's main functions, such as calculating all ECS-related metrics and evaluating and plotting clustering stability in the dimensionality reduction, graph building, and graph clustering components.

The package allows for a data-driven assessment of optimal parameter values for dimensionality reduction (e.g. choosing between UMAP and PCA), graph type (e.g., shared-nearest-neighbors vs nearest neighbors), identification of the most stable community detection algorithm (e.g., leiden vs louvain), resolution that produces the most stable partitions.
