Abstract
Genomic selection (GS) is presently revolutionizing dairy cattle breeding. GS, however, is efficient in populations with large reference populations of genotyped and progeny tested bulls to train the genomic prediction models, but less efficient in smaller populations. Two main approaches are currently being tested to improve prediction accuracies in numerically small breeds. The first is to develop across breed predictions. Because the linkage disequilibrium (LD) across distantly related breeds is not strong enough to simply use large scale 50K genotype information, this approach is strongly linked to use of high density genotypes and whole genome sequence (WGS) information. The second approach is to genotype vast number of cows with phenotypes to enlarge the reference population.
In chapter 2.1 and 2.2, we took advantage of both approaches by using genotype information on many cows in five breeds. These cows were genotyped with customized chips including an increasing number of confirmed or candidate causative mutations (potential QTN), discovered in large scale genome-wide association studies (GWAS) combining genotyping and whole genome sequencing (WGS). The large-scale genotyping of potential QTN across multiple breeds provided a powerful approach to prioritize between potential QTN, thus providing evidence on which markers explain genetic variance across breeds and therefore can be used to improve predictions. In chapter 2.3, we identified several novel QTL that can be used to select for reduced milking time without causing deterioration in important health traits such as clinical mastitis and somatic cell score.
In contrast with Daughter Yield Deviations, which primarily reflect the additive value of the bulls, a cow’s performance is the result of additive and non-additive effects. In chapter 2.4 we studied additive effects using the Associated Weight Matrix (AWM), in combination with Partial Correlation Coefficient with Information Theory (PCIT), a network inference algorithm, to generate gene networks with regulatory and functional significance for udder related phenotypes. By exploiting correlated udder phenotypes, we increased accuracy of statistical inference and identified ten genes that directly affect mammary gland development. In chapter 2.5, we pre-selected SNP with major effect and using a Bayesian approach, estimated dominance and epistasis as explained by second order gene interactions. Dominance variance was consistent across breeds and traits and represented about 20% of the additive genetic variance. The epistatic variance estimates were more variable, from nearly zero to 19% of the additive genetic variance, with an average of 7%. This provides original research on use of sequence data to study epistasis in dairy cattle.
In chapter 2.1 and 2.2, we took advantage of both approaches by using genotype information on many cows in five breeds. These cows were genotyped with customized chips including an increasing number of confirmed or candidate causative mutations (potential QTN), discovered in large scale genome-wide association studies (GWAS) combining genotyping and whole genome sequencing (WGS). The large-scale genotyping of potential QTN across multiple breeds provided a powerful approach to prioritize between potential QTN, thus providing evidence on which markers explain genetic variance across breeds and therefore can be used to improve predictions. In chapter 2.3, we identified several novel QTL that can be used to select for reduced milking time without causing deterioration in important health traits such as clinical mastitis and somatic cell score.
In contrast with Daughter Yield Deviations, which primarily reflect the additive value of the bulls, a cow’s performance is the result of additive and non-additive effects. In chapter 2.4 we studied additive effects using the Associated Weight Matrix (AWM), in combination with Partial Correlation Coefficient with Information Theory (PCIT), a network inference algorithm, to generate gene networks with regulatory and functional significance for udder related phenotypes. By exploiting correlated udder phenotypes, we increased accuracy of statistical inference and identified ten genes that directly affect mammary gland development. In chapter 2.5, we pre-selected SNP with major effect and using a Bayesian approach, estimated dominance and epistasis as explained by second order gene interactions. Dominance variance was consistent across breeds and traits and represented about 20% of the additive genetic variance. The epistatic variance estimates were more variable, from nearly zero to 19% of the additive genetic variance, with an average of 7%. This provides original research on use of sequence data to study epistasis in dairy cattle.
Originalsprog | Engelsk |
---|
Antal sider | 161 |
---|---|
ISBN (Trykt) | 978-87-93643-74-1 |
Status | Udgivet - 2018 |