How to calculate PRS weights for FinnGen data

There is a pipeline ready in sandbox that allows to calculate PRS weights based on the FG summary statistics. In order to do so we've build a custom LD panel based on the Finnish panel used for imputation. More information on the pipeline can be found from github.

Example: .json and .wdl- files for running the pipeline can be found in sandbox at /finngen/library-green/scripts/prs_weights_fg:

• .json- file: sandbox_fg_weights.json and

• .wdl- file: sandbox_fg_weights.wdl

You can copy these files and edit the .json- file according to your needs as follow:

• "finngen_weights.pheno_list": path to a tab- delimited .txt file including phenotype(s) to run and their sample sizes, each phenotype on its own row, if multiple phenotypes:

RX_STATIN    218792
T2D    211766
E4_HYTHY_AI_STRICT    198472

• 'finngen_weights.munge_sumstats.file_root': Path to summary statistics file(s). Replace the name of the phenotype with PHENO . Example for DF5 summary statistics file: /finngen/library-green/finngen_R5/finngen_R5_analysis_data/summary_stats/release/finngen_R5_PHENO.gz

In addition, check the column names in "finngen.weights.mungesumstats.columns" to match the ones in your summary statistics file. The ones In the example .json do correspond to the DF5 summary statistics.

Once your .json- file has been edited accordingly, you can submit your job via command line using command:

finngen-cli request-workflow --wd /path/to/sandbox_fg_weights.wdl \
    --input /path/to/sandbox_fg_weights.json

Remember to save your [WORKFLOW_ID] to track your job and check the results!

Once your job is ready, you can find your results in: /finngen/pipelines/cromwell/workflows/finngen_weights/[WORKFLOW_ID]/call-weights/shard-#

There you will find two weight files for each endpoint:

• sandbox_test_RX_STATIN.weights.txt (variants as chrom_pos_ref_alt) and

• sandbox_test_RX_STATIN.weights.rsid (variants as rsid)

Please see also Pietros presentation and slides at User's meeting.