Abstract:
|
Genetic polygenic scores (PGS) are a weighted sum of risk alleles carried by an individual, and can be used to represent an individual’s genetic risk of disease. Recent Bayesian PGS methods jointly estimate effect estimates, adjusting for linkage disequilibrium between variants, to determine posterior effect estimates from the results of genome-wide association studies (GWAS). In European Ancestry samples, these new methods have been widely successful at increasing PGS performance to identify high genetic risk individuals. We have adapted PRS-CS, a Bayesian hierarchical PGS method that uses continuous shrinkage priors, to incorporate GWAS summary statistics from multiple ancestries to create a single score that can be applied to all ancestries. Consistent with the hypothesis that most causal variants are shared across ancestries, our method adjusts each ancestry-specific GWAS by ancestry-specific LD and then combine effect estimates in a Bayesian meta-analysis with random effects for ancestry, with an option to include ancestry-specific shrinkage parameters. We demonstrate our methods in simulations and multi-ancestry cohorts.
|