Across breed multi‐trait random regression genomic predictions in the Nordic Red dairy cattle

The current study evaluates reliability of genomic predictions in selection candidates using multi‐trait random regression model, which accounts for interactions between marker effects and breed of origin in the Nordic Red dairy cattle (RDC). The population structure of the RDC is admixed. Data consisted of individual animal breed proportions calculated from the full pedigree, deregressed proofs (DRP) of published estimated breeding values (EBV) for yield traits and genotypic data for 37 595 single nucleotide polymorphic markers. The analysed data included 3330 bulls in the reference population and 812 bulls that were used for validation. Direct genomic breeding values (DGV) were estimated using the model under study, which accounts for breed effects and also with GBLUP, which assume uniform population. Validation reliability was calculated as a coefficient of determination from weighted regression of DRP on DGV (), scaled by the mean reliability of DRP. Using the breed‐specific model increased the reliability of DGV by 2 and 3% for milk and protein, respectively, when compared to homogeneous population GBLUP. The exception was for fat, where there was no gain in reliability. Estimated validation reliabilities were low for milk (0.32) and protein (0.32) and slightly higher (0.42) for fat.     

Influence of model specifications on the reliabilities of genomic prediction in a Swedish-Finnish red breed cattle population

Using a combined multi‐breed reference population, this study explored the influence of model specification and the effect of including a polygenic effect on the reliability of genomic breeding values (DGV and GEBV). The combined reference population consisted of 2986 Swedish Red Breed (SRB) and Finnish Ayrshire (FAY) dairy cattle. Bayesian methodology (common prior and mixture models with different prior distribution settings for the marker effects) as well as a best linear unbiased prediction with a genomic relationship matrix [genomic best linear unbiased predictor (GBLUP)] was used in the prediction of DGV. Mixture models including a polygenic effect were used to predict GEBV. In total, five traits with low, high and medium heritability were analysed. For the models using a mixture prior distribution, reliabilities of DGV tended to decrease with an increasing proportion of markers with small effects. The influence of the inclusion of a polygenic effect on the reliability of DGV varied across traits and model specifications. Average correlation between DGV with the Mendelian sampling term, across traits, was highest (R² = 0.25) for the GBLUP model and decreased with increasing proportion of markers with large effects. Reliabilities increased when DGV and parent average information were combined in an index. The GBLUP model with the largest gain across traits in the reliability of the index achieved the highest DGV mean reliability. However, the polygenic models showed to be less biased and more consistent in the estimation of DGV regardless of the model specifications compared with the mixture models without the polygenic effect.     

Comparison between genomic predictions using daughter yield deviation and conventional estimated breeding value as response variables

This study compared genomic predictions using conventional estimated breeding values (EBV) and daughter yield deviations (DYD) as response variables based on simulated data. Eight scenarios were simulated in regard to heritability (0.05 and 0.30), number of daughters per sire (30, 100, and unequal numbers with an average of 100 per sire) and numbers of genotyped sires (all or half of sires were genotyped). The simulated genome had a length of 1200 cM with 15,000 equally spaced Single-nucleotide polymorphism (SNP) markers and 500 randomly distributed Quantitative trait locus (QTL). In the simulated scenarios, the EBV approach was as effective as or slightly better than the DYD approach at predicting breeding value, dependent on simulated scenarios and statistical models. Applying a Bayesian common prior model (the same prior distribution of marker effect variance) and a linear mixed model (GBLUP), the EBV and DYD approaches provided similar genomic estimated breeding value (GEBV) reliabilities, except for scenarios with unequal numbers of daughters and half of sires without genotype, for which the EBV approach was superior to the DYD approach (by 1.2 and 2.4%). Using a Bayesian mixture prior model (mixture prior distribution of marker effect variance), the EBV approach resulted in slightly higher reliabilities of GEBV than the DYD approach (by 0.3-3.6% with an average of 1.9%), and more obvious in scenarios with low heritability, small or unequal numbers of daughters, and half of sires without genotype. Moreover, the results showed that the correlation between GEBV and conventional parent average (PA) was lower (corresponding to a relatively larger gain by including PA) when using the DYD approach than when using the EBV approach. Consequently, the two approaches led to similar reliability of an index combining GEBV and PA in most scenarios. These results indicate that EBV can be used as an alternative response variable for genomic prediction.     

