Don't BLUP Twice

Large, complex data sets can be difficult to model in a single comprehensive genome-wide association study (GWAS). The best practice for two-stage analyses is to consider lines as fixed effects in the first stage statistical model. Best linear unbiased estimates of lines can then be used as input phenotypes to the second stage analysis. In the second stage, lines can be modeled as random effects with genomic relationships to adjust for population structure when estimating individual SNP effects in GWAS.