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

Genomic dissection of the correlation between milk yield and various health traits using functional and evolutionary information about imputed sequence variants of 34,497 German Holstein cows

BACKGROUND

Over the last decades, it was subject of many studies to investigate the genomic connection of milk production and health traits in dairy cattle. Thereby, incorporating functional information in genomic analyses has been shown to improve the understanding of biological and molecular mechanisms shaping complex traits and the accuracies of genomic prediction, especially in small populations and across-breed settings. Still, little is known about the contribution of different functional and evolutionary genome partitioning subsets to milk production and dairy health. Thus, we performed a uni- and a bivariate analysis of milk yield (MY) and eight health traits using a set of ~34,497 German Holstein cows with 50K chip genotypes and ~17 million imputed sequence variants divided into 27 subsets depending on their functional and evolutionary annotation. In the bivariate analysis, eight trait-combinations were observed that contrasted MY with each health trait. Two genomic relationship matrices (GRM) were included, one consisting of the 50K chip variants and one consisting of each set of subset variants, to obtain subset heritabilities and genetic correlations. In addition, 50K chip heritabilities and genetic correlations were estimated applying merely the 50K GRM.

RESULTS

In general, 50K chip heritabilities were larger than the subset heritabilities. The largest heritabilities were found for MY, which was 0.4358 for the 50K and 0.2757 for the subset heritabilities. Whereas all 50K genetic correlations were negative, subset genetic correlations were both, positive and negative (ranging from -0.9324 between MY and mastitis to 0.6662 between MY and digital dermatitis). The subsets containing variants which were annotated as noncoding related, splice sites, untranslated regions, metabolic quantitative trait loci, and young variants ranked highest in terms of their contribution to the traits` genetic variance. We were able to show that linkage disequilibrium between subset variants and adjacent variants did not cause these subsets` high effect.

CONCLUSION

Our results confirm the connection of milk production and health traits in dairy cattle via the animals` metabolic state. In addition, they highlight the potential of including functional information in genomic analyses, which helps to dissect the extent and direction of the observed traits` connection in more detail.

Increased Accuracy of Genomic Prediction Using Preselected SNPs from GWAS with Imputed Whole-Genome Sequence Data in Pigs

Enhancing the accuracy of genomic prediction is a key goal in genomic selection (GS) research. Integrating prior biological information into GS methods using appropriate models can improve prediction accuracy for complex traits. Genome-wide association study (GWAS) is widely utilized to identify potential candidate loci associated with complex traits in livestock and poultry, offering essential genomic insights. In this study, a GWAS was conducted on 685 Duroc × Landrace × Yorkshire (DLY) pigs to extract significant single-nucleotide polymorphisms (SNPs) as genomic features. We compared two GS models, genomic best linear unbiased prediction (GBLUP) and genomic feature BLUP (GFBLUP), by using imputed whole-genome sequencing (WGS) data on 651 Yorkshire pigs. The results revealed that the GBLUP model achieved prediction accuracies of 0.499 for backfat thickness (BFT) and 0.423 for loin muscle area (LMA). By applying the GFBLUP model with GWAS-based SNP preselection, the average prediction accuracies for BFT and LMA traits reached 0.491 and 0.440, respectively. Specifically, the GFBLUP model displayed a 4.8% enhancement in predicting LMA compared to the GBLUP model. These findings suggest that, in certain scenarios, the GFBLUP model may offer superior genomic prediction accuracy when compared to the GBLUP model, underscoring the potential value of incorporating genomic features to refine GS models.

Alternative SNP weighting for multi-step and single-step genomic BLUP in the presence of causative variants

The accuracy of genetic selection in dairy can be increased by the adoption of new technologies, such as the inclusion of sequence data. In simulation studies, assigning different weights to causative single-nucleotide polymorphism (SNP) markers led to better predictions depending on the genomic prediction method used. However, it is still not clear how the weights should be calculated. Our objective was to evaluate the accuracy of a multi-step method (GBLUP) and single-step GBLUP with simulated data using regular SNP, causatives variants (QTN) and the combination of both. Additionally, we compared the accuracies of all previous scenarios using alternatives for SNP weighting. The data were simulated assuming a single trait with a heritability of 0.3. The effective population size (Ne) was approximately 200. The pedigree contained 440,000 animals, and approximately 16,800 individuals were genotyped. A total of 49,974 SNP markers were evenly placed throughout the genome, and 100, 1000 and 2000 causative QTN were simulated. Both GBLUP and ssGBLUP were used in this study. We evaluated quadratic and nonlinear SNP weights in addition to the unweighted G. The inclusion of QTN to panels led to significant accuracy gains. Nonlinear A was demonstrated to be superior to quadratic weighting and unweighted approaches; however, results from Nonlinear A were dependent on the equation parameters. The unweighted approach was more suitable for less polygenic scenarios. Finally, SNP weighting might help elucidate trait architecture features based on changes in the accuracy of genomic prediction.