Efficient approximation of reliabilities for single-step genomic best linear unbiased predictor models with the Algorithm for Proven and Young

The objectives of this study were to develop an efficient algorithm for calculating prediction error variances (PEVs) for genomic best linear unbiased prediction (GBLUP) models using the Algorithm for Proven and Young (APY), extend it to single-step GBLUP (ssGBLUP), and apply this algorithm for approximating the theoretical reliabilities for single- and multiple-trait models in ssGBLUP. The PEV with APY was calculated by block sparse inversion, efficiently exploiting the sparse structure of the inverse of the genomic relationship matrix with APY. Single-step GBLUP reliabilities were approximated by combining reliabilities with and without genomic information in terms of effective record contributions. Multi-trait reliabilities relied on single-trait results adjusted using the genetic and residual covariance matrices among traits. Tests involved two datasets provided by the American Angus Association. A small dataset (Data1) was used for comparing the approximated reliabilities with the reliabilities obtained by the inversion of the left-hand side of the mixed model equations. A large dataset (Data2) was used for evaluating the computational performance of the algorithm. Analyses with both datasets used single-trait and three-trait models. The number of animals in the pedigree ranged from 167,951 in Data1 to 10,213,401 in Data2, with 50,000 and 20,000 genotyped animals for single-trait and multiple-trait analysis, respectively, in Data1 and 335,325 in Data2. Correlations between estimated and exact reliabilities obtained by inversion ranged from 0.97 to 0.99, whereas the intercept and slope of the regression of the exact on the approximated reliabilities ranged from 0.00 to 0.04 and from 0.93 to 1.05, respectively. For the three-trait model with the largest dataset (Data2), the elapsed time for the reliability estimation was 11 min. The computational complexity of the proposed algorithm increased linearly with the number of genotyped animals and with the number of traits in the model. This algorithm can efficiently approximate the theoretical reliability of genomic estimated breeding values in ssGBLUP with APY for large numbers of genotyped animals at a low cost.     

Effect of marker‐data editing on the accuracy of genomic prediction

Genomic selection is a method to predict breeding values using genome‐wide single‐nucleotide polymorphism (SNP) markers. High‐quality marker data are necessary for genomic selection. The aim of this study was to investigate the effect of marker‐editing criteria on the accuracy of genomic predictions in the Nordic Holstein and Jersey populations. Data included 4429 Holstein and 1071 Jersey bulls. In total, 48 222 SNP for Holstein and 44 305 SNP for Jersey were polymorphic. The SNP data were edited based on (i) minor allele frequencies (MAF) with thresholds of no limit, 0.001, 0.01, 0.02, 0.05 and 0.10, (ii) deviations from Hardy–Weinberg proportions (HWP) with thresholds of no limit, chi‐squared p‐values of 0.001, 0.02, 0.05 and 0.10, and (iii) GenCall (GC) scores with thresholds of 0.15, 0.55, 0.60, 0.65 and 0.70. The marker data sets edited with different criteria were used for genomic prediction of protein yield, fertility and mastitis using a Bayesian variable selection and a GBLUP model. De‐regressed EBV were used as response variables. The result showed little difference between prediction accuracies based on marker data sets edited with MAF and deviation from HWP. However, accuracy decreased with more stringent thresholds of GC score. According to the results of this study, it would be appropriate to edit data with restriction of MAF being between 0.01 and 0.02, a p‐value of deviation from HWP being 0.05, and keeping all individual SNP genotypes having a GC score over 0.15.     

Genomic predictions across Nordic Holstein and Nordic Red using the genomic best linear unbiased prediction model with different genomic relationship matrices

This study investigated genomic predictions across Nordic Holstein and Nordic Red using various genomic relationship matrices. Different sources of information, such as consistencies of linkage disequilibrium (LD) phase and marker effects, were used to construct the genomic relationship matrices (G‐matrices) across these two breeds. Single‐trait genomic best linear unbiased prediction (GBLUP) model and two‐trait GBLUP model were used for single‐breed and two‐breed genomic predictions. The data included 5215 Nordic Holstein bulls and 4361 Nordic Red bulls, which was composed of three populations: Danish Red, Swedish Red and Finnish Ayrshire. The bulls were genotyped with 50 000 SNP chip. Using the two‐breed predictions with a joint Nordic Holstein and Nordic Red reference population, accuracies increased slightly for all traits in Nordic Red, but only for some traits in Nordic Holstein. Among the three subpopulations of Nordic Red, accuracies increased more for Danish Red than for Swedish Red and Finnish Ayrshire. This is because closer genetic relationships exist between Danish Red and Nordic Holstein. Among Danish Red, individuals with higher genomic relationship coefficients with Nordic Holstein showed more increased accuracies in the two‐breed predictions. Weighting the two‐breed G‐matrices by LD phase consistencies, marker effects or both did not further improve accuracies of the two‐breed predictions.     

