XPXP: improving polygenic prediction by cross-population and cross-phenotype analysis

Area of Science:

• Genetics
• Bioinformatics
• Personalized Medicine

Background:

• Genome-wide association studies (GWASs) enable polygenic risk score (PRS) development for disease prediction.
• Current PRSs show reduced accuracy in non-European populations due to biases in training data.
• Improving PRS for under-represented populations like African and East Asian groups is an urgent clinical need.

Purpose of the Study:

• To develop a novel method for constructing accurate PRSs in under-represented populations.
• To leverage large European datasets and genetically correlated phenotypes for improved PRS.
• To enhance PRS accuracy by incorporating population-specific and phenotype-specific effects.

Main Methods:

• Proposed a cross-population and cross-phenotype (XPXP) method for PRS construction.
• Utilized biobank-scale European datasets and multiple GWASs of correlated traits.
• Incorporated population-specific and phenotype-specific effects into the PRS model.

Main Results:

• XPXP demonstrated superior performance compared to existing PRS methods in simulation and real-world analyses.
• Height PRSs constructed by XPXP showed 9% and 18% improvement in predicted R2 for East Asian and African populations, respectively.
• XPXP significantly enhanced the ability to stratify individuals by genetic risk for type 2 diabetes.

Conclusions:

• The XPXP method offers a robust approach to constructing accurate PRSs for under-represented populations.
• This advancement holds significant potential for improving personalized medicine and disease risk prediction globally.
• The XPXP software and analysis code are publicly available to facilitate further research and application.