Genetic disease risks can be misestimated across global populations

Area of Science:

• Genomics
• Population Genetics
• Health Disparities Research

Background:

• Accurate health disparity assessment necessitates unbiased genetic risk knowledge across diverse populations.
• Current genome-wide association studies (GWAS) predominantly utilize European samples and genotyping arrays, potentially introducing bias.
• Integrating global whole genome sequence data is crucial for understanding population-specific genetic disease risks.

Purpose of the Study:

• To investigate how genetic disease risks are misestimated in different global populations.
• To identify biases in current genome-wide association studies (GWAS) methodologies.
• To evaluate the impact of ancestral versus derived risk alleles on disease risk prediction.

Main Methods:

• Integration of whole genome sequence data from global populations.
• Analysis of results from thousands of genome-wide association studies (GWAS).
• Extensive computer simulations to model GWAS with varying study populations and ascertainment biases.

Main Results:

• Significant differences in risk allele frequencies at known disease loci were observed between African and other continental populations.
• Ancestral risk alleles are more frequent in Africa (9.51%), while derived risk alleles are less frequent (5.40%).
• Simulations revealed that non-African cohorts in GWAS lead to biased allele frequencies, whereas African cohorts provide relatively unbiased associations. Genotyping arrays and SNP ascertainment bias contribute to these continental differences, leading to misestimated polygenic risk scores for individuals of African descent.

Conclusions:

• Extrapolation of GWAS results across different populations requires significant caution due to inherent biases.
• Continental differences in allele frequencies persist even with whole genome sequencing and large sample sizes.
• Considering ancestral versus derived alleles is essential for improving the accuracy of genetic risk scores.