Comparison of genomic and traditional BLUP-estimated breeding value accuracy and selection response under alternative trait and genomic parameters

Accuracy of prediction of estimated breeding values based on genome-wide markers (GEBV) and selection based on GEBV as compared with traditional Best Linear Unbiased Prediction (BLUP) was examined for a number of alternatives, including low heritability, number of generations of training, marker density, initial distributions, and effective population size (Ne). Results show that the more the generations of data in which both genotypes and phenotypes were collected, termed training generations (TG), the better the accuracy and persistency of accuracy based on GEBV. GEBV excelled for traits of low heritability regardless of initial equilibrium conditions, as opposed to traditional marker-assisted selection, which is not useful for traits of low heritability. Effective population size is critical for populations starting in Hardy-Weinberg equilibrium but not for populations started from mutation-drift equilibrium. In comparison with traditional BLUP, GEBV can exceed the accuracy of BLUP provided enough TG are included. Unfortunately selection rapidly reduces the accuracy of GEBV. In all cases examined, classic BLUP selection exceeds what was possible for GEBV selection. Even still, GEBV could have an advantage over traditional BLUP in cases such as sex-limited traits, traits that are expensive to measure, or can only be measured on relatives. A combined approach, utilizing a mixed model with a second random effect to account for quantitative trait loci in linkage equilibrium (the polygenic effect) was suggested as a way to capitalize on both methodologies.     

Application of imputation methods to genomic selection in Chinese Holstein cattle

Missing genotypes are a common feature of high density SNP datasets obtained using SNP chip technology and this is likely to decrease the accuracy of genomic selection. This problem can be circumvented by imputing the missing genotypes with estimated genotypes. When implementing imputation, the criteria used for SNP data quality control and whether to perform imputation before or after data quality control need to consider. In this paper, we compared six strategies of imputation and quality control using different imputation methods, different quality control criteria and by changing the order of imputation and quality control, against a real dataset of milk production traits in Chinese Holstein cattle. The results demonstrated that, no matter what imputation method and quality control criteria were used, strategies with imputation before quality control performed better than strategies with imputation after quality control in terms of accuracy of genomic selection. The different imputation methods and quality control criteria did not significantly influence the accuracy of genomic selection. We concluded that performing imputation before quality control could increase the accuracy of genomic selection, especially when the rate of missing genotypes is high and the reference population is small.     

Dimension reduction and variable selection for genomic selection: application to predicting milk yield in Holsteins

Genome‐assisted prediction of genetic merit of individuals for a quantitative trait requires building statistical models that can handle data sets consisting of a massive number of markers and many fewer observations. Numerous regression models have been proposed in which marker effects are treated as random variables. Alternatively, multivariate dimension reduction techniques [such as principal component regression (PCR) and partial least‐squares regression (PLS)] model a small number of latent components which are linear combinations of original variables, thereby reducing dimensionality. Further, marker selection has drawn increasing attention in genomic selection. This study evaluated two dimension reduction methods, namely, supervised PCR and sparse PLS, for predicting genomic breeding values (BV) of dairy bulls for milk yield using single‐nucleotide polymorphisms (SNPs). These two methods perform variable selection in addition to reducing dimensionality. Supervised PCR preselects SNPs based on the strength of association of each SNP with the phenotype. Sparse PLS promotes sparsity by imposing some penalty on the coefficients of linear combinations of original SNP variables. Two types of supervised PCR (I and II) were examined. Method I was based on single‐SNP analyses, whereas method II was based on multiple‐SNP analyses. Supervised PCR II was clearly better than supervised PCR I in predictive ability when evaluated on SNP subsets of various sizes, and sparse PLS was in between. Supervised PCR II and sparse PLS attained similar predictive correlations when the size of the SNP subset was below 1000. Supervised PCR II with 300 and 500 SNPs achieved correlations of 0.54 and 0.59, respectively, corresponding to 80 and 87% of the correlation (0.68) obtained with all 32 518 SNPs in a PCR model. The predictive correlation of supervised PCR II reached a plateau of 0.68 when the number of SNPs increased to 3500. Our results demonstrate the potential of combining dimension reduction and variable selection for accurate and cost‐effective prediction of genomic BV.     

