Polygenic risk scores from many GWAS, in minutes

GigaPRS constructs polygenic risk scores from summary statistics of one or many GWAS, spanning multiple traits and ancestries. It matches the accuracy of the best existing tools, is substantially more accurate for African and admixed individuals, and typically runs in under 30 minutes on a single CPU.

Run GigaPRS in your browser Browse results for our traits Download pre-computed scores

+10–12%

Higher PRS accuracy than PRS-CSx for African and admixed individuals, across 18 quantitative traits.

~50×

Faster than PRS-CSx and JointPRS when combining GWAS of several ancestries.

101

Disease endpoints with pre-computed multi-ancestry SNP effect sizes.

How it works

Overview of the two GigaPRS steps — **Overview of GigaPRS.** Step 1 estimates SNP effect sizes for each ancestry, combining primary and informative secondary summary statistics; Step 2 computes ancestry-specific PRS for the target individuals and weights them by each individual's inferred ancestry.

Step 1 — estimate SNP effect sizes

For each ancestry, GigaPRS builds a single set of combined summary statistics: a weighted average of the GWAS for the target trait (primary) and of GWAS for genetically correlated traits (secondary). A stepwise algorithm keeps a secondary set only if it raises the effective sample size by at least 5%. Effect sizes are then estimated with variational Bayes under a BayesR-type prior, using the BLD heritability model. Because the joint analysis becomes a series of single analyses, run time grows only linearly with the number of ancestries.

Step 2 — compute scores

The output is one file containing effect sizes and allele frequencies for every ancestry. When scoring target individuals, GigaPRS infers each person's ancestry proportions and weights the ancestry-specific scores accordingly, so it works for individuals of unknown, mixed or admixed ancestry.

Read the full method and technical details →

Ways to use GigaPRS

In this website. Run GigaPRS runs Step 1 entirely in your browser — your summary statistics never leave your computer. Use our hosted LD references and summary statistics, or your own.

On the command line. GigaPRS is part of LDAK:

./ldak.out --giga-prs gigaprs --corslist cors.txt --sumslist sums.txt
./ldak.out --calc-scores prs --scorefile gigaprs.effects --bfile target

Pre-computed scores. The PRS library has ready-to-use multi-ancestry effect sizes; Browse traits to explore and download summary statistics.

Citation

Speed D, Ning C, Yang S, Pain O, Balding D, Cao C, Hof J. Fast and accurate construction of polygenic risk scores using summary statistics from multiple GWAS. In preparation.