When conducting GWAS meta-analysis with multiple small cohorts (<2,000 samples), what strategies do you use to address issues such as limited statistical power, population stratification, phenotype harmonization, and result interpretation?
Most of this is covered in the day 4 lectures, so Iād encourage you to revisit those first (see: Genome wide association studies | Institute for Behavioral Genetics | University of Colorado Boulder). But briefly: with small cohorts, the statistical power is limited, and meta-analysis is the main solution. Pooling multiple small cohorts is the scenario GWAS meta-analysis was designed for. Population stratification is handled through PCs and (especially if relatedness is an issue) a GRM, as covered in the practical. Phenotype harmonization can be the hardest part and has no perfect solution; the best practice is agreeing on definitions upfront before cohorts run their GWAS. For result interpretation with modest N, try to be conservative and focus on whether your findings replicate rather than treating nominal hits as discoveries.
1 Like
Very insightful answer thank you @Abdel_Abdellaoui .