J. Přibyl

Contribution of domestic production records, Interbull estimated breeding values, and single nucleotide polymorphism genetic markers to the single-step genomic evaluation of milk production

Estimated breeding values (EBV) for first-lactation milk production of Holstein cattle in the Czech Republic were calculated using a conventional animal model and by single-step prediction of the genomic enhanced breeding value. Two overlapping data sets of milk production data were evaluated: (1) calving years 1991 to 2006, with 861,429 lactations and 1,918,901 animals in the pedigree and (2) calving years 1991 to 2010, with 1,097,319 lactations and 1,906,576 animals in the pedigree. Global Interbull (Uppsala, Sweden) deregressed proofs of 114,189 bulls were used in the analyses. Reliabilities of Interbull values were equivalent to an average of 8.53 effective records, which were used in a weighted analysis. A total of 1,341 bulls were genotyped using the Illumina BovineSNP50 BeadChip V2 (Illumina Inc., San Diego, CA). Among the genotyped bulls were 332 young bulls with no daughters in the first data set but more than 50 daughters (88.41, on average) with performance records in the second data set. For young bulls, correlations of EBV and genomic enhanced breeding value before and after progeny testing, corresponding average expected reliabilities, and effective daughter contributions (EDC) were calculated. The reliability of prediction pedigree EBV of young bulls was 0.41, corresponding to EDC=10.6. Including Interbull deregressed proofs improved the reliability of prediction by EDC=13.4 and including genotyping improved prediction reliability by EDC=6.2. Total average expected reliability of prediction reached 0.67, corresponding to EDC=30.2. The combination of domestic and Interbull sources for both genotyped and nongenotyped animals is valuable for improving the accuracy of genetic prediction in small populations of dairy cattle.

Short communication: Reliability of single-step genomic BLUP breeding values by multi-trait test-day model analysis

The purpose of our study was to develop an approximation procedure to estimate reliabilities of single-step genomic BLUP breeding values in a test-day model for routine evaluation of milk yield in a dairy cattle population. Input data consisted of 20,220,047 first-, second-, and third-lactation test-day milk yield records of 1,126,102 Czech Holstein cows (each lactation being considered a separate trait), with 1,844,679 animals in the pedigree file and with genomic data from 2,236 bulls. Evaluation was according to a multi-lactation model. The procedure was based on the effective number of records per animal from milk recording as well as from genomic and pedigree relationships. Traits were analyzed individually, and genetic covariances among traits were subsequently taken into account. The use of genomic information increased average reliability in young bulls from 0.276 to 0.505, but increased reliability in proven bulls only from 0.828 to 0.855. The reliabilities of genomic breeding values in multi-trait evaluation for first, second and third lactations, respectively, averaged 0.652, 0.673, and 0.633 for young bulls and 0.907, 0.894, and 0.852 for proven bulls. For an index combining all 3 lactations, the average reliability of a single-step genomic BLUP prediction was 0.712 and 0.925 for younger and proven bulls, respectively. Increased reliability due to genotyping in the population of all genotyped and nongenotyped animals was very small (<0.01) because of the small proportion of genotyped animals in the population.

Contribution of domestic and Interbull records to reliabilities of single-step genomic breeding values in dairy cattle

The method of approximating reliabilities of genomic breeding values in the single-step genomic BLUP evaluation procedure of Misztal et al. (2013) was used to evaluate the increase in reliability of breeding values for milk production in dairy cattle brought about by the inclusion of genomic data. Three strategies for approximation of reliabilities were compared: using only domestic records from performance testing of cows in the Czech Holstein dairy cattle population, using the same records in combination with Interbull breeding values of sires expressed as deregressed proofs, and using only the Interbull breeding values of sires expressed as deregressed proofs. The highest average reliability of genomic breeding values was achieved by the strategy using both domestic and Interbull data, for which the approximated reliabilities of genotyped bulls increased by 0.063. This general increase in reliability of genomic breeding values was small due to the small number of reference bulls available for the study. The overall increase in reliabilities for the entire population of dairy cattle was low but detectable. That modest increase was partially dependent on the unfavourable ratio of the number of genotyped bulls to the size of the analyzed population. Inclusion of Interbull data dramatically increased the benefits of genotyping in our test case - a relatively small population with substantial genetic contributions of foreign genes.

Preliminary analysis of utilization of genomic relationship in mating plan of Old Kladruber horse

The study analyzed 48 Old Kladruber horses genotyped by Illumina Equine SNP70 BeadChip for usefulness of genomic data in determining of mating plan. Totally 12 variants of data filtering and their impact on calculations in dependence of different parameters of GenCall Score, Minor Allele Frequency and assumed average values of loci of ancestors (l) was investigated. For possibility of comparison between genomic and commonly evaluated relationships, pedigree based relationship matrix was constructed and subsequently subtraction of pedigree from genomic matrix was performed. All matrices were thoroughly inspected and most suitable setting of parameters was chosen. Evaluation of genomic relationships can be successfully implemented in more precise method of mating plan design of Old Kladruber horses. Further genotyping and development of method for rescaling of differences between genomic and pedigree relationship matrices´ elements is advised for a purpose of better interpretation of results by breeders.

Imputation of missing genetic markers SNP using linear regression models

For imputation of missing SNP are used softwares which require known relationship between genotyped individuals. In common breeding business the genotypes of parents are not always known. That is why our own methodological process was used. The aim of this study is to map the current research of genetic chips and to verify the calculation process. The testing was processed at chosen loci in two datasets and in 8 models with different amount of SNPs. For the dataset A was prediction of missing values almost accurate with model reliability 100% with the exception of one homozygous locus where the reliability reached only 55%. In the dataset B the most extensive model reached the reliability of 80 - 90% even in case of homozygous loci. The prediction error value was higher than in the first case. It was proven that missing values prediction is possible to calculate using the neighbouring SNPs.

Methods for calculating the breeding values of missing traits and their comparison

The objective of this study was to calculate the breeding values (BVs) of traits missing in a selection index. Different traits can be evaluated within the breeding programs of given countries. The BV of a trait can be calculated based on genetic correlations with other traits. Similarly, the BV of a missing trait can be calculated for imported bulls. Two methods of calculation were used. Method A was based on a regression of BVs. Method B was based on performing a de-regression of BVs and their retroactive calculation. Both of these methods were tested using a Czech and a Canadian database of BVs for Holstein bulls. The Czech database of Holstein bulls contained 766 bulls and the Canadian database 851. Two calculations were performed for bulls with low reliability of estimated BVs, the first calculation with their genetic correlation matrix and the second with a genetic correlation matrix created from a set of bulls with high reliability of BVs. These newly calculated BVs (CBVs) were then compared with the national BVs (NBVs) using correlation coefficients. The highest correlations were achieved with high reliability bulls when all traits were included into the calculation (34 evaluated traits). The correlations of these bulls averaged 0.82, with an average standard deviation of 0.19. The lowest correlations were found when low reliability bulls were included and the genetic correlation matrix from the high reliability bulls was applied. That average correlation was 0.74 and standard deviation 0.25. When only 15 traits were evaluated in the model, the average correlation for all sets was 0.68 with standard deviation of 0.28. These results show that calculating the BV of a missing trait is possible using both methods. Method B was slightly more accurate in its prediction.