Genomic selection is a new technology in which selection decisions are based on direct genomic values (DGVs) or genomic enhanced breeding values (GEBVs). The objective of this study was to evaluate the relations between DGVs and several milk traits important for both the nutritional value and processability of milk. This is a new approach and can be used to increase the knowledge on how genomic selection can be used in practice. Morning milk samples from Swedish Holstein cows were analyzed for milk composition and technological properties. DGVs were received for each cow for milk, protein and fat yield, milk index, udder health, Nordic total merit and a quota was calculated between fat and milk yield as well as protein and milk yield. The results show that linear correlations exist (P<0·10) between the studied DGVs and contents and yields of parameters in the protein (P=0·002–0·097), fat (P=0·024–0·055) and mineral profiles (P=0·001–0·099) as well as for cheese characteristics (P=0·004–0·065), thus making it possible to obtain detailed information on milk traits that are not registered in the milk recording scheme. Hence, genomic selection will be an efficient tool for breeding and dairy industry to select cows early in life for targeted milk production.

In order to optimize the use of genomic selection in breeding plans, it is essential to have reliable estimates of the genomic breeding values. This study investigated reliabilities of direct genomic values (DGVs) in the Jersey population estimated by three different methods. The validation methods were (i) fivefold cross-validation and (ii) validation on the most recent 3 years of bulls. The reliability of DGV was assessed using squared correlations between DGV and deregressed proofs (DRPs). In the recent 3-year validation model, estimated reliabilities were also used to assess the reliabilities of DGV. The data set consisted of 1003 Danish Jersey bulls with conventional estimated breeding values (EBVs) for 14 different traits included in the Nordic selection index. The bulls were genotyped for Single-nucleotide polymorphism (SNP) markers using the Illumina 54 K chip. A Bayesian method was used to estimate the SNP marker effects. The corrected squared correlations between DGV and DRP were on average across all traits 0.04 higher than the squared correlation between DRP and the pedigree index. This shows that there is a gain in accuracy due to incorporation of marker information compared with parent index pre-selection only. Averaged across traits, the estimates of reliability of DGVs ranged from 0.20 for validation on the most recent 3 years of bulls and up to 0.42 for expected reliabilities. Reliabilities from the cross-validation were on average 0.24. For the individual traits, the reliability varied from 0.12 (direct birth) to 0.39 (milk). Bulls whose sires were included in the reference group had an average reliability of 0.25, whereas the bulls whose sires were not included in the reference group had an average reliability that was 0.05 lower.