Genomic selection for crossbred performance accounting for breed-specific effects

Imputation of genotypes with low-density chips and its effect on reliability of direct genomic values in Dutch Holstein cattle

Multiple-breed genomic evaluation by principal component analysis in small size populations

The objective of this study was to estimate accuracies of direct genomic breeding values (DGV) for nationally evaluated traits of 1,081 American (AMH), 100 Argentine (ARH), 75 Canadian (CAH), and 395 Uruguayan (URH) Hereford animals genotyped using the Illumina BovineSNP50 BeadChip. Deregressed EBV (DEBV) were used as observations in a weighted analysis to derive DGV using BayesB and BayesC methods. The AMH animals were clustered into 4 groups, using either K-means or random clustering. Cross validation was performed with the group not used in training providing validation of the accuracies of estimated DGV. Genomic predictions were also evaluated for AMH animals by training on older animals and validating on younger animals. Bivariate animal models were used for each trait to estimate genetic correlations between DEBV and DGV. Genomic predictions were separately evaluated for foreign animals from each country using marker estimates from training on AMH or pooled international data. Pedigree estimated breeding values were developed for AMH animals, using traditional, pedigree-based BLUP (PBLUP) for comparison purposes. Using BayesB (BayesC) method, the average simple correlations between DGV and DEBV in AMH animals was 0.24 (0.21), 0.39 (0.36), and 0.32 (0.30) when training and validation sets were formed by K-means clustering, random allocation or year of birth of the animals, respectively. Genetic correlations between DEBV and DGV ranged from 0.20 (0.18) to 0.52 (0.45) in AMH animals. The DGV from BayesB were more accurate than from BayesC for most traits in AMH animals. Genomic predictions for foreign animals were less accurate than those obtained in AMH animals. Among foreign animals, genomic predictions were more accurate for CAH animals, which reflect the greater use of AMH sires in CAH in comparison with ARH and URH populations. Small changes in accuracies of DGV were observed for foreign animals by using admixed training populations. On average, genomic predictions across countries were more accurate for CAH and URH animals using BayesB. On average, accuracies of genomic predictions using BayesB (BayesC) method were 66% (55%) greater than those obtained from PBLUP. These results demonstrate the feasibility of developing DGV for American Hereford beef cattle. However, foreign breeders, especially South American Hereford breeders, need to genotype more animals to obtain more accurate genomic predictions.

One of the main animal health problems in tropical and subtropical cattle production is the bovine tick, which causes decreased performance, hide devaluation, increased production costs with acaricide treatments, and transmission of infectious diseases. This study investigated the utility of genomic prediction as a tool to select Braford (BO) and Hereford (HH) cattle resistant to ticks. The accuracy and bias of different methods for direct and blended genomic prediction was assessed using 10,673 tick counts obtained from 3,435 BO and 928 HH cattle belonging to the Delta G Connection breeding program. A subset of 2,803 BO and 652 HH samples were genotyped and 41,045 markers remained after quality control. Log transformed records were adjusted by a pedigree repeatability model to estimate variance components, genetic parameters, and breeding values (EBV) and subsequently used to obtain deregressed EBV. Estimated heritability and repeatability for tick counts were 0.19 ± 0.03 and 0.29 ± 0.01, respectively. Data were split into 5 subsets using k-means and random clustering for cross-validation of genomic predictions. Depending on the method, direct genomic value (DGV) prediction accuracies ranged from 0.35 with Bayes least absolute shrinkage and selection operator (LASSO) to 0.39 with BayesB for k-means clustering and between 0.42 with BayesLASSO and 0.45 with BayesC for random clustering. All genomic methods were superior to pedigree BLUP (PBLUP) accuracies of 0.26 for k-means and 0.29 for random groups, with highest accuracy gains obtained with BayesB (39%) for k-means and BayesC (55%) for random groups. Blending of historical phenotypic and pedigree information by different methods further increased DGV accuracies by values between 0.03 and 0.05 for direct prediction methods. However, highest accuracy was observed with single-step genomic BLUP with values of 0.48 for -means and 0.56, which represent, respectively, 84 and 93% improvement over PBLUP. Observed random clustering cross-validation breed-specific accuracies ranged between 0.29 and 0.36 for HH and between 0.55 and 0.61 for BO, depending on the blending method. These moderately high values for BO demonstrate that genomic predictions could be used as a practical tool to improve genetic resistance to ticks and in the development of resistant lines of this breed. For HH, accuracies are still in the low to moderate side and this breed training population needs to be increased before genomic selection could be reliably applied to improve tick resistance.