Prediction of response to marker-assisted and genomic selection using selection index theory

Selection index methods can be used for deterministic assessment of the potential benefit of including marker information in genetic improvement programmes using marker-assisted selection (MAS). By specifying estimates of breeding values derived from marker information (M-EBV) as a correlated trait with heritability equal to 1, it was demonstrated that marker information can be incorporated in standard software for selection index predictions of response and rates of inbreeding, which requires specifying phenotypic traits and their genetic parameters. Path coefficient methods were used to derive genetic and phenotypic correlations between M-EBV and the phenotypic data. Methods were extended to multi-trait selection and to the case when M-EBV are based on high-density marker genotype data, as in genomic selection. Methods were applied to several example scenarios, which confirmed previous results that MAS substantially increases response to selection but also demonstrated that MAS can result in substantial reductions in the rates of inbreeding. Although further validation by stochastic simulation is required, the developed methodology provides an easy means of deterministically evaluating the potential benefits of MAS and to optimize selection strategies with availability of marker data.     

Effects of a breeding scheme combined by genomic pre‐selection and progeny testing on annual genetic gain in a dairy cattle population

The effectiveness of the incorporation of genomic pre‐selection into dairy cattle progeny testing (GS‐PT) was compared with that of progeny testing (PT) where the fraction of dam to breed bull (DB) selected was 0.01. When the fraction of sires to breed bulls (SB) selected without being progeny tested to produce young bulls (YB) in the next generation was 0.2, the annual genetic gain from GS‐PT was 13% to 43% greater when h² = 0.3 and 16% to 53% greater when h² = 0.1 compared with that from PT. Given h² = 0.3, a selection accuracy of 0.8 for both YB and DB, and selected fractions of 0.117 for YB and 0.04 for DB, GS‐PT produced 40% to 43% greater annual genetic gain than PT. Given h² = 0.1, a selection accuracy of 0.6 for both YB and DB, and selected fractions of 0.117 for YB and 0.04 for DB, annual genetic gain from GS‐PT was 48% to 53% greater than that from PT. When h² = 0.3, progeny testing capacity had little effect on annual genetic gain from GS‐PT. However, when h² = 0.1, annual genetic gain from GS‐PT increased with increasing progeny testing capacity.     

Improving genomic prediction accuracy for meat tenderness in Nellore cattle using artificial neural networks

The goal of this study was to compare the predictive performance of artificial neural networks (ANNs) with Bayesian ridge regression, Bayesian Lasso, Bayes A, Bayes B and Bayes Cπ in estimating genomic breeding values for meat tenderness in Nellore cattle. The animals were genotyped with the Illumina Bovine HD Bead Chip (HD, 777K from 90 samples) and the GeneSeek Genomic Profiler (GGP Indicus HD, 77K from 485 samples). The quality control for the genotypes was applied on each Chip and comprised removal of SNPs located on non-autosomal chromosomes, with minor allele frequency <5%, deviation from HWE (p < 10^–6), and with linkage disequilibrium >0.8. The FImpute program was used for genotype imputation. Pedigree-based analyses indicated that meat tenderness is moderately heritable (0.35), indicating that it can be improved by direct selection. Prediction accuracies were very similar across the Bayesian regression models, ranging from 0.20 (Bayes A) to 0.22 (Bayes B) and 0.14 (Bayes Cπ) to 0.19 (Bayes A) for the additive and dominance effects, respectively. ANN achieved the highest accuracy (0.33) of genomic prediction of genetic merit. Even though deep neural networks are recognized to deliver more accurate predictions, in our study ANN with one single hidden layer, 105 neurons and rectified linear unit (ReLU) activation function was sufficient to increase the prediction of genetic merit for meat tenderness. These results indicate that an ANN with relatively simple architecture can provide superior genomic predictions for meat tenderness in Nellore cattle.     