Genomic best linear unbiased prediction method reflecting the degree of linkage disequilibrium

The degree of linkage disequilibrium (LD) between markers differs depending on the location of the genome; this difference biases genetic evaluation by genomic best linear unbiased prediction (GBLUP). To correct this bias, we used three GBLUP methods reflecting the degree of LD (GBLUP‐LD). In the three GBLUP‐LD methods, genomic relationship matrices were conducted from single nucleotide polymorphism markers weighted according to local LD levels. The predictive abilities of GBLUP‐LD were investigated by estimating variance components and assessing the accuracies of estimated breeding values using simulation data. When quantitative trait loci (QTL) were located at weak LD regions, the predictive abilities of the three GBLUP‐LD methods were superior to those of GBLUP and Bayesian lasso except when the number of QTL was small. In particular, the superiority of GBLUP‐LD increased with decreasing trait heritability. The rates of QTL at weak LD regions would increase when selection by GBLUP continues; this consequently decreases the predictive ability of GBLUP. Thus, the GBLUP‐LD could be applicable for populations selected by GBLUP for a long time. However, if QTL were located at strong LD regions, the accuracies of three GBLUP‐LD methods were lower than GBLUP and Bayesian lasso.     

Application of multiple shrinkage methods to genomic predictions

New challenges have arisen with the development of large marker panels for livestock species. Models easily become overparameterized when all available markers are included. Solutions have led to the development of shrinkage or regularization techniques. The objective of this study was the application and comparison of Bayesian LASSO (B-L), thick-tailed (Student-t), and semiparametric multiple shrinkage methods. The B-L and Student-t methods were also each analyzed within a single shrinkage and a multiple shrinkage framework. Simulated and real data were used to evaluate each method's performance. Real data consisted of SNP genotypes of 4,069 Holstein sires. Traits included in analysis of real data were milk, fat, protein yield, and somatic cell score. The performance of each model was compared based on correlations between true and predicted genomic predicted transmitting abilities. Model performance was also compared with the performance of routinely used methods such as Bayes-A and GBLUP through cross-validation techniques. When using simulated data regardless of shrinkage framework, shrinkage models outperformed genomic BLUP (GBLUP). The average advantage of shrinkage models ranged from 1% to approximately 8% depending on the prior specification. When analyzing real data, shrinkage models slightly outperformed GBLUP for most traits. Shrinkage models were better able to model traits for which 1 or more SNP of large effect have been identified. Overall, results suggested a relatively small advantage in multiple shrinkage models. Multiple shrinkage methods could represent a useful alternative to current methods of prediction; however, their performance in a variety of scenarios needs to be investigated further.     

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.     

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.     

Accuracy of genomic prediction using imputed whole‐genome sequence data in white layers

There is an increasing interest in using whole‐genome sequence data in genomic selection breeding programmes. Prediction of breeding values is expected to be more accurate when whole‐genome sequence is used, because the causal mutations are assumed to be in the data. We performed genomic prediction for the number of eggs in white layers using imputed whole‐genome resequence data including ~4.6 million SNPs. The prediction accuracies based on sequence data were compared with the accuracies from the 60 K SNP panel. Predictions were based on genomic best linear unbiased prediction (GBLUP) as well as a Bayesian variable selection model (BayesC). Moreover, the prediction accuracy from using different types of variants (synonymous, non‐synonymous and non‐coding SNPs) was evaluated. Genomic prediction using the 60 K SNP panel resulted in a prediction accuracy of 0.74 when GBLUP was applied. With sequence data, there was a small increase (~1%) in prediction accuracy over the 60 K genotypes. With both 60 K SNP panel and sequence data, GBLUP slightly outperformed BayesC in predicting the breeding values. Selection of SNPs more likely to affect the phenotype (i.e. non‐synonymous SNPs) did not improve the accuracy of genomic prediction. The fact that sequence data were based on imputation from a small number of sequenced animals may have limited the potential to improve the prediction accuracy. A small reference population (n = 1004) and possible exclusion of many causal SNPs during quality control can be other possible reasons for limited benefit of sequence data. We expect, however, that the limited improvement is because the 60 K SNP panel was already sufficiently dense to accurately determine the relationships between animals in our data.     

