Meta-analysis

We conducted genome-wide association studies (GWAS) on cattle flight time for three populations and conducted a meta-analysis of the results for a combined sample of 9223 animals and 28.4 million imputed biallelic variants, including both SNPs and small insertions and deletions (INDEL). The number of animals and variants varied slightly between cohorts, ranging from 2112 to 4586 animals and 24.8 to 28.2 million variants. Within each cohort, the model was fitted as:

where y is a vector of phenotypes (standardized natural logarithm of flight time), β is a vector of fixed effects including the genotype for the candidate SNP and all covariates (mean, log(age), and contemporary group (year, stud, and sex)), g is a vector of total additive genetic effects with g∼N0,Gσg2, where G is the genomic relationship matrix (GRM) generated from imputed sequence variants, σg2 is the additive genetic variance, and ε is a vector of random residuals ε∼N(0,Iσε2). X and Z are design matrices for the fixed and random effects, respectively. The GRM was constructed following [46]. Note that this is an additive model for the SNP that assumes that the effect of having two copies of the non-reference allele is twice the effect of having one only.

GWAS were performed in GCTA [47]. Results for all cohorts were combined using a fixed effects inverse-variance weighted meta-analysis as implemented in METAL [48]. We used a significance threshold of P < 5×10⁻⁸, which corresponded to a false discovery rate (FDR) of 0.01. Figures throughout the article were generated using R [49]. The meta-analysis lead variants were identified using clumping in PLINK1.9 [45] with parameters: P <5 ×10⁻⁸, 5-Mb windows, and r² = 0.1.