Download LD Reference

Linkage-Disequilibrium (LD) matrices, which record pairwise correlations between genetic variants, are required as input to the VIPRS model. To facilitate running the model on GWAS data from diverse ancestries, we computed LD matrices for 6 continental populations represented in the UK Biobank.

The sample sizes reported below are restricted to unrelated individuals in the UK Biobank. The matrices were computed using the block LD estimator, where we only record pairwise correlations between variants in the same LD block. The LD blocks are defined by LDetect. The matrices were computed using the sister package magenpy and were then quantized to int8 data type for enhanced compressibility.

For European samples, we also provide LD matrices that record pairwise correlations for up to 18 million variants. This matrix is available for download via Zenodo.

For more details on QC criteria, data preparation, etc., please consult our manuscript:

Zabad, S., Haryan, C. A., Gravel, S., Misra, S., Li, Y. (2025). Toward whole-genome inference of polygenic scores with fast and memory-efficient algorithms. The American Journal of Human Genetics, 112(7), 1528-1546. https://doi.org/10.1016/j.ajhg.2025.05.002

New since July 2026: cloud-streamable LD matrices

VIPRS can now read the UK Biobank LD matrices directly from Hugging Face using hf:// paths. This means you can point VIPRS to the LD matrices without pre-downloading and extracting the archive files.

Streaming LD matrices from Hugging Face 🤗

The LD matrices are available as chromosome-level zip files on Hugging Face. To use this feature, install huggingface_hub in the environment where you run VIPRS:

python -m pip install huggingface_hub

You can then pass the appropriate hf:// wildcard path to the -l / --ld-panel argument, for example:

viprs_fit \
  --ld-panel "hf://datasets/shz9/ukb-ld/EUR/chr_*.zip" \
  --sumstats path/to/sumstats.txt \
  --output-dir output/viprs_fit

The available ancestry-specific Hugging Face paths are:

Code	Ancestry group	Sample size	Hugging Face LD path
EUR	European	362446	hf://datasets/shz9/ukb-ld/EUR/chr_*.zip
CSA	Central/South Asian	8284	hf://datasets/shz9/ukb-ld/CSA/chr_*.zip
AFR	African	6255	hf://datasets/shz9/ukb-ld/AFR/chr_*.zip
EAS	East Asian	2700	hf://datasets/shz9/ukb-ld/EAS/chr_*.zip
MID	Middle Eastern	1567	hf://datasets/shz9/ukb-ld/MID/chr_*.zip
AMR	Admixed American	987	hf://datasets/shz9/ukb-ld/AMR/chr_*.zip

For a single chromosome, provide the exact chromosome path instead of the wildcard, for example hf://datasets/shz9/ukb-ld/EUR/chr_22.zip.

Downloading LD matrix archives

If you prefer to pre-download the matrices, the six ancestry groups and their corresponding archive links are listed below:

Code	Ancestry group	Sample size	Download
EUR	European	362446	GitHub or Zenodo
CSA	Central/South Asian	8284	GitHub or Zenodo
AFR	African	6255	GitHub or Zenodo
EAS	East Asian	2700	GitHub or Zenodo
MID	Middle Eastern	1567	GitHub or Zenodo
AMR	Admixed American	987	GitHub or Zenodo

To access and use these matrices for downstream tasks, consult the codebase of magenpy, our sister python package that implements specialized data structures for computing and processing large-scale LD matrices.

Bash Script for downloading/extracting LD matrices

Here is a bash script that can be used to download and extract the LD matrices for all 6 populations. The script uses the GitHub links provided above. Feel free to modify the script to suit your needs.

#!/bin/bash
output_dir="LD_matrices"
populations=("EUR" "CSA" "AFR" "EAS" "MID" "AMR")
extract=true

mkdir -p $output_dir

for pop in "${populations[@]}"
do
    echo "Downloading LD matrix for $pop"
    wget -O $output_dir/$pop.tar.gz "https://github.com/shz9/viprs/releases/download/v0.1.2/$pop.tar.gz"
    if [ "$extract" = true ]; then
        mkdir -p $output_dir/$pop
        tar -xf $output_dir/$pop.tar.gz -C $output_dir/$pop
    fi
done