Interplay between heritability,genetic correlation and economic weighting in a selection index with and without genomic information

The availability of genomic information demands proper evaluation on how the kind (phenotypic versus genomic) and the amount of information influences the interplay of heritability (h²), genetic correlation () and economic weighting of traits with regard to the standard deviation of the index (σ_I). As σ_I is directly proportional to response to selection, it was the chosen parameter for comparing the indices. Three selection indices incorporating conventional and genomic information for a two trait (i and j) breeding goal were compared. Information sources were chosen corresponding to pig breeding applications. Index I incorporating an own performance in trait j served as reference scenario. In index II, additional information in both traits was contributed by a varying number of full‐sibs (2, 7, 50). In index III, the conventional own performance in trait j was combined with genomic information for both traits. The number of animals in the reference population (N_P = 1000, 5000, 10 000) and thus the accuracy of GBVs were varied. With more information included in the index, σ_I became more independent of , and relative economic weighting. This applied for index II (more full‐sibs) and for index III (more accurate GBVs). Standard deviations of index II with seven full‐sibs and index III with N_P = 1000 were similar when both traits had the same heritability. If the heritability of trait j was reduced ( = 0.1), σ_I of index III with N_P = 1000 was clearly higher than for index II with seven full‐sibs. When enhancing the relative economic weight of trait j, the decrease in σ_I of the conventional full‐sib index was much stronger than for index III. Our results imply that N_P = 1000 can be considered a minimum size for a reference population in pig breeding. These conclusions also hold for comparing the accuracies of the indices.     

Effect of selection and selective genotyping for creation of reference on bias and accuracy of genomic prediction

Reference populations for genomic selection usually involve selected individuals, which may result in biased prediction of estimated genomic breeding values (GEBV). In a simulation study, bias and accuracy of GEBV were explored for various genetic models with individuals selectively genotyped in a typical nucleus breeding program. We compared the performance of three existing methods, that is, Best Linear Unbiased Prediction of breeding values using pedigree‐based relationships (PBLUP), genomic relationships for genotyped animals only (GBLUP) and a Single‐Step approach (SSGBLUP) using both. For a scenario with no‐selection and random mating (RR), prediction was unbiased. However, lower accuracy and bias were observed for scenarios with selection and random mating (SR) or selection and positive assortative mating (SA). As expected, bias disappeared when all individuals were genotyped and used in GBLUP. SSGBLUP showed higher accuracy compared to GBLUP, and bias of prediction was negligible with SR. However, PBLUP and SSGBLUP still showed bias in SA due to high inbreeding. SSGBLUP and PBLUP were unbiased provided that inbreeding was accounted for in the relationship matrices. Selective genotyping based on extreme phenotypic contrasts increased the prediction accuracy, but prediction was biased when using GBLUP. SSGBLUP could correct the biasedness while gaining higher accuracy than GBLUP. In a typical animal breeding program, where it is too expensive to genotype all animals, it would be appropriate to genotype phenotypically contrasting selection candidates and use a Single‐Step approach to obtain accurate and unbiased prediction of GEBV.     

Performance of genomic selection under a single-step approach in autochthonous Spanish beef cattle populations

This study evaluated different strategies for implementing a single-step genomic selection programme in two autochthonous Spanish beef cattle populations (Pirenaica—Pi and Rubia Gallega—RG). The strategies were compared in terms of accuracy attained under different scenarios by simulating genomic data over the known genealogy. Several genotyping approaches were tested, as well as, other factors like marker density, effective population size, mutation rate and heritability of the trait. The results obtained showed gains in accuracy with respect to pedigree BLUP evaluation in all cases. The greatest benefit was obtained when the candidates to selection had their genotypes included in the evaluation. Moreover, genotyping the individuals with the most accurate predictions maximized the gains but other suboptimal strategies also yielded satisfactory results. Furthermore, the gains in accuracy increased with the marker density reaching a plateau at around 50,000 markers. Likewise, the effective population size and the mutation rate have also shown an effect, both increasing the accuracy with decreasing values of these population parameters. Finally, the results obtained for the RG population showed greater gains compared to the Pi population, probably attributed to the wider implantation of artificial insemination.     

