The pgsc_benchmark tool aims to facilitate the widespread performance benchmarking
of polygenic scores (PGS) curated in the PGS Catalog.
It has been designed to run directly on the output files produced by the PGS Catalog
Calculator (pgsc_calc), providing a
reproducible method for evaluating score performance. Users are able to run the tool
using their own study cohorts and phenotype data, generating a consistent and directly
comparable set of performance metrics across all desired PGS. The aim is to provide
the human genomics community with a more complete understanding of each score’s predictive
ability, allowing the best performing PGS to be identified across a range of diverse
populations and ancestries.
The pgsc_benchmark tool is available to download as either a Docker or Singularity image. Please
see the user manual
for download instructions.
- Docker image
- Singularity image
An R implementation is also provided in case any issues are encountered running Docker or Singularity on your system. Please see the separate R user manual here.
In brief, four main steps are involved in the benchmarking process:
- Calculate PGS for your cohort using the PGS Catalog Calculator
- Prepare a demographic/phenotype file for your cohort
- Run the benchmarking tool on the outputs of (1) and (2)
- Post-processing and data interpretation
A full description of each step is provided in the user manual, along with detailed explanations of the output files produced.
Written by Joel T. Gibson
Cardiovascular Epidemiology Unit
Department of Public Health and Primary Care
Victor Phillip Dahdaleh Heart and Lung Research Institute
University of Cambridge