The long-term effects of genomic selection: 2. Changes in allele frequencies of causal loci and new mutations

Genetic selection has been applied for many generations in animal, plant, and experimental populations. Selection changes the allelic architecture of traits to create genetic gain. It remains unknown whether the changes in allelic architecture are different for the recently introduced technique of genomic selection compared to traditional selection methods and whether they depend on the genetic architectures of traits. Here, we investigate the allele frequency changes of old and new causal loci under 50 generations of phenotypic, pedigree, and genomic selection, for a trait controlled by either additive, additive and dominance, or additive, dominance, and epistatic effects. Genomic selection resulted in slightly larger and faster changes in allele frequencies of causal loci than pedigree selection. For each locus, allele frequency change per generation was not only influenced by its statistical additive effect but also to a large extent by the linkage phase with other loci and its allele frequency. Selection fixed a large number of loci, and 5 times more unfavorable alleles became fixed with genomic and pedigree selection than with phenotypic selection. For pedigree selection, this was mainly a result of increased genetic drift, while genetic hitchhiking had a larger effect on genomic selection. When epistasis was present, the average allele frequency change was smaller (∼15% lower), and a lower number of loci became fixed for all selection methods. We conclude that for long-term genetic improvement using genomic selection, it is important to consider hitchhiking and to limit the loss of favorable alleles.

Improving the accuracy of genomic prediction for meat quality traits using whole genome sequence data in pigs

BACKGROUND

Pork quality can directly affect customer purchase tendency and meat quality traits have become valuable in modern pork production. However, genetic improvement has been slow due to high phenotyping costs. In this study, whole genome sequence (WGS) data was used to evaluate the prediction accuracy of genomic best linear unbiased prediction (GBLUP) for meat quality in large-scale crossbred commercial pigs.

RESULTS

We produced WGS data (18,695,907 SNPs and 2,106,902 INDELs exceed quality control) from 1,469 sequenced Duroc × (Landrace × Yorkshire) pigs and developed a reference panel for meat quality including meat color score, marbling score, L* (lightness), a* (redness), and b* (yellowness) of genomic prediction. The prediction accuracy was defined as the Pearson correlation coefficient between adjusted phenotypes and genomic estimated breeding values in the validation population. Using different marker density panels derived from WGS data, accuracy differed substantially among meat quality traits, varied from 0.08 to 0.47. Results showed that MultiBLUP outperform GBLUP and yielded accuracy increases ranging from 17.39% to 75%. We optimized the marker density and found medium- and high-density marker panels are beneficial for the estimation of heritability for meat quality. Moreover, we conducted genotype imputation from 50K chip to WGS level in the same population and found average concordance rate to exceed 95% and r² = 0.81.

CONCLUSIONS

Overall, estimation of heritability for meat quality traits can benefit from the use of WGS data. This study showed the superiority of using WGS data to genetically improve pork quality in genomic prediction.

Variance component estimates for growth traits in beef cattle using selected variants from imputed low-pass sequence data

