Dissecting the Predictive Accuracy of Polygenic Indexes for Behavioral Phenotypes Across Genetic Ancestries

Area of Science:

• Genetics
• Bioinformatics
• Population Genetics

Background:

• Polygenic indexes (PGIs) trained on European ancestry data exhibit reduced predictive power in non-European populations.
• The portability of PGIs across diverse ancestries, particularly for behavioral and social traits, is understudied.
• Understanding factors contributing to PGI accuracy loss in different ancestries is crucial for equitable genomic applications.

Purpose of the Study:

• To systematically analyze the portability of genome-wide PGIs across 52 health, behavioral, and social phenotypes in non-European ancestries.
• To investigate the influence of genetic ancestry, trait type, and GWAS methodology on PGI predictive performance.
• To identify key genetic factors (e.g., linkage disequilibrium, allele frequency differences) driving PGI portability differences.

Main Methods:

• Utilized UK Biobank and Health and Retirement Study data for genome-wide PGI analysis.
• Assessed PGI portability across African, East Asian, and South Asian genetic ancestries.
• Compared PGIs derived from standard Genome-Wide Association Studies (GWAS) with those from family-based GWAS.
• Quantified cross-ancestry heritability and analyzed factors contributing to accuracy loss.

Main Results:

• Significant reductions in PGI predictive power were observed in non-European ancestries: African (24%), East Asian (37%), and South Asian (51%) relative accuracy.
• Biologically proximal traits demonstrated greater portability than behavioral and social traits.
• Linkage disequilibrium and allele frequency differences were major contributors to accuracy loss in African ancestries (82%), with smaller contributions in East (34%) and South Asian (25%) ancestries.
• Family-based GWAS PGIs showed modest improvements in portability for specific traits, such as BMI in African ancestry.

Conclusions:

• PGI portability is a significant challenge, with substantial accuracy loss in non-European populations.
• Trait heritability and underlying genetic architecture differences across ancestries influence PGI performance.
• Addressing population-specific genetic variation and refining GWAS methodologies are essential for improving PGI equity and utility.