| The overall objective of this study was to evaluate milk release parameters in dairy cattle. This was accomplished by 1) estimating variance components for all milk flow traits of interest including production traits that are related to milk release; 2) identifying genomic regions harboring QTL associated with milk flow traits; and 3) evaluating the opportunity for implementation of genomic selection within Italian Brown Swiss for milk flow traits. Data were available from 37,213 Italian Brown Swiss cows over a span of 12 years (1997--2008) from 1661 herds. Milking flows were recorded for each individual once during lactation. Sires (n = 1351) of cows with milk flow information were genotyped for 33,074 markers distributed across 29 bos taurus autosomes (BTA). Heritabilities for milkability traits ranged from 0.02 to 0.42 with genetic correlations between production traits and flow traits ranging from low to moderate values. Positive genetic correlations were found among production, somatic cell score (SCS) and milkability traits. A Bayesian LASSO analysis was employed for each of the milk traits resulting in 6,929 to 14,585 significant single nucleotide polymorphisms (SNPs) marker effects identified for each trait across all 29 bos Taurus autosomes. Unique significant marker effects were found for each of the six traits providing evidence that each individual milk flow trait offers distinct genetic information about milk flow. Furthermore a number of quantitative trait loci associated with milk yield, somatic cell count, somatic cell score, milking speed and udder morphometrics were co-located with markers identified. Genotyped sires were partitioned based on their estimated breeding values (EBV) reliabilities into a training set (EBV reliability > 0.60) and a validation set (EBV reliability < 0.60). Pseudo-phenotypes for these sires were obtained by EBV de-regression and subsequently employed in obtaining genomic breeding values through GBLUP, Student-t, and Bayesian LASSO models. Genomic breeding values were also obtained using single-step methods combining genomic and pedigree based information using dense (HBLUP) and reduced (HrBLUP) SNP panels. Reliabilities estimated from breeding values incorporating genomic information from bulls in the validation set were compared with parental averages (PA) and breeding values obtained using traditional BLUP methods (PBLUP). Breeding values obtained using GBLUP resulted in reliabilities within the prediction set of 0.34 (0.29 PA), 0.32 (0.26 PA), 0.23 (0.30 PA), 0.35 (0.26 PA), 0.43 (0.30 PA), and 0.45 (0.29 PA) for TMT, AT, TP, DT, AVGF and MMF, respectively. Student-t and Bayesian LASSO models obtained similar reliabilities that were slightly better than GBLUP; 0.45, 0.32, 0.24, 0.47, 0.47, and 0.46 for TMT, AT, TP, DT, AVGF and MMF, respectively. The largest increase in reliability was accomplished by combining genomic and pedigree information in a single step (HBLUP and HrBLUP). Average gain in reliability from HBLUP from PA was 0.20, 0.24, 0.20, 0.17, 0.20 and 0.21 while gains from HrBLUP were slightly more than HBLUP. For all milk flow traits, use of a genomic model resulted in increased accuracy of prediction for sires with little or no progeny information. Inclusion of genomic information increased accuracy of prediction for milk flow measures providing evidence that implementation of genomic selection would likely increase response to selection. |