Catfish is an important aquaculture species in the USA. Columnaris disease is distributed worldwide, affecting a wide variety of fish species including catfish . It leads to huge economic losses each year to the US catfish industry. Channel catfish in general is highly resistant to the disease, while blue catfish is highly susceptible. Genomic selection is an effective and accurate way to predict the breeding values and thus was expected to improve the prediction veracity of columnaris disease resistance in catfish effectively. In this study, two different methods, elastic net genomic best linear unbiased prediction (ENGBLUP) and genomic best linear unbiased prediction (GBLUP), were used to predict the columnaris disease resistance evaluated by binary survival status. Cross-validation showed that the prediction accuracy of ENGBLUP and GBLUP was 0.7347 and 0.4868, respectively, showing that ENGBLUP had a high prediction accuracy. It was shown that fitting QTL and polygenic effect with different distribution will improve genomic prediction accuracy for binary traits. In this study, an accurate and effective genomic selection method was proposed to predict the columnaris resistance in catfish, and its application should be beneficial to catfish breeding.

Genomic selection for meat quality traits in Nelore cattle

Cross-validation of genetic and genomic predictions of temperament in Nellore–Angus crossbreds

The importance of accurate genomic prediction of phenotypes in plant breeding is undeniable, as higher prediction accuracy can increase selection responses. In this regard, epistasis models have shown to be capable of increasing the prediction accuracy while their high computational load is challenging. In this study, we investigated the predictive ability obtained in additive and epistasis models when utilizing haplotype blocks versus pruned sets of SNPs by including phenotypic information from the last growing season. This was done by considering a single biological trait in two growing seasons (2017 and 2018) as separate traits in a multi-trait model. Thus, bivariate variants of the Genomic Best Linear Unbiased Prediction (GBLUP) as an additive model, Epistatic Random Regression BLUP (ERRBLUP) and selective Epistatic Random Regression BLUP (sERRBLUP) as epistasis models were compared with respect to their prediction accuracies for the second year. The prediction accuracies of bivariate GBLUP, ERRBLUP and sERRBLUP were assessed with eight phenotypic traits for 471/402 doubled haploid lines in the European maize landrace Kemater Landmais Gelb/Petkuser Ferdinand Rot. The results indicate that the obtained prediction accuracies are similar when utilizing a pruned set of SNPs or haplotype blocks, while utilizing haplotype blocks reduces the computational load significantly compared to the pruned sets of SNPs. The number of interactions considered in the model was reduced from 323.5/456.4 million for the pruned SNP panel to 4.4/5.5 million in the haplotype block dataset for Kemater and Petkuser landraces, respectively. Since the computational load scales linearly with the number of parameters in the model, this leads to a reduction in computational time of 98.9% from 13.5 hours for the pruned set of markers to 9 minutes for the haplotype block dataset. We further investigated the impact of genomic correlation, phenotypic correlation and trait heritability as factors affecting the bivariate models’ prediction accuracy, identifying the genomic correlation between years as the most influential one. As computational load is substantially reduced, while the accuracy of genomic prediction is unchanged, the here proposed framework to use haplotype blocks in sERRBLUP provided a solution for the practical implementation of sERRBLUP in real breeding programs. Furthermore, our results indicate that sERRBLUP is not only suitable for prediction across different locations, but also for the prediction across growing seasons.

Genotype by environment (G × E) interaction is fundamental in the biology of complex traits and diseases. However, most of the existing methods for genomic prediction tend to ignore G × E interaction (GEI). In this study, we proposed the genomic prediction method G × EBLUP by considering GEI. Meanwhile, G × EBLUP can also detect the genome-wide single nucleotide polymorphisms (SNPs) subject to GEI. Using comprehensive simulations and analysis of real data from pigs and maize, we showed that G × EBLUP achieved higher efficiency in mapping GEI SNPs and higher prediction accuracy than the existing methods, and its superiority was more obvious when the GEI variance was large. For pig and maize real data, compared with GBLUP, G × EBLUP showed improvement by 3% in the prediction accuracy for backfat thickness, while our findings indicated that the trait of days to 100 kg of pig was not affected by GEI and G × EBLUP did not improve the accuracy of genomic prediction for the trait. A significant advantage was observed for G × EBLUP in maize; the prediction accuracy was improved by ∼5.0 and 7.7% for grain weight and water content, respectively. Furthermore, G × EBLUP was not influenced by the number of environment levels. It could determine a favourable environment using SNP Bayes factors for each environment, implying that it is a robust and useful method for market-specific animal and plant breeding. We proposed G × EBLUP, a novel method for the estimation of genomic breeding value by considering GEI. This method identified the genome-wide SNPs that were susceptible to GEI and yielded higher genomic prediction accuracies and lower mean squared error compared with the GBLUP method.

