projectPCA

Project genomes onto pre-computed principal components widely used in ancient DNA. Enables fast analysis without re-computing the principal components. The software accepts ancient DNA data in eigenstrat or PLINK format as input. No modern samples are required, as the packages include the pre-computed PCA weights and PC coordinates for relevant modern samples (based on publicly available Human Origin array data).

Installation

The package projectPCAis available as a Python package via pip. To install, simply run a version of:

python3 -m pip install projectPCA

List of available PCAs

As of early 2026, three pre-computed PCAs are officially bundled into projectPCA. The bracket denotes the code you can use for all this PCA.

• HO Westeurasia (HO) Standard Western Eurasian PCA, which is widely used in aDNA studies. PC1 corresponds to West-East, and PC2 to North-South.

• HO Eurasian (EUAS) Standard whole-Eurasian PCA, widely used in aDNA studies. Excellent to resolve West versus East Asian ancestry (on PC1). PC2 generally corresponds to North-South.

• HO Mediterranean (MED) This PCA is an extension of the standard Western Eurasian PCA (see above) and includes additional North African populations. That gives a more useful resolution for North African ancestry. The PCA is based on the one from the Puncic paper (Ringbauer et al 2025). Important: You have to use maf=0.001 here, which is different from the default. See an example Vignette here.

Usage

Project single Samples

To project onto a PCA, the key function is project_eigenstrat. To import it and run a single sample, use:

from projectPCA.run import project_eigenstrat

project_eigenstrat(es_path="/mnt/archgen/Autorun_eager/eager_outputs/TF/SUA/SUA002/genotyping/pileupcaller.double",
                   pca="HO", es_type="default")

This function also returns the dataframe with PCA coordinates. Note that the input path is the path of the eigenstrat files up to .geno but without the suffix.

The keyword pca denotes which PCA type to project onto (see above).

If you want to save the figure, you can add the keyword fig_path="". If this string is filled in, the program saves the resulting figure there. If the path ends in .html, the figure is saved as an interactive plot, where you can hover over the individuals to see their labels (both ancient and modern reference samples). Otherwise, the standard matplotlib libraries are used to plot and save the figure (including in .png or .pdf format, based on the extension you provide).

project_eigenstrat(es_path="/mnt/archgen/Autorun_eager/eager_outputs/TF/SUA/SUA002/genotyping/pileupcaller.double",
                   pca="EUAS", es_type="unpacked_fast", plot_bgrd_c=False, fig_path='./figs/SUA002_EUAS.html')

Project multiple samples

It is also possible to project multiple samples. For this, you can use the keyword iids=[]. If the keyword is empty (the default), all samples in a file are projected and plotted. If you specify a list of individuals, only individuals with these IDs are projected.

Project PLINK files

To project PLINK files, you can use the keyword es_type="plink", and provide the path of the PLINK file up to the suffix:

project_eigenstrat(es_path="/mnt/archgen/users/hringbauer/git/EPIDEMIC/output/plink/bd_ptn_335",
                   pca="EUAS", es_type="plink", iids=[],
                   plot_bgrd_c=False, verbose=True, flip=True, 
                   fig_path='/mnt/archgen/users/hringbauer/git/projectPCA/figs/ptn335PLINK_EUAS.html')

@Harald Ringbauer, 2026