projectionSVD (v0.2.0)

projectionSVD is a small command-line program written in Python/Cython to project a dataset onto a principal component space based on genotype data. It takes binary PLINK files as genotype input and works with PCA output from programs like halkoSVD, PLINK, and PCAone. projectionSVD requires estimated allele frequencies, eigenvalues and SNP loadings to perform the projection.

Installation

# Option 1: Build and install via PyPI
pip install projectionSVD

# Option 2: Download source and install via pip
git clone https://github.com/Rosemeis/projectionSVD.git
cd projectionSVD
pip install .

# Option 3: Download source and install in a new Conda environment
git clone https://github.com/Rosemeis/projectionSVD.git
conda env create -f projectionSVD/environment.yml
conda activate projectionSVD

You can now run the program with the projectionSVD command.

Quick usage

# Check help message of the program
projectionSVD -h

# Perform projection using PCAone output
projectionSVD --bfile new --freqs old.afreq --eigvals old.eigvals --loadings old.loadings --threads 32 --out new

# Outputs eigenvectors of new dataset (new.eigvecs)

Options

• --freqs-col, specify which column to use in frequency file (6)
• --batch, process projection in batches of specified number of SNPs (8192)
• --raw, only output eigenvectors without FID/IID