Genomic selection for dry matter intake using a combined European and Australian reference population

Dairy cow dry matter intake (DMI) data from Australia (AU), the United Kingdom (UK) and the Netherlands (NL) were combined (1801 cows) for this study. The aim was to explore the impact on the accuracy of genomic estimated breeding values of pooling data across key reference populations. A total of 843 Australian growing heifers with records on DMI measured over 60 to 70 d at approximately 200 d of age, 359 Scottish and 599 Dutch lactating heifers with records on DMI during the first 100 d in milk were included in the data set. Genotypes were obtained using the Illumina BovineSNP50 BeadChip for European (UK+NL) cows, and Illumina High Density Bovine SNP chip for AU heifers. The AU and EU genomic data were matched on SNP-name and genotypes were compared for quality control using 40 bulls that were genotyped in both data sets. This resulted in a total of 30,949 SNPs being used in the analyses. Genomic predictions were with both single-trait and multi-trait genomic REML models, using ASReml. The accuracy of genomic prediction was evaluated in 11 single-country validation sets, and the reference set (where animals had both DMI phenotypes and genotypes) were either a reference set within AU or EU, or with a multi-country reference set consisting of all data except the validation set. When DMI was considered to be the same trait for each country, using a multi-country reference set, the accuracy of genomic prediction for DMI increased for EU and UK, but not for AU and NL. Extending to a bivariate (AU-EU) or trivariate (AU-UK-NL) model increased the accuracy of genomic prediction for DMI in all countries. The highest accuracies were estimated for all countries when data was analyzed with a trivariate model, with increases of up to 5.5% compared with a single-trait analysis with a multi-country reference set