Multi-population genome-wide association studies involving four distinct barley populations

The power of genome-wide association studies (GWAS) relies heavily on the sample size. A strategy to increase sample size is to combine datasets from different populations. However, this approach introduces challenges due to heterogeneity between populations. With this data, we set up a statistically sound model to account for such heterogeneity. Using this model, we combined up to four distinct barley populations in GWAS to detect genomic regions associated with heading date and stem lodging. Each population represented an applied breeding program with unique combinations of growth habit (winter versus spring) and row type (2-rowed versus 6-rowed). 

By comparing single-population GWAS with multi-population GWAS, we identified both quantitative trait loci (QTLs) that were shared across populations and population-specific QTLs. We found that multi-population GWAS provided greater statistical power than single-population analyses, revealed QTLs that were undetectable in small populations, and explained an overall larger proportion of the phenotypic variance. 

Our findings offer a promising approach to accelerate genomics-based breeding in new breeding populations with limited data. This methodology is applicable to a wide range of datasets where sample sizes are limited for various reasons.