Improving polygenic prediction in ancestrally diverse populations

Area of Science:

• Genetics
• Bioinformatics
• Population Genetics

Background:

• Polygenic risk scores (PRS) show limited predictive performance across different ancestral populations.
• This limitation stems from genome-wide association studies (GWAS) predominantly featuring European ancestry, hindering PRS clinical utility in non-European groups and potentially worsening health disparities.
• Current efforts to increase diversity in genomic research have expanded non-European GWAS, but many studies remain underpowered.

Purpose of the Study:

• To develop and evaluate PRS-CSx, a novel method for constructing polygenic risk scores.
• To enhance cross-population prediction accuracy by integrating summary statistics from multiple GWAS.
• To address the challenge of limited PRS transferability and improve its clinical value in diverse populations.

Main Methods:

• Introduced PRS-CSx, a new PRS construction method that integrates GWAS summary statistics from multiple populations.
• Employed a shared continuous shrinkage (CS) prior to couple genetic effects across populations.
• Leveraged linkage disequilibrium (LD) diversity across discovery samples for improved effect size estimation.

Main Results:

• PRS-CSx demonstrated superior performance compared to existing methods across various traits, genetic architectures, and GWAS sample sizes in simulations.
• The method showed improved prediction of quantitative traits in non-European populations.
• PRS-CSx also enhanced the prediction of schizophrenia risk in non-European populations.

Conclusions:

• PRS-CSx offers a significant advancement in cross-population polygenic prediction by effectively integrating multi-ancestry GWAS data.
• The method improves the accuracy and clinical utility of PRS for non-European populations, potentially mitigating health disparities.
• PRS-CSx provides a computationally efficient and robust approach for personalized genetic risk prediction across diverse ancestries.