Accuracy of genomic prediction using different models and response variables in the Nordic Red cattle population

Breeding animals can be accurately evaluated using appropriate genomic prediction models, based on marker data and phenotype information. In this study, direct genomic values (DGV) were estimated for 16 traits of Nordic Total Merit (NTM) Index in Nordic Red cattle population using three models and two different response variables. The three models were as follows: a linear mixed model (GBLUP), a Bayesian variable selection model similar to BayesA (BayesA*) and a Bayesian least absolute shrinkage and selection operator model (Bayesian Lasso). The response variables were deregressed proofs (DRP) and conventional estimated breeding values (EBV). The reliability of genomic predictions was measured on bulls in the validation data set as the squared correlation between DGV and DRP divided by the reliability of DRP. Using DRP as response variable, the reliabilities of DGV among the 16 traits ranged from 0.151 to 0.569 (average 0.317) for GBLUP, from 0.152 to 0.576 (average 0.318) for BayesA* and from 0.150 to 0.570 (average 0.320) for Bayesian Lasso. Using EBV as response variable, the reliabilities ranged from 0.159 to 0.580 (average 0.322) for GBLUP, from 0.157 to 0.578 (average 0.319) for BayesA* and from 0.159 to 0.582 (average 0.325) for Bayesian Lasso. In summary, Bayesian Lasso performed slightly better than the other two models, and EBV performed slightly better than DRP as response variable, with regard to prediction reliability of DGV. However, these differences were not statistically significant. Moreover, using EBV as response variable would result in problems with the scale of the resulting DGV and potential problem due to double counting.     

Review: optimizing genomic selection for crossbred performance by model improvement and data collection

Breeding programs aiming to improve the performance of crossbreds may benefit from genomic prediction of crossbred (CB) performance for purebred (PB) selection candidates. In this review, we compared genomic prediction strategies that differed in 1) the genomic prediction model used or 2) the data used in the reference population. We found 27 unique studies, two of which used deterministic simulation, 11 used stochastic simulation, and 14 real data. Differences in accuracy and response to selection between strategies depended on i) the value of the purebred crossbred genetic correlation (rpc), ii) the genetic distance between the parental lines, iii) the size of PB and CB reference populations, and iv) the relatedness of these reference populations to the selection candidates. In studies where a PB reference population was used, the use of a dominance model yielded accuracies that were equal to or higher than those of additive models. When rpc was lower than ~0.8, and was caused mainly by G × E, it was beneficial to create a reference population of PB animals that are tested in a CB environment. In general, the benefit of collecting CB information increased with decreasing rpc. For a given rpc, the benefit of collecting CB information increased with increasing size of the reference populations. Collecting CB information was not beneficial when rpc was higher than ~0.9, especially when the reference populations were small. Collecting only phenotypes of CB animals may slightly improve accuracy and response to selection, but requires that the pedigree is known. It is, therefore, advisable to genotype these CB animals as well. Finally, considering the breed-origin of alleles allows for modeling breed-specific effects in the CB, but this did not always lead to higher accuracies. Our review shows that the differences in accuracy and response to selection between strategies depend on several factors. One of the most important factors is rpc, and we, therefore, recommend to obtain accurate estimates of rpc of all breeding goal traits. Furthermore, knowledge about the importance of components of rpc (i.e., dominance, epistasis, and G × E) can help breeders to decide which model to use, and whether to collect data on animals in a CB environment. Future research should focus on the development of a tool that predicts accuracy and response to selection from scenario specific parameters.     

