Add TAU Community Detection

[HillelCharbit]

Package submission

Name: TAU Community Detection

TAU is an evolutionary graph community-detection algorithm combining genetic
search with Leiden refinement. It provides native support for clustering
Scanpy neighbor graphs stored in AnnData.

Use of scverse data structures

TAU reads graph connectivities from AnnData.obsp, including Scanpy's default
neighbor graph, custom obsp keys, and graphs selected through
neighbors_key.

It stores categorical cluster assignments in AnnData.obs and records
clustering parameters and modularity in AnnData.uns.

Mandatory checklist

• Publicly available under an OSI-approved license
• Provides versioned releases
• Installable from PyPI
• Includes automated tests
• Includes continuous integration
• Public API documentation is available in the README
• Uses AnnData where appropriate
• I am an author and maintainer and agree to the package being listed

Links

• Repository: https://github.com/HillelCharbit/TAU
• PyPI: https://pypi.org/project/tau-community-detection/
• Publication: https://doi.org/10.1093/pnasnexus/pgad180

Scanpy integration

Installation:

pip install "tau-community-detection[scanpy]==1.4.7"

Example:

import scanpy as sc
from tau_community_detection.tl import tau

sc.pp.neighbors(adata)
tau(
    adata,
    key_added="tau",
    resolution=1.0,
    random_state=42,
)

The integration is tested for graph selection, weighted and unweighted
clustering, reproducibility, isolated vertices, copy and in-place behavior,
multiple resolutions, and comparisons with Leiden and Louvain.