Polygenic scoring accuracy varies across the genetic ancestry continuum

Area of Science:

• Genetics
• Bioinformatics
• Population Health

Background:

• Polygenic scores (PGSs) are crucial for predicting disease risk but face challenges in portability across diverse populations.
• Current assessments of PGS portability often overlook individual variations within populations.
• Equitable application of PGSs is hindered by their limited generalizability.

Purpose of the Study:

• To investigate the impact of genetic distance on polygenic score accuracy across diverse ancestries.
• To evaluate the effectiveness of a continuous measure of genetic distance versus discrete ancestry clusters for PGS portability.
• To identify factors influencing the variation in PGS performance within and between populations.

Main Methods:

• Utilized large biobanks (ATLAS and UK Biobank) with diverse participants.
• Assessed PGS accuracy across a continuum of genetic ancestries.
• Quantified the relationship between genetic distance (GD) and PGS accuracy using Pearson correlation.
• Compared PGS performance between individuals with European and Hispanic/Latino American ancestries.

Main Results:

• PGS accuracy decreases along the continuum of genetic ancestries, even within ostensibly homogeneous groups.
• Genetic distance (GD) strongly correlates with PGS accuracy (r = -0.95) across 84 traits.
• Individuals with greater genetic distance from training data exhibit significantly lower PGS accuracy.
• Hispanic/Latino individuals closest in GD to training data showed similar PGS performance to distant European individuals.

Conclusions:

• The continuum of genetic ancestries, measured by GD, is a critical factor in PGS portability and accuracy.
• Discrete genetic ancestry clusters are insufficient for accurately assessing and interpreting PGS performance.
• Future PGS development and application must account for the continuous nature of genetic variation for equitable use.