pgs

utility for downloading and preprocessing Polygenic risk scores

Utilities to download PGS catalog data, and then apply those weights to compute polygenic risk scores for individuals in the UK Biobank. Only works for PGS files that provide rsids (instead of only chromo/bp), but this is most of the PGS catalog data.

Download PGS Catalog scores

use the functionality in download_pgs.py:

download_multiple_pgs('all')
# or e.g.
download_multiple_pgs(['PGS000001', 'PGS000002'])

This will download all scoring files and metadata to DST (defaults to pgs_data in the working directory).

Apply risk scores on genetic data

Using imputed data

This will first convert the UKB bgen files into plink pgen files, then join all chromosomes into one big file, and then run plink2 --score on this. There might be a slightly better way using e.g. PRSIce2 or LDpred2 instead of plink2.

To merge and convert the bgens use merge_bgen_to_pgen.sh. First, replace iids_with_retinal_scan.txt with a list of your iids with retinal fundus images. If you change the MLOCAL variable, you also need to change the paths in the pmerge_list.txt files. You will also need to update the paths to your .bgen and .sample files. Finally, run the script.

Next, run the create_pgs.sh script. You need to be in an environment with python and pandas installed. If you're using a newer version of the PGS catalog, you might need to update the number of PGS scores to include the most current number.

Note that some of the PGS Catalog files will be skipped because they only provide weights based on chromosome-positions, instead of rsids; plink2 only works with rsids so far.

Using only microarray data

You can run the analysis on unimputed microarray data (bed-files) only: you only need to modify a couple of commands in the two .sh files and won't need to convert away from bed files (but still need to merge them, afaik).

Note: this should give reasonable results but will only use a small fraction of the actual SNPs provided by PGS catalog. It's going to be a lot faster and more disk-space efficient than with imputed data, though.