Mixture models detect large effect QTL better than GBLUP and result in more accurate and persistent predictions

Background: Accurate evaluation of SNP effects is important for genome wide association studies and for genomic prediction. The genetic architecture of quantitative traits differs widely, with some traits exhibiting few if any quantitative trait loci(QTL) with large effects, while other traits have one or several easily detectable QTL with large effects.Methods: Body weight in broilers and egg weight in layers are two examples of traits that have QTL of large effect.A commonly used method for genome wide association studies is to fit a mixture model such as Bayes B that assumes some known proportion of SNP effects are zero. In contrast, the most commonly used method for genomic prediction is known as GBLUP, which involves fitting an animal model to phenotypic data with the variance-covariance or genomic relationship matrix among the animals being determined by genome wide SNP genotypes. Genotypes at each SNP are typically weighted equally in determining the genomic relationship matrix for GBLUP. We used the equivalent marker effects model formulation of GBLUP for this study. We compare these two classes of models using egg weight data collected over 8 generations from 2,324 animals genotyped with a42 K SNP panel.Results: Using data from the first 7 generations, both Bayes B and GBLUP found the largest QTL in a similar well-recognized QTL region, but this QTL was estimated to account for 24 % of genetic variation with Bayes B and less than 1 % with GBLUP. When predicting phenotypes in generation 8 Bayes B accounted for 36 % of the phenotypic variation and GBLUP for 25 %. When using only data from any one generation, the same QTL was identified with Bayes B in all but one generation but never with GBLUP. Predictions of phenotypes in generations 2 to 7 based on only 295 animals from generation 1 accounted for 10 % phenotypic variation with Bayes B but only6 % with GBLUP. Predicting phenotype using only the marker effects in the 1 Mb region that accounted for the largest effect on egg weight from generation 1 data alone accounted for almost 8 % variation using Bayes B but had no predictive power with GBLUP.Conclusions: In conclusion, In the presence of large effect QTL, Bayes B did a better job of QTL detection and its genomic predictions were more accurate and persistent than those from GBLUP.     

The importance of haplotype length and heritability using genomic selection in dairy cattle

Reliabilities for genomic estimated breeding values (GEBV) were investigated by simulation for a typical dairy cattle breeding setting. Scenarios were simulated with different heritabilites ( h ²) and for different haplotype sizes, and seven generations with only genotypes were generated to investigate reliability of GEBV over time. A genome with 5000 single nucleotide polymorphisms (SNP) at distances of 0.1 cM and 50 quantitative trait loci (QTL) was simulated, and a Bayesian variable selection model was implemented to predict GEBV. Highest reliabilities were obtained for 10 SNP haplotypes. At optimal haplotype size, reliabilities in generation 1 without phenotypes ranged from 0.80 for h ² = 0.02 to 0.93 for h ² = 0.30, and in the seventh generation without phenotypes ranged from 0.69 for h ² = 0.02 to 0.86 for h ² = 0.30. Reliabilities of GEBV were found sufficiently high to implement dairy selection schemes without progeny testing in which case a data time-lag of two to three generations may be present. Reliabilities were also relatively high for low heritable traits, implying that genomic selection could be especially beneficial to improve the selection on, e.g. health and fertility.     

