Installation¶

Step 1: Create and activate a conda env (recommended)¶

ONTraC supports Python 3.10, 3.11, and 3.12 for now

conda create -y -n ONTraC python=3.11
conda activate ONTraC

Add this kernel to jupyter (recommended)

pip install ipykernel
python -m ipykernel install --user --name ONTraC --display-name "Python 3.11 (ONTraC)"

Step 2: Install ONTraC¶

Using pip

pip install ONTraC
python -m ipykernel install --user --name ONTraC --display-name "Python 3.11 (ONTraC)"

Use this command if you want to visualise the results by ONTraC_analysis.

pip install "ONTraC[analysis]""

Using conda

NOTE: For ARM-based macOS, we recommend installing via pip for now, as the dependency package, pytorh-geometric, does not have a conda build for it. NOTE: For x86-based macOS, conda install only supports Python 3.10 and 3.11 for now. WARNING: Installing with conda can be very slow.

conda install -c gyuanlab -c pytorch -c pyg -c default -c nvidia -c conda-forge ontrac

Install the development version from GitHub

git clone git@github.com:gyuanlab/ONTraC.git .
cd ONTraC
pip install .
pip install ".[analysis]"

Optional dependencies¶

Deconvolution

STdeconvolve is a reference-free cell-type deconvolution method. Please install through the following command if you want to use it in the preprocessing step when analyzing low resolution data.

conda install bioconda::bioconductor-stdeconvolve -y