Identification of candidate genes and genomic prediction for early heifer pregnancy in Nelore beef cattle

Heidaritabar, M. (2016). Genomic selection in egg-laying chickens. PhD thesis, Wageningen University, the Netherlands In recent years, prediction of genetic values with DNA markers, or genomic selection (GS), has become a very intense field of research. Many initial studies on GS have focused on the accuracy of predicting the genetic values with different genomic prediction methods. In this thesis, I assessed several aspects of GS. I started with evaluating results of GS against results of traditional pedigree-based selection (BLUP) in data from a selection experiment that applied both methods side by side. The impact of traditional selection and GS on the overall genome variation as well as the overlap between regions selected by GS and the genomic regions predicted to affect the traits were assessed. The impact of selection on genome variation was assessed by measuring changes in allele frequencies that allowed the identification of regions in the genome where changes must be due to selection. These frequency changes were shown to be larger than what could be expected from random fluctuations, indicating that selection is really affecting the allele frequencies and that this effect is stronger in GS compared with BLUP. Next, concordance was tested between the selected regions and regions that affect the traits, as detected by a genome-wide association study. Results showed a low concordance overall between the associated regions and the selected regions. However, markers in associated regions did show larger changes in allele frequencies compared with the average changes across the genome. The selection experiment was performed using a medium density of DNA markers (60K). I subsequently explored the potential benefits of whole-genome sequence data for GS by comparing prediction accuracy from imputed sequence data with the accuracy obtained from the 60K genotypes. Before sequencing, the selection of key animals that should be sequenced to maximize imputation accuracy was assessed with the original 60K genotypes. The accuracy of genotype imputation from lower density panels using a small number of selected key animals as reference was compared with a scenario where random animals were used as the reference population. Even with a very small number of animals as reference, reasonable imputation accuracy could be obtained. Moreover, selecting key animals as reference considerably improved imputation accuracy of rare alleles compared with a set of random reference animals. While imputation from a small reference set was successful, imputation to whole-genome sequence data hardly improved genomic prediction accuracy compared with the predictions based on 60K genotypes. Using only those markers from the whole-genome sequence that are more likely to affect the phenotype was expected to remove noise from the data, but resulted in slightly lower prediction accuracy compared with the complete genome sequence. Finally, I evaluated the inclusion of dominance effects besides additive effects in GS models. The proportion of variance due to additive and dominance effects were estimated for egg production and egg quality traits of a purebred line of layers. The proportion of dominance variance to the total phenotypic variance ranged from 0 to 0.05 across traits. Also, the impact of fitting dominance besides additive effects on prediction accuracy was investigated, but was not found to improve accuracy of genomic prediction of breeding values.

Key messageAccurate prediction of genomic breeding values for Timothy was possible using genomic best linear unbiased prediction.Timothy (Phleum pratense L.) is a grass species of great importance for Finnish agricultural production systems. Genotyping-by-sequencing along with genomic prediction methods offer the possibility to develop breeding materials efficiently. In addition, knowledge about the relationships among traits may be used to increase rates of genetic gain. Still, the quality of the genotypes and the validation population may affect the accuracy of predictions. The objectives of the study were (i) to estimate variance components for yield, winter damage and digestibility traits, and (ii) to assess the accuracy of genomic predictions. Variance components were estimated using genomic residual maximum likelihood where the genomic relationship matrix was scaled using a novel approach. Genomic breeding values were estimated using genomic best linear unbiased prediction in single- and multiple-trait settings, and for different marker filtering criteria. Estimates of heritability ranged from 0.13 ± 0.03 to 0.86 ± 0.05 for yield at first cut and organic matter digestibility at second cut, respectively. Genetic correlations ranged from −0.72 ± 0.12 to 0.59 ± 0.04 between yield at first cut and winter damage, and between digestibility at first and second cuts, respectively. Accuracy of prediction was not severely affected by the quality of genotyping. Using family cross-validation and single-trait models, predictive ability ranged from 0.18 to 0.62 for winter damage and digestibility at second cut, respectively. In addition, validation using forward prediction showed that estimated genomic breeding values were moderately accurate with little dispersion. Thus, genomic prediction constitutes a valuable tool for improving Timothy in Finland.

