GENOMIC ASPECTS OF GENETIC IMPROVEMENT FOR MASTITIS RESISTANCE IN DAIRY CATTLE

The overall objective of this work is to evaluate the mastitis resistances genetic aspects in cattle with genetic and genomic tools. This was accomplished by: 1) the estimation of genetic parameters for traits related to udder health in the Valdostana cattle breed; 2) the evaluation of the effect on the genomic breeding value estimation of mastitis traits of the assumption of different prior probability values for the proportion of markers with a large effect; 3) the estimation of the fraction of the genetic variance not explained by the 54K Illumina SNP chip while using different marker-based relationship matrices; and 4) the exploration of the influence of the level of phenotype accuracy on the genomic predictions, by including phenotypes with different minimum level of reliability in the training population to estimate the marker effects. For the objective one, data of the Valdostana cattle breed were used for a total of 34,291 milking cows, collected born from 2002, with the milk bacteriological analysis that reported the presence of Staphylococcus aureus, Streptococcus agalactie, Staphylococcus ssp., Streptococcus ssp., Escherichia coli, minor pathogens gram positive, minor pathogens gram negative and fungi. Data used also included information on SCS and milk yield (MY) and genealogical data were extracted from the national herd book. The threshold model analysis was evaluated to be the most appropriate approach for the binary data under analysis concerning the presence/absence of pathogens in milk. Generally moderate heritability values (from 0.02 to 0.09) were estimated for the specific presence of the pathogens. This suggested that bacteriological data can be considered for the genetic selection to improve udder health. Promising results in mastitis selection were predicted through the aggregation of the indirect indicator (SCS), the trait commonly worldwide used in dairy cattle mastitis selection, and innate resistance to some of the major pathogens causing mastitis, such as Staphylococcus aureus, Streptococcus agalactie and Escherichia coli. The Bovine SNP50 Illumina genotypes of 1,089 Brown Swiss bulls were used for the second purpose. A total of 51,582 SNP markers were considered in the analyses after the exclusion of markers on chromosome X. The estimated breeding values of bulls were provided by the Italian Brown Cattle Breeders Association for SCS and for the following production traits: milk yield, fat yield, protein yield, fat percentage, and protein percentage. To perform genomic breeding value estimation, data available were split in two populations subsets: training (the 846 bulls born before 2001) and test population (243 bulls born from 2001 and 2005). The value assumed (0.001, 0.005, 0.01, 0.05, 0.1 and 0.5) for the number of SNP with a large effect did not have impact on the marker effect estimates or on genomic predictions accuracies. For the objective three, genotypes from Bovine SNP50 chip of 1,086 sires were available for a total of 35,706 SNP with a 99.34% total genotyping rate. Phenotypic information consisted on EBV estimated by the National Breeder Association of Brown Swiss dairy cattle for SCS, production and type traits as follows: milk yield, fat yield, protein yield, SCS, overall conformation, stature, and rear leg side view, fore udder attachment, rear udder width, udder support, udder depth, feet and legs and foot height. Three generations of genealogical information were used (4,988 animals). The proportion of the genetic variance addressed by markers was estimated using different marker-based relationship matrices. In all traits considered the fraction of the genetic variance not explained by the genetic markers did not significantly differ from 0 for all the designs including in the training population bulls with high accuracy. Indeed no substantial differences were found with the use of different genomic relationship matrices. The only exception was the genomic matrix corrected by the heterozygosity per SNP. All the analysis with that genomic matrix converged and the genetic variance explained was bigger than with the other matrices. For the objective 4; the phenotype was associated with the genotype at 35,546 SNP markers for 1357 sires for all available traits and genomic EBV estimated using different strategy to identify the training population. Results indicate that selecting the training population with accurate phenotypes yield genomic EBV with larger accuracy. The genetic selection of a complex trait, as mastitis selection, involved several aspects and it would be better searched through the integration of classical genetic aspects and of the genomic selection. The latter genomic approach would benefit of the use of all the aspect considered in this thesis. In particular suggestions were drawn for the use of best prior parameters, the definition of the most informative training population, the calculation of genetic parameters for binary and non-binary traits, and the calculation of genomic relationship variances to be used in genomic breeding value estimations.