A comparison of methods to estimate genomic relationships using pedigree and markers in livestock populations

Accurate prediction of breeding values depends on capturing the variability in genome sharing of relatives with the same pedigree relationship. Here, we compare two approaches to set up genomic relationship matrices for precision of genomic relationships (GR) and accuracy of estimated breeding values (GEBV). Real and simulated data (pigs, 60k SNP) were analysed, and GR were estimated using two approaches: (i) identity by state, corrected with either the observed ( G _{VR ‐O} ) or the base population ( G _{VR ‐B} ) allele frequencies and (ii) identity by descent using linkage analysis ( G _{IBD ‐L} ). Estimators were evaluated for precision and empirical bias with respect to true pedigree IBD GR. All three estimators had very low bias. G _{IBD ‐L} displayed the lowest sampling error and the highest correlation with true genome‐shared values. G _{VR ‐B} approximated G _{IBD ‐L} 's correlation and had lower error than G _{VR ‐O} . Accuracy of GEBV for selection candidates was significantly higher when G _{IBD ‐L} was used and identical between G _{VR ‐O} and G _{VR ‐B} . In real data, G _{IBD ‐L} 's sampling standard deviation was the closest to the theoretical value for each pedigree relationship. Use of pedigree to calculate GR improved the precision of estimates and the accuracy of GEBV.     

Machine learning classification procedure for selecting SNPs in genomic selection: application to early mortality in broilers

Summary Genome‐wide association studies using single nucleotide polymorphisms (SNPs) can identify genetic variants related to complex traits. Typically thousands of SNPs are genotyped, whereas the number of phenotypes for which there is genomic information may be smaller. When predicting phenotypes, options for statistical model building range from incorporating all possible markers into the specification to including only sets of relevant SNPs (features). In the latter case, an efficient method of selecting influential features is required. A two‐step feature selection method for binary traits was developed, which consisted of filtering (using information gain), and wrapping (using naïve Bayesian classification). The filter reduces the large number of SNPs to a much smaller size, to facilitate the wrapper step. As the procedure is tailored for discrete outcomes, an approach based on discretization of phenotypic values was developed, to enable feature selection in a classification framework. The method was applied to chick mortality rates (0–14 days of age) on progeny from 201 sires in a commercial broiler line, with the goal of identifying SNPs (over 5000) related to progeny mortality. To mimic a case–control study, sires were clustered into two groups, low and high, according to two arbitrarily chosen mortality rate cut points. By varying these thresholds, 11 different ‘case–control’ samples were formed, and the SNP selection procedure was applied to each sample. To compare the 11 sets of chosen SNPs, predicted residual sum of squares (PRESS) from a linear model was used. The two‐step method improved naïve Bayesian classification accuracy over the case without feature selection (from around 50 to above 90% without and with feature selection in each case–control sample). The best case–control group (63 sires above or below the thresholds) had the smallest PRESS statistic among groups with model p‐values below 0.003. The 17 SNPs selected using this group accounted for 31% of the variation in raw mortality rates between sire families.     

Tag-SNP selection using Bayesian genomewide association study for growth traits in Hereford and Braford cattle

The aim of this study was to perform a Bayesian genomewide association study (GWAS) to identify genomic regions associated with growth traits in Hereford and Braford cattle, and to select Tag-SNPs to represent these regions in low-density panels useful for genomic predictions. In addition, we propose candidate genes through functional enrichment analysis associated with growth traits using Medical Subject Headings (MeSH). Phenotypic data from 126,290 animals and genotypes for 131 sires and 3,545 animals were used. The Tag-SNPs were selected with BayesB (π = 0.995) method to compose low-density panels. The number of Tag-single nucleotide polymorphism (SNP) ranged between 79 and 103 SNP for the growth traits at weaning and between 78 and 100 SNP for the yearling growth traits. The average proportion of variance explained by Tag-SNP with BayesA was 0.29, 0.23, 0.32 and 0.19 for birthweight (BW), weaning weight (WW205), yearling weight (YW550) and postweaning gain (PWG345), respectively. For Tag-SNP with BayesA method accuracy values ranged from 0.13 to 0.30 for k-means and from 0.30 to 0.65 for random clustering of animals to compose reference and validation groups. Although genomic prediction accuracies were higher with the full marker panel, predictions with low-density panels retained on average 76% of the accuracy obtained with BayesB with full markers for growth traits. The MeSH analysis was able to translate genomic information providing biological meanings of more specific gene products related to the growth traits. The proposed Tag-SNP panels may be useful for future fine mapping studies and for lower-cost commercial genomic prediction applications.     

