NicheCompass Reproducibility

This repository contains the code to reproduce the analyses and benchmarking experiments performed in the NicheCompass manuscript. The NicheCompass source code can be found here.

Installation

Standard

1. Clone the nichecompass-reproducibility repository and navigate into it:
   git clone https://github.com/Lotfollahi-lab/nichecompass-reproducibility.git
cd nichecompass-reproducibility

2. (Optional) Install the Libmamba solver to make the installation faster:
   conda update -n base conda
conda install -n base conda-libmamba-solver
conda config --set solver libmamba

3. Create the nichecompass-reproducibility conda environment:
   conda env create -f envs/environment.yaml

   Install pyg dependencies with GPU support:
   conda activate nichecompass-reproducibility
pip install pyg_lib torch_scatter torch_sparse -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html
   where ${TORCH} and ${CUDA} should be replaced by the specific PyTorch and CUDA versions, respectively.

   To enable GPU support for JAX, after the installation run:
   conda activate nichecompass-reproducibility
pip install jaxlib==0.3.25+cuda${CUDA}.cudnn${CUDNN} -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

   For example, for CUDA 11.7, type:
   conda activate nichecompass-reproducibility
pip install jaxlib==0.4.7+cuda11.cudnn86 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

4. Create the deeplinc conda environment (for benchmarking deeplinc method which relies on legacy packages and is incompatible with dependencies of other methods):
   conda env create -f envs/environment_deeplinc.yaml

5. Create the cellcharter conda environment (for benchmarking cellcharter method which relies on legacy packages and is incompatible with dependencies of other methods):
   conda env create -f envs/environment_cellcharter.yaml

6. Create the stalign conda environment (for starmap_plus_mouse_cns analysis which requires specific dependencies and is incompatible with dependencies of other analyses):
   conda env create -f envs/environment_stalign.yaml

   Install STalign into the environment:
   conda activate stalign
pip install --upgrade "git+https://github.com/JEFworks-Lab/STalign.git"

Docker / Charliecloud Container

1. Clone the nichecompass-reproducibility repository:
   git clone https://github.com/Lotfollahi-lab/nichecompass-reproducibility.git
   

2. Clone the nichecompass repository:
   git clone https://github.com/Lotfollahi-lab/nichecompass.git

3. From the root repository that contains both the nichecompass and nichecompass-reproducibility repositories, run:
   docker buildx build --load --platform linux/amd64 --file nichecompass-reproducibility/envs/Dockerfile --tag nichecompass . --no-cache (macOS)

4. (For Charliecloud) Export the Docker image to a tarball:
   docker export $(docker create nichecompass) | gzip -c > ./nichecompass.tar.gz

Data & Models

All preprocessed data used in the manuscript and trained models are downloadable from GDrive.

Reference

@article{Birk2024Quantitative,
  title = {Quantitative characterization of cell niches in spatial atlases},
  author = {Birk, S. and others},
  journal = {bioRxiv},
  year = {2024},
  doi = {10.1101/2024.02.21.581428},
  url = {https://www.biorxiv.org/content/10.1101/2024.02.21.581428}
}