GWAS And Genomic Prediction Based On Markers Of SNP-Chips And Sequence Data In Dairy Cattle Populations

Genome wide association study (GWAS) and genomic selection are important approaches for genetic improvement of livestock. The objective of this PhD project was to investigate the methods and models for GWAS and genomic prediction using markers of different densities. The project includes the following four phases.In paper I, GWAS was carried out to identify single nucleotide polymorphisms (SNPs) that are associated with body conformation traits, using marker data of the Illumina BovineSNP50BeadChip (54K markers). A least absolute shrinkage and selection operator (LASSO) was applied to detect multiple SNPs simultaneously for29body conformation traits using data of1,314Chinese Holsteins. Totally,59genome-wide significant SNPs associated with26conformation traits were detected. By combining information of previously reported QTL regions and annotated biological functional genes, we identified DARC, GAS1, MTPN, HTR2A, ZNF521, PDIA6, and TMEM130as the most promising candidate genes for capacity and body depth, chest width, foot angle, angularity, rear leg side view, teat length, and animal size traits, respectively. We also found four SNPs that affected four pairs of traits, and the genetic correlation between each pair of traits ranged from0.35to0.86, suggesting that these SNPs may have a pleiotropic effect on each pair of traits.In paper II, we investigated the performance of three marker densities for refining six previously detected QTL regions for mastitis traits:54K. markers of a medium density SNP chip (MD),770K. markers of a high density SNP chip (HD), and whole genome sequencing data (SEQ) of4,496Danish Holsteins. GWAS was performed using the linear mixed model (LM) and the Bayesian variable selection model (BVS). After quality control, there were587,7,825, and78,856SNPs in the six regions available in MD, HD, and SEQ, respectively. In general, the association patterns between SNPs and traits based on the three marker densities were similar when tested using the same statistical model. With the LM, a total of120(MD),967(HD), and7,209(SEQ) SNPs exhibited significant association with mastitis. In addition,1,16, and33QTL peaks for MD, HD, and SEQ data were detected by using the QTL intensity profile. According the positions of the QTL peaks, NPFFR2, SLC4A4, DCK, LIFR, and EDN3were suggested as candidate genes associated with mastitis.In paper III, we investigated reliability of genomic prediction in various scenarios with regard to relationship between test and training animals and between animals within the training data set. Different training data sets were generated from EuroGenomics data and a group of Nordic Holstein bulls (born in2005and afterward) as a common test data set. Genomic breeding values were predicted using a GBLUP model and a Bayesian mixture model. Result showed that a closer relationship between test and training animals led to a higher reliability of genomic predictions for the test animals, while a closer relationship between training animal resulted in a lower reliability. In addition, the Bayesian mixture model in general led to a slightly higher reliability of genomic prediction especially for the scenario of distant relationship between training and test animals.In paper IV, we investigate the efficiency of combining GWAS information into GBLUP model on accuracy of genomic prediction, based on data of Nordic Holstein and Red populations. The analysis was carried out by several models were compared on reliability of genomic prediction. Results show that integrating GWAS information by weighting G-matrix doesn’t show advantages in our research. However, a model includes QTL markers and the other markers as two different components performed slightly better than the model without considering GWAS information. In addition, joint reference population led to slight increase of prediction accuracy, compared with single-breed reference population.