Methods To Improve Genomic Prediction And GWAS Using Combined Holstein Populations

The economically important traits in livestock are usually complex traits influenced by genes, environment and genotype by environment (GxE) interaction. A situation in which GxE interaction occurs but is ignored could lead to biased estimates of breeding values and suboptimal selection decisions. Meanwhile, genome-wide association studies (GWAS) and genomic predictions have been widely applied in the genetic improvement of livestock. Therefore, the objectives of this PhD project are to perform GWAS and genomic prediction in multiple populations from different countries, and to investigate interaction between genotypes and production environments in different countries.In paper Ⅰ, we investigated the improvement of prediction reliabilities for three traits (milk yield, fat yield and protein yield) of Brazilian Holsteins by adding Nordic and French Holsteins. Here, performances of the same trait in Brazilian and Nordic (Nordic and French) populations were treated as two different but genetically correlated traits. Four methods, single-trait pedigree BLUP (pBLUP), single-trait single step genomic BLUP (ssGBLUP), two-trait pBLUP and two-trait ssGBLUP, were used. We found that the estimated across country genetic correlations (ranging from 0.579 to 0.713) indicated that there was an important GxE interaction between Brazilian and Nordic (or Nordic and French) populations. The prediction reliabilities for Brazilian genotyped bulls increased greatly and two-trait ssGBLUP performed much better than the corresponding pBLUP. In contrast, a minor improvement in prediction reliabilities was observed for Brazilian non-genotyped cows as two-trait ssGBLUP performed similar to the corresponding pBLUP.In paper Ⅱ, a joint GWAS was carried out for 16 milk fatty acid traits based on data of 784 Chinese and 371 Danish Holstein cows. According to the analysis of either the Chinese or the Danish data separately, the total numbers of overlapping SNP that were significant at the chromosome level were 94 for Cl4:1, and 208 for the C14 index. Joint analysis using the combined data of the two populations detected greater numbers of significant SNP compared to single population alone for 7 and 10 traits at the genome- and chromosome-wide significance levels, respectively. However, greater numbers of significant SNP were detected for C18:0 and the C18 index in the analysis using Chinese population alone, compared to the joint analysis. Ten fatty acid traits were influenced by a quantitative trait loci (QTL) region including DGAT1 in the combined population. A large region (14.9-24.9 Mbp) in BTA26 significantly influenced C14:1 and the C14 index in both populations, mostly likely due to SCD1. A QTL region (69.97-73.69 Mbp) on BTA9 showed a significantly different effect on C 18:0 between the two populations. Detection of these important SNP and the corresponding QTL regions will be helpful for follow-up studies to identify casual mutations and their interaction with environments for milk fatty acid traits in dairy cattle.In paper Ⅲ, we performed an investigation on the patterns of genomic variances, covariances and correlations across genome between Chinese and Nordic Holstein populations for three milk production traits at three different levels of genome region (all SNP as one region, each chromosome as one region and every 100 SNP as one region). The results showed that BTA 14 and BTA 5 accounted for very large genomic variance and covariance for milk yield and fat yield, whereas no specific chromosome showed very large genomic variance and covariance for protein yield. In the scenario of every 100 SNP as one region, most regions explained <0.50% of genomic variance and covariance for milk yield and fat yield, and most regions explained <0.30% for protein yield. Although overall correlations between two populations for the three traits were high, a few of regions showed high negative genomic correlations for milk yield and fat yield. Those estimated genomic parameters could be useful to improve the genomic prediction accuracy for Chinese and Nordic Holstein populations using a joint reference data in the future.