Polygenic risk scores based on European GWAS correlate to disease prevalence differences around the world

Pritesh Jain, Myson Burch, Melanie Martinez, Pablo Mir, Jakub Fichna, Cezary Żekanowski, Renata Rizzo, Zeynep Tümer, Csaba Barta, Evangelia Yannaki, J Stamatoyannopoulos, Petros Drineas, Peristera PaschoumedRxiv 2021 · DOI

Abstract

The vast majority of Genome-Wide Association Studies (GWAS) have focused on populations of European descent, raising significant concerns about the transferability of polygenic risk scores (PRS) to non-European groups. This study investigates whether European-derived PRS can accurately reflect disease prevalence across 24 worldwide populations. By analyzing 20 complex disorders, researchers found that for 13 conditions—including obesity and type 2 diabetes—genetic risk scores significantly correlate with actual global prevalence. These findings suggest that many disease-associated loci are evolutionarily conserved across human ancestries, providing a degree of cross-population utility for current genetic risk models while underscoring the ongoing need for ancestral diversity in genomic research.

Introduction: The Genomic Diversity Gap

In contemporary genomics, a "diversity gap" persists: approximately 70% of GWAS participants are of European ancestry. This imbalance poses a challenge for the implementation of personalized medicine globally. Because different populations exhibit distinct linkage disequilibrium (LD) structures and allele frequencies, it has been hypothesized that risk scores optimized for European populations might fail to predict disease liability in other ancestral groups. This study addresses this "transferability" problem by testing the correlation between European-based PRS and real-world epidemiological data across diverse global populations.

Methodology

The researchers employed a multi-step approach to evaluate the global applicability of European GWAS data:

  1. PRS Calculation: Polygenic Risk Scores were derived for 20 complex disorders using summary statistics from large-scale European GWAS.
  2. Population Modeling: These scores were applied to a target dataset of 3,953 individuals from the 1000 Genomes Project, representing 24 populations across five major ancestral groups: African, European, South Asian, East Asian, and Admixed American.
  3. Epidemiological Correlation: Average genetic risk scores were correlated with disease prevalence data from the Global Burden of Disease (GBD) database and the World Health Organization (WHO).
  4. Conservation Analysis: To identify the biological basis for transferability, the researchers analyzed the genetic architecture ($F_{ST}$ and LD structure) of the loci used in the PRS compared to random genetic markers.

Key Insights: The Conserved Loci Hypothesis

A primary finding of the study is that the genetic regions associated with the 13 correlated disorders are significantly more "conserved" across populations than random genomic regions. This suggests that GWAS often identifies fundamental biological "choke points"—variants that play a consistent role in disease biology regardless of an individual's specific ancestral background.

Furthermore, genetic risk scores were found to follow geographic patterns. Populations within the same region tended to exhibit similar risk profiles, reinforcing the link between shared ancestry and genetic liability.

Regional Vulnerabilities and Results

The study identified specific genetic liabilities across different ancestral groups:

  • African Populations: Demonstrated 1.3 to 2.5 times higher PRS for obesity and Polycystic Ovary Syndrome (PCOS), but a lower risk for Type 1 Diabetes.
  • East Asian Populations: Carried the highest genetic risk for Type 2 Diabetes and Chronic Kidney Disease, while showing lower liability for autoimmune conditions like asthma and multiple sclerosis.
  • European Populations: Showed higher genetic liability for Alzheimer’s disease and several psychiatric disorders compared to non-European groups.

Quantitatively, the correlations were robust for 13 of the 20 disorders. Obesity (OBY) showed the strongest correlation ($R^2 = 0.73, p=0.001$), followed by Type 2 Diabetes ($R^2 = 0.55$) and Chronic Respiratory Disease ($R^2 = 0.60$). Psychiatric conditions, including ADHD ($R^2=0.46$) and Schizophrenia ($R^2=0.45$), also showed significant correlations.

Discussion and Limitations

These results provide a vital "sanity check" for GWAS, confirming that the variants identified are biologically relevant across diverse populations. However, the authors note several limitations:

  • Missing Variants: European-based GWAS may overlook variants that are rare in Europeans but common and highly impactful in other populations.
  • Reporting Bias: Lower prevalence rates for certain conditions (e.g., psychiatric disorders) in developing nations may reflect diagnostic disparities rather than lower biological risk.
  • The Need for Diversity: While the transferability of European PRS is higher than previously feared, it does not negate the necessity for trans-ancestral studies. Expanding GWAS to include diverse populations remains essential for improving the precision of risk estimates and ensuring equitable access to genomic medicine.

Conclusion

Polygenic risk scores derived from European GWAS can significantly capture differences in disease risk worldwide. By identifying populations with high genetic liability, these scores offer a valuable tool for global health research, even as the field moves toward more inclusive and representative genomic datasets.

Summary updated

Abstract

Abstract Background Complex disorders are caused by a combination of genetic, environmental and lifestyle factors, and their prevalence can vary greatly across different populations. Genome wide association studies (GWAS) can help identify common variants that underlie disease risk. However, despite their increasing number, the vast majority of studies focuses on European populations, leading to questions regarding the transferability of findings to non-Europeans. Here, we investigated whether polygenic risk scores (PRS) based on European GWAS correlate to disease prevalence within Europe and around the world. Results GWAS summary statistics of 20 different disorders were used to estimate PRS in nine European and 24 worldwide reference populations. We estimated the correlation between average genetic risk for each of the 20 disorders and their prevalence in Europe and around the world. A clear variation in genetic risk was observed based on ancestry and we identified populations that have a higher genetic liability for developing certain disorders both within European and global regions. We also found significant correlations between worldwide disease prevalence and PRS for 13 of the studied disorders with obesity genetic risk having the highest correlation to disease prevalence. For these 13 disorders we also found that the loci used in PRS are significantly more conserved across the different populations compared to randomly selected SNPs as revealed by Fst and linkage disequilibrium structure. Conclusion Our results show that PRS of world populations calculated based on European GWAS data can significantly capture differences in disease risk and identify populations with the highest genetic liability to develop various conditions. Our findings point to the potential transferability of European-based GWAS results to non-European populations and provide further support for the validity of GWAS.
BibTeX
@misc{jain2021polygenic,
  author = {Pritesh Jain and Myson Burch and Melanie Martinez and Pablo Mir and Jakub Fichna and Cezary Żekanowski and Renata Rizzo and Zeynep Tümer and Csaba Barta and Evangelia Yannaki and J Stamatoyannopoulos and Petros Drineas and Peristera Paschou},
  title = {{Polygenic risk scores based on European GWAS correlate to disease prevalence differences around the world}},
  year = {2021},
  doi = {10.1101/2021.11.13.21265898},
  howpublished = {medRxiv},
}
Figure