ktest

This package implements kernel tests such as kernel Fisher Discriminant Analysis (kFDA) and Minimum Mean Discrepency (MMD) introduced in [1].

Note regarding code used in publication

The code for the analysis presented in [1] are available in a dedicated repository here.

The source code for experimental_data and squair scripts are based on the branch publication_genome_biology. The reversion scripts are based on the branch publication_genome_biology_reversion.

See https://github.com/AnthoOzier/ktest_experiment_genome_biology_2024 for more details.

Licensing and authorship

See the dedicated LICENSE.md and AUTHORS.md files.

Python package

See python directory.

Platform Support

• macOS x86 (before Apple switched to ARM processors): Requires Python < 3.13 (due to PyTorch limitations).
• Linux/macOS ARM: Supports Python >= 3.8.

See python/pyproject.toml file for more details.

Install

Latest development version:

pip install ktest@git+https://github.com/LMJL-Alea/ktest@main#subdirectory=python

Tutorials

See the dedicated tutorials folder in the project root directory for notebook tutorials in Python.

R package

See r-pkg directory.

Install

remotes::install_github("LMJL-Alea/ktest", ref = "main", subdir = "r-pkg")

References

[1] Anthony Ozier-Lafontaine, Camille Fourneaux, Ghislain Durif, Polina Arsenteva, Céline Vallot, Olivier Gandrillon, Sandrine Gonin-Giraud, Bertrand Michel, Franck Picard. Kernel-based testing for single-cell differential analysis. Genome Biol 25, 114 (2024). doi:10.1186/s13059-024-03255-1, hal-04214858