Estimation of variance components for carcass traits in Japanese Black cattle using 50K SNP genotype data

Genomic selection using high‐density single nucleotide polymorphism (SNP) genotype data may accelerate genetic improvements in livestock animals. In this study, we attempted to estimate the variance components of six carcass traits in fattened Japanese Black steers using SNP genotype data. Six hundred and seventy‐three steers were genotyped using an Illumina Bovine SNP50 BeadChip and phenotyped for cold carcass weight, ribeye area, rib thickness, subcutaneous fat thickness, estimated yield percent and marbling score. Additive polygenic variance and the variance attributable to a set of SNPs that had statistically significant effects on the trait were estimated via Gibbs sampling with two models: (i) a model with the chosen SNPs and the additive polygenic effects; and (ii) a model with the polygenic effects alone. The proportion of the estimated variance attributable to the SNPs became higher as the number of SNP effects that fit increased. High correlations between breeding values estimated with the model containing the polygenic effect alone and those estimated by chosen SNPs were obtained. No fraction of the total genetic variance was explained by SNPs associated with the trait at P ≥ 0.1. Our results suggest that for the carcass traits of Japanese Black cattle, a maximum of half of the total additive genetic variance may be explained by SNPs between 100 several tens to several 100s.     

Impact of relationships between test and training animals and among training animals on reliability of genomic prediction

One of the factors affecting the reliability of genomic prediction is the relationship among the animals of interest. This study investigated the reliability of genomic prediction in various scenarios with regard to the relationship between test and training animals, and among animals within the training data set. Different training data sets were generated from EuroGenomics data and a group of Nordic Holstein bulls (born in 2005 and afterwards) as a common test data set. Genomic breeding values were predicted using a genomic best linear unbiased prediction model and a Bayesian mixture model. The results showed that a closer relationship between test and training animals led to a higher reliability of genomic predictions for the test animals, while a closer relationship among training animals resulted in a lower reliability. In addition, the Bayesian mixture model in general led to a slightly higher reliability of genomic prediction, especially for the scenario of distant relationships between training and test animals. Therefore, to prevent a decrease in reliability, constant updates of the training population with animals from more recent generations are required. Moreover, a training population consisting of less‐related animals is favourable for reliability of genomic prediction.     

Genomic prediction for meat and carcass traits in Nellore cattle using a Markov blanket algorithm

This study was carried out to evaluate the advantage of preselecting SNP markers using Markov blanket algorithm regarding the accuracy of genomic prediction for carcass and meat quality traits in Nellore cattle. This study considered 3675, 3680, 3660 and 524 records of rib eye area (REA), back fat thickness (BF), rump fat (RF), and Warner–Bratzler shear force (WBSF), respectively, from the Nellore Brazil Breeding Program. The animals have been genotyped using low-density SNP panel (30 k), and subsequently imputed for arrays with 777 k SNPs. Four Bayesian specifications of genomic regression models, namely Bayes A, Bayes B, Bayes Cπ and Bayesian Ridge Regression methods were compared in terms of prediction accuracy using a five folds cross-validation. Prediction accuracy for REA, BF and RF was all similar using the Bayesian Alphabet models, ranging from 0.75 to 0.95. For WBSF, the predictive ability was higher using Bayes B (0.47) than other methods (0.39 to 0.42). Although the prediction accuracies using Markov blanket of SNP markers were lower than those using all SNPs, for WBSF the relative gain was lower than 13%. With a subset of informative SNPs markers, identified using Markov blanket, probably, is possible to capture a large proportion of the genetic variance for WBSF. The development of low-density and customized arrays using Markov blanket might be cost-effective to perform a genomic selection for this trait, increasing the number of evaluated animals, improving the management decisions based on genomic information and applying genomic selection on a large scale.     