Comparison of heritabilities of dairy traits in Australian Holstein‐Friesian cattle from genomic and pedigree data and implications for genomic evaluations

The reliability of genomic evaluations depends on the proportion of genetic variation explained by the DNA markers. In this study, we have estimated the proportion of variance in daughter trait deviations (DTDs) of dairy bulls explained by 45 993 genome wide single‐nucleotide poly‐ morphism (SNP) markers for 29 traits in Australian Holstein‐Friesian dairy cattle. We compare these proportions to the proportion of variance in DTDs explained by the additive relationship matrix derived from the pedigree, as well as the sum of variance explained by both pedigree and marker information when these were fitted simultaneously. The propor‐ tion of genetic variance in DTDs relative to the total genetic variance (the total genetic variance explained by the genomic relationships and pedigree relationships when both were fitted simultaneously) varied from 32% for fertility to approximately 80% for milk yield traits. When fitting genomic and pedigree relationships simultaneously, the variance unexplained (i.e. the residual variance) in DTDs of the total variance for most traits was reduced compared to fitting either individually, suggesting that there is not complete overlap between the effects. The proportion of genetic variance accounted by the genomic relationships can be used to modify the blending equations used to calculate genomic estimated breeding value (GEBV) from direct genomic breeding value (DGV) and parent average. Our results, from a validation population of young dairy bulls with DTD, suggest that this modification can improve the reliability of GEBV by up to 5%.     

Impact of fitting dominance and additive effects on accuracy of genomic prediction of breeding values in layers

Most genomic prediction studies fit only additive effects in models to estimate genomic breeding values (GEBV). However, if dominance genetic effects are an important source of variation for complex traits, accounting for them may improve the accuracy of GEBV. We investigated the effect of fitting dominance and additive effects on the accuracy of GEBV for eight egg production and quality traits in a purebred line of brown layers using pedigree or genomic information (42K single‐nucleotide polymorphism (SNP) panel). Phenotypes were corrected for the effect of hatch date. Additive and dominance genetic variances were estimated using genomic‐based [genomic best linear unbiased prediction (GBLUP)‐REML and BayesC] and pedigree‐based (PBLUP‐REML) methods. Breeding values were predicted using a model that included both additive and dominance effects and a model that included only additive effects. The reference population consisted of approximately 1800 animals hatched between 2004 and 2009, while approximately 300 young animals hatched in 2010 were used for validation. Accuracy of prediction was computed as the correlation between phenotypes and estimated breeding values of the validation animals divided by the square root of the estimate of heritability in the whole population. The proportion of dominance variance to total phenotypic variance ranged from 0.03 to 0.22 with PBLUP‐REML across traits, from 0 to 0.03 with GBLUP‐REML and from 0.01 to 0.05 with BayesC. Accuracies of GEBV ranged from 0.28 to 0.60 across traits. Inclusion of dominance effects did not improve the accuracy of GEBV, and differences in their accuracies between genomic‐based methods were small (0.01–0.05), with GBLUP‐REML yielding higher prediction accuracies than BayesC for egg production, egg colour and yolk weight, while BayesC yielded higher accuracies than GBLUP‐REML for the other traits. In conclusion, fitting dominance effects did not impact accuracy of genomic prediction of breeding values in this population.     

基于简化基因组测序技术和基因芯片技术比较研究黄羽肉鸡基因组选择

旨在比较简化基因组测序技术和基因芯片技术实施基因组选择的基因组估计育种值（GEBV）准确性。本研究在AH肉鸡资源群体F₂代中随机选取395个个体（其中公鸡212只,母鸡183只,来自8个半同胞家系）,同时采用10×SLAF测序技术和Illumina Chicken 60K SNP芯片进行基因标记分型。采用基因组最佳无偏估计法（GBLUP）和BayesCπ对6周体重、12周体重、日均增重、日均采食量、饲料转化率和剩余采食量等6个性状进行GEBV准确性比较研究,并采用5折交叉验证法验证。结果表明,采用同一基因标记分型平台,两种育种值估计方法所得GEBV准确性差异不显著（P>0.05）;不同的性状对基因标记分型平台的选择存在差异,对于6周体重,使用基因芯片可获得更高的GEBV准确性（P<0.05）,对于剩余采食量,则使用简化基因组测序可获得更高的GEBV准确性（P<0.05）。综合6个性状GEBV均值比较,两个基因标记分型平台之间差异不到0.01,高通量测序技术和基因芯片技术都可以用于黄羽肉鸡基因组选择。     

Comparison between Reduced-Representation Genome Sequencing and SNP Chip for Genomic Selection in Yellow-feathered Broiler