Biological control is a sustainable strategy to combat agricultural pests. Yet, legislation increasingly restricts importing nonnative biocontrol agents. Thus, selective breeding of biocontrol traits is suggested to enhance performance of existing biocontrol agents. Genomic prediction, where genomic data are used to estimate the genetic merit of an individual for specific traits, is an alternative to exploit genetic variation for the improvement of native biocontrol agents. This study aims to establish a proof of principle for genomic prediction in insect biocontrol agents, using wing morphology traits in the model parasitoid Nasonia vitripennis Walker (Pteromalidae). We performed genomic prediction using a genomic best linear unbiased prediction (GBLUP) model, using 1,230 individuals with 8,639 SNPs generated by genotyping-by-sequencing (GBS). We used individuals from 2 generations from the outbred HVRx population, 717 individuals from generation 169 (G169) and 513 individuals from generation 172 (G172). To assess genomic prediction accuracy, we used across generation validation (forward validation for G172 from G169 and backward validation for G169 from G172) and also 5-fold cross-validation. For size-related traits, including tibia length, wing length, wing width, and second moment area, the accuracy of genomic prediction was close to 0 in both across generation validations but much higher in 5-fold cross-validation (ranging from 0.54 to 0.68). For the shape-related trait wing aspect ratio, a high accuracy was found for all 3 validation strategies, with 0.47 for across generation forward validation (AGFV), 0.65 for across generation backward validation (AGBV), and 0.54 for 5-fold cross-validation. Overall, genomic selection in insect biocontrol agents with a relative small effective population size seems promising. However, factors such as the biology of insects, phenotyping techniques, and large-scale genotyping costs still challenge the application of genomic selection to biocontrol agents.

The approximately 20-year breeding cycle has severely restricted the progress of genetic improvement in poplar. Genomic selection (GS) breeding has been demonstrated as an effective approach to accelerate this process. However, its application in forest tree species remains at an early stage. To advance the genetic improvement of target traits in Populus deltoides, the primary species of poplar plantations in China, we systematically implemented GS breeding using 765 hybrid progenies from 32 multi-generational full-sib families. Firstly, we assembled a high-quality genome of one core parent P. deltoides 'Danhong', with a genome size of 419.4Mb and scaffold N50 of 22.0Mb, which is also the first telomere-to-telomere (T2T) level genome of P. deltoides. Through comparative genomic analysis, we identified 1395 specific structural variants closely associated with growth and development. Subsequently, through genome-wide association studies (GWAS), we identified 135 quantitative trait nucleotides (QTNs) associated with growth and wood quality traits. By systematically evaluating reference genomes, statistical models, and various marker selection strategies, we developed optimal genomic prediction (GP) models for six traits, with the highest prediction accuracy (PA) reaching 0.730 for DBH. Compared with using all markers, the PA was improved by an average of 136.34%. Furthermore, by integrating GP, GWAS, and RNA-seq results, we identified core breeding parents and elite clones for P. deltoides genetic improvement and discovered important candidate genes. Our results provide a promising strategy for accelerating breeding cycles and genetic improvement, offering valuable breeding and genetic resources for forest tree improvement.

BackgroundGenomic selection (GS) may improve selection response over conventional pedigree-based selection if markers capture more detailed information than pedigrees in recently domesticated tree species and/or make it more cost effective. Genomic prediction accuracies using 1748 trees and 6932 SNPs representative of as many distinct gene loci were determined for growth and wood traits in white spruce, within and between environments and breeding groups (BG), each with an effective size of Ne ≈ 20. Marker subsets were also tested.ResultsModel fits and/or cross-validation (CV) prediction accuracies for ridge regression (RR) and the least absolute shrinkage and selection operator models approached those of pedigree-based models. With strong relatedness between CV sets, prediction accuracies for RR within environment and BG were high for wood (r = 0.71–0.79) and moderately high for growth (r = 0.52–0.69) traits, in line with trends in heritabilities. For both classes of traits, these accuracies achieved between 83% and 92% of those obtained with phenotypes and pedigree information. Prediction into untested environments remained moderately high for wood (r ≥ 0.61) but dropped significantly for growth (r ≥ 0.24) traits, emphasizing the need to phenotype in all test environments and model genotype-by-environment interactions for growth traits. Removing relatedness between CV sets sharply decreased prediction accuracies for all traits and subpopulations, falling near zero between BGs with no known shared ancestry. For marker subsets, similar patterns were observed but with lower prediction accuracies.ConclusionsGiven the need for high relatedness between CV sets to obtain good prediction accuracies, we recommend to build GS models for prediction within the same breeding population only. Breeding groups could be merged to build genomic prediction models as long as the total effective population size does not exceed 50 individuals in order to obtain high prediction accuracy such as that obtained in the present study. A number of markers limited to a few hundred would not negatively impact prediction accuracies, but these could decrease more rapidly over generations. The most promising short-term approach for genomic selection would likely be the selection of superior individuals within large full-sib families vegetatively propagated to implement multiclonal forestry.