In this repository, you can find the codes and datasets that were used to write [1].

The pCMF package is available here: https://gitlab.inria.fr/gdurif/pCMF

[1] Durif, Ghislain, Laurent Modolo, Jeff E. Mold, Sophie Lambert-Lacroix, et Franck Picard. « Probabilistic Count Matrix Factorization for Single Cell Expression Data Analysis ». arXiv:1710.11028 [stat], 2019. http://arxiv.org/abs/1710.11028.

Copyright (c) [2019] [Ghislain DURIF]

Applies to all files except the ones located in the 'third_party_src' and 'data/single_cell' sub-directories, which belong to the respective authors of the respective softwares or data.

Licensing: all files except the ones located in the 'third_party_src' and 'data/single_cell' sub-directories are distributed under the CECILL v2.1 license. You can check the license agreement in the './LICENSE.txt' file.

If you encounter any issue, you can contact us at 'gd DOT dev AT libertymail DOT net'.

This is some research code, that is not packaged for distribution, however it should be quite easy to use it to replicate the experiment from the paper, at least on Unix OS, tested on GNU/Linux (Ubuntu 16.04 and 18.04) but not on MacOS.

Download the repository:

git clone https://gitlab.inria.fr/gdurif/pCMF_experiments
cd pCMF_experiments

Set the environment (R and Python packages are installed locally in pCMF_experiments/.R_libs and pCMF_experiments/.pyenv respectively):

cd pCMF_experiments
source set_env.sh
Rscript set_pkg.R
bash set_python_zifa.sh

To benefit from mutli-threading (which increases the performance), you can set the NCORE variable in the pCMF_experiments/.Renviron file or set the environment variable NCORE before sourcing the bash script pCMF_experiments/set_env.sh:

export NCORE=8
source set_env.sh

You can then run scripts in the pCMF_experiments/experiment sub-folders, either in Rstudio (there is a specific project file pCMF_experiments/pCMF_experiments.Rproj), or with R command line tools (source the file pCMF_experiments/set_env.sh will set environment variables pointing to the project specific .Renviron and .Rprofile files.

Source codes defining experiment are located in the corresponding sub-folders. Some utilitary functions are defined in pCMF_experiments/src/R.

We mention that, for the moment, our codes use the low level functions run_XXX of the pCMF package, and not the wrapper pCMF. A new version based on this more simple wrapper will be soon published.