Identification of differential selection traces in two Polish cattle breeds

Genetic improvement of animals based on artificial selection is leading to changes in the frequency of genes related to desirable production traits. The changes are reflected by the neutral, intergenic single nucleotide polymorphims (SNPs) being in long‐range linkage disequilibrium with functional polymorphisms. Genome‐wide SNP analysis tools designed for cattle, allow for scanning divergences in allelic frequencies between distinct breeds and thus for identification of genomic regions which were divergently selected in breeds' histories. In this study, by using Bovine SNP50 assay, we attempted to identify genomic regions showing the highest differences in allele frequencies between two distinct cattle breeds – preserved, unselected Polish Red breed and highly selected Holstein cattle. Our study revealed 19 genomic regions encompassing 55 protein‐coding genes and numerous quantitative trait loci which potentially may underlie some of the phenotypic traits distinguishing the breeds.     

Genomic selection strategies in dairy cattle: Strong positive interaction between use of genotypic information and intensive use of young bulls on genetic gain

We tested the following hypotheses: (i) breeding schemes with genomic selection are superior to breeding schemes without genomic selection regarding annual genetic gain of the aggregate genotype (ΔG(AG) ), annual genetic gain of the functional traits and rate of inbreeding per generation (ΔF), (ii) a positive interaction exists between the use of genotypic information and a short generation interval on ΔG(AG) and (iii) the inclusion of an indicator trait in the selection index will only result in a negligible increase in ΔG(AG) if genotypic information about the breeding goal trait is known. We examined four breeding schemes with or without genomic selection and with or without intensive use of young bulls using pseudo-genomic stochastic simulations. The breeding goal consisted of a milk production trait and a functional trait. The two breeding schemes with genomic selection resulted in higher ΔG(AG) , greater contributions of the functional trait to ΔG(AG) and lower ΔF than the two breeding schemes without genomic selection. Thus, the use of genotypic information may lead to more sustainable breeding schemes. In addition, a short generation interval increases the effect of using genotypic information on ΔG(AG) . Hence, a breeding scheme with genomic selection and with intensive use of young bulls (a turbo scheme) seems to offer the greatest potential. The third hypothesis was disproved as inclusion of genomically enhanced breeding values (GEBV) for an indicator trait in the selection index increased ΔG(AG) in the turbo scheme. Moreover, it increased the contribution of the functional trait to ΔG(AG) , and it decreased ΔF. Thus, indicator traits may still be profitable to use even when GEBV for the breeding goal traits are available.     

Use of different statistical models to predict direct genomic values for productive and functional traits in Italian Holsteins

One of the main issues in genomic selection was the huge unbalance between number of markers and phenotypes available. In this work, principal component analysis is used to reduce the number of predictors for calculating direct genomic breeding values (DGV) for production and functional traits. 2093 Italian Holstein bulls were genotyped with the 54 K Illumina beadchip, and 39 555 SNP markers were retained after data editing. Principal Components (PC) were extracted from SNP matrix, and 15 207 PC explaining 99% of the original variance were retained and used as predictors. Bulls born before 2001 were included in the reference population, younger animals in the test population. A BLUP model was used to estimate the effect of principal component on deregressed proof (DRPF) for 35 traits and results were compared to those obtained by using SNP genotypes as predictors either with BLUP or with Bayes_A models. Correlations between DGV and DRPF did not substantially differ among the three methods except for milk fat content. The lowest prediction bias was obtained for the method based on the use of principal component. Regression coefficients of DRPF on DGV were lower than one for the approach based on the use of PC and higher than one for the other two methods. The use of PC as predictors resulted in a large reduction of number of predictors (approximately 38%) and of computational time that was approximately 2% of the time needed to estimate SNP effects with the other two methods. Accuracies of genomic predictions were in most of cases only slightly higher than those of the traditional pedigree index, probably due to the limited size of the considered population.