This study aimed to compare the accuracy of the genomic estimated breeding value (GEBV) using reduced-representation genome sequencing technology and SNP chip technology to implement genomic selection. A total of 395 individuals (212♂+ 183♀, from 8 half-sib families) were randomly selected from F₂ generation of AH broiler resource population, and genotyped with 10×specific-locus amplified fragment sequencing (SLAF-seq) and Illumina Chicken 60K SNP BeadChip. Genomic best linear unbiased prediction (GBLUP) and BayesCπ were used to compare the accuracy of genomic estimated breeding values (GEBV) for 6 traits: body weight at the 6^th week, body weight at the 12^th week, average daily gain (ADG), average daily feed intake (ADFI), feed conversion ratio (FCR) and residual feed intake (RFI). A 5-fold cross validation procedure was used to verify the accuracies of GEBV between prediction models and between genotyping platforms. The results showed that there was no significant difference between accuracies of GEBV predicted by GBLUP and BayesCπ using the same genotyping platform(P>0.05). The superiority of the two genotyping platforms was different for different traits. For body weight at the 6^th week, the accuracy of GEBV was higher using chip SNPs (P<0.05). On the contrary, the accuracy was higher using SLAF-seq for residual feed intake (P<0.05). Comprehensive comparison of the means of GEBV for 6 traits, the difference between the two genotyping platforms was less than 0.01, therefore, both high throughput sequencing and chip SNPs can be used for genomic selection in yellow-feathered broiler.     

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.     

Genomic prediction in a numerically small breed population using prioritized genetic markers from whole-genome sequence data

The objective of this study was to investigate the accuracy of genomic prediction of body weight and eating quality traits in a numerically small sheep population (Dorper sheep). Prediction was based on a large multi-breed/admixed reference population and using (a) 50k or 500k single nucleotide polymorphism (SNP) genotypes, (b) imputed whole-genome sequencing data (~31 million), (c) selected SNPs from whole genome sequence data and (d) 50k SNP genotypes plus selected SNPs from whole-genome sequence data. Furthermore, the impact of using a breed-adjusted genomic relationship matrix on accuracy of genomic breeding value was assessed. The selection of genetic variants was based on an association study performed on imputed whole-genome sequence data in an independent population, which was chosen either randomly from the base population or according to higher genetic proximity to the target population. Genomic prediction was based on genomic best linear unbiased prediction (GBLUP), and the accuracy of genomic prediction was assessed according to the correlation between genomic breeding value and corrected phenotypes divided by the square root of trait heritability. The accuracy of genomic prediction was between 0.20 and 0.30 across different traits based on common 50k SNP genotypes, which improved on average by 0.06 (absolute value) on average based on using prioritized genetic markers from whole-genome sequence data. Using prioritized genetic markers from a genetically more related GWAS population resulted in slightly higher prediction accuracy (0.02 absolute value) compared to genetic markers derived from a random GWAS population. Using high-density SNP genotypes or imputed whole-genome sequence data in GBLUP showed almost no improvement in genomic prediction accuracy however, accounting for different marker allele frequencies in reference population according to a breed-adjusted GRM resulted to on average 0.024 (absolute value) increase in accuracy of genomic prediction.     

Assessment of bagging GBLUP for whole‐genome prediction of broiler chicken traits

Bootstrap aggregation (bagging) is a resampling method known to produce more accurate predictions when predictors are unstable or when the number of markers is much larger than sample size, because of variance reduction capabilities. The purpose of this study was to compare genomic best linear unbiased prediction (GBLUP) with bootstrap aggregated sampling GBLUP (Bagged GBLUP, or BGBLUP) in terms of prediction accuracy. We used a 600 K Affymetrix platform with 1351 birds genotyped and phenotyped for three traits in broiler chickens; body weight, ultrasound measurement of breast muscle and hen house egg production. The predictive performance of GBLUP versus BGBLUP was evaluated in different scenarios consisting of including or excluding the TOP 20 markers from a standard genome‐wide association study (GWAS) as fixed effects in the GBLUP model, and varying training sample sizes and allelic frequency bins. Predictive performance was assessed via five replications of a threefold cross‐validation using the correlation between observed and predicted values, and prediction mean‐squared error. GBLUP overfitted the training set data, and BGBLUP delivered a better predictive ability in testing sets. Treating the TOP 20 markers from the GWAS into the model as fixed effects improved prediction accuracy and added advantages to BGBLUP over GBLUP. The performance of GBLUP and BGBLUP at different allele frequency bins and training sample sizes was similar. In general, results of this study confirm that BGBLUP can be valuable for enhancing genome‐enabled prediction of complex traits.