Julie-Alexia Dias (Harvard University)
will speak on
Pooled vs. meta-analytic models in multi-ancestry genome-wide association studies: Power, population structure, and prac
Time: 3:00PM
Date: Thu 15th January 2026
Location: E0.32 (beside Pi restaurant)
[map]
Abstract: Multi-ancestry biobanks now make it possible to conduct genome-wide association studies (GWAS) across diverse populations, but there remains no consensus on the most statistically powerful and robust analytical strategy. Two dominant approaches are widely used: pooled analysis, which jointly models individuals from all ancestries with population-structure adjustment (typically via principal components), and meta-analysis, which fits ancestry-specific GWAS followed by summary-statistic aggregation. Each approach has advantages and potential pitfalls arising from sample size imbalance, allele frequency differences, and the complexity of population structure.
In this talk, I will compare these strategies using large-scale simulations spanning a range of ancestry compositions and sample sizes, as well as empirical analyses of eight continuous and five binary traits from the UK Biobank (N≈324,000) and the All of Us Research Program (N≈207,000). Our findings show that pooled analysis consistently achieves higher statistical power while adequately controlling for population stratification. I will also introduce a theoretical framework that links the observed power differences to cross-population allele frequency variation. Together, these results—replicated across both biobanks—demonstrate that pooled analysis offers a robust and scalable approach for multi-ancestry GWAS, improving discovery while maintaining rigorous control of confounding.
If time permits, I will present my new research in developing a survival PRS (polygenic risk score) method, leveraging time-to-event data.
(This talk is part of the Statistics and Actuarial Science series.)
PDF notice
Return to all seminars