snputils 0.1.1

SNP utils

snputils

Installation

We recommend creating a fresh Python >= 3.8 conda environment. This package can be easily installed using pip after cloning the repository:

(base) $ conda create -n snputils python=3.10
(base) $ conda activate snputils
(snputils) $ git clone https://github.com/AI-sandbox/snputils.git && cd snputils
(snputils) $ pip install -e '.[testing,jupyter]'

Alternatively, you can install the package directly from the repository without cloning it:

(base) $ pip install git+https://github.com/AI-sandbox/snputils.git

Optional Dependencies

By default, torch is not installed when you install snputils using pip install snputils. This is because torch is only required for certain PCA functionalities within snputils.

If you wish to use the optimized PCA implementation for GPU using the PyTorch backend, you can install the extra dependencies using:

pip install snputils[gpu]

This will install torch along with snputils.

Alternatively, you can install torch manually if you prefer:

pip install torch

Please note that installing torch might also install additional NVIDIA packages if CUDA support is enabled.

Use snputils tools

PCA

To directly run Principal Component Analysis (PCA) on SNP data, plotting the results, and saving the principal components on a .npy file, simply call snputils pca with the required arguments, see snputils pca --help for more information.

An example of running PCA on a .vcf file, saving the plot in fig.png and the principal components in pc.npy:

> snputils pca --vcf_file /dataset/path/hapmap3.vcf --fig_path fig.png --npy_path pc.npy --backend sklearn