A beef cattle population (n = 2,343) was used to assess the impact of variants identified from the imputed low-pass sequence (LPS) on the estimation of variance components and genetic parameters of birth weight (BWT) and post-weaning gain (PWG). Variants were selected based on functional impact and were partitioned into four groups (low, modifier, moderate, high) based on predicted functional impact and re-partitioned based on the consequence of mutation, such as missense and untranslated region variants, into six groups (G1-G6). Each subset was used to construct a genomic relationship matrix (GRM) for univariate animal models. Multiple analyses were conducted to compare the proportion of additive genetic variation explained by the different subsets individually and collectively, and these estimates were benchmarked against all LPS variants in a single GRM and array (e.g., GeneSeek Genomic Profiler 100K) genotypes. When all variants were included in a single GRM, heritability estimates for BWT and PWG were 0.43 ± 0.05 and 0.38 ± 0.05, respectively. Heritability estimates for BWT ranged from 0.10 to 0.42 dependent on which variant subsets were included. Similarly, estimates for PWG ranged from 0.05 to 0.38. Results showed that variants in the subsets modifier and G1 (untranslated region) yielded the highest heritability estimates and were similar to the inclusion of all variants, while estimates from GRM containing only variants in the categories High, G4 (non-coding transcript exon), and G6 (start and stop loss/gain) were the lowest. All variants combined provided similar heritability estimates to chip genotypes and provided minimal to no additional information when combined with chip data. This suggests that the chip single nucleotide polymorphisms and the variants from LPS predicted to be less consequential are in relatively high linkage disequilibrium with the underlying causal variants as a whole and sufficiently spread throughout the genome to capture larger proportions of additive genetic variation.

Genomic prediction in the wild: A case study in Soay sheep

Genomic prediction, the technique whereby an individual's genetic component of their phenotype is estimated from its genome, has revolutionised animal and plant breeding and medical genetics. However, despite being first introduced nearly two decades ago, it has hardly been adopted by the evolutionary genetics community studying wild organisms. Here, genomic prediction is performed on eight traits in a wild population of Soay sheep. The population has been the focus of a >30 year evolutionary ecology study and there is already considerable understanding of the genetic architecture of the focal Mendelian and quantitative traits. We show that the accuracy of genomic prediction is high for all traits, but especially those with loci of large effect segregating. Five different methods are compared, and the two methods that can accommodate zero-effect and large-effect loci in the same model tend to perform best. If the accuracy of genomic prediction is similar in other wild populations, then there is a real opportunity for pedigree-free molecular quantitative genetics research to be enabled in many more wild populations; currently the literature is dominated by studies that have required decades of field data collection to generate sufficiently deep pedigrees. Finally, some of the potential applications of genomic prediction in wild populations are discussed.

Progress and opportunities through use of genomics in animal production

The rearing of farmed animals is a vital component of global food production systems, but its impact on the environment, human health, animal welfare, and biodiversity is being increasingly challenged. Developments in genetic and genomic technologies have had a key role in improving the productivity of farmed animals for decades. Advances in genome sequencing, annotation, and editing offer a means not only to continue that trend, but also, when combined with advanced data collection, analytics, cloud computing, appropriate infrastructure, and regulation, to take precision livestock farming (PLF) and conservation to an advanced level. Such an approach could generate substantial additional benefits in terms of reducing use of resources, health treatments, and environmental impact, while also improving animal health and welfare.

Genomic prediction with whole-genome sequence data in intensely selected pig lines

Background

Early simulations indicated that whole-genome sequence data (WGS) could improve the accuracy of genomic predictions within and across breeds. However, empirical results have been ambiguous so far. Large datasets that capture most of the genomic diversity in a population must be assembled so that allele substitution effects are estimated with high accuracy. The objectives of this study were to use a large pig dataset from seven intensely selected lines to assess the benefits of using WGS for genomic prediction compared to using commercial marker arrays and to identify scenarios in which WGS provides the largest advantage.

Methods

We sequenced 6931 individuals from seven commercial pig lines with different numerical sizes. Genotypes of 32.8 million variants were imputed for 396,100 individuals (17,224 to 104,661 per line). We used BayesR to perform genomic prediction for eight complex traits. Genomic predictions were performed using either data from a standard marker array or variants preselected from WGS based on association tests.

Results

The accuracies of genomic predictions based on preselected WGS variants were not robust across traits and lines and the improvements in prediction accuracy that we achieved so far with WGS compared to standard marker arrays were generally small. The most favourable results for WGS were obtained when the largest training sets were available and standard marker arrays were augmented with preselected variants with statistically significant associations to the trait. With this method and training sets of around 80k individuals, the accuracy of within-line genomic predictions was on average improved by 0.025. With multi-line training sets, improvements of 0.04 compared to marker arrays could be expected.

Conclusions

Our results showed that WGS has limited potential to improve the accuracy of genomic predictions compared to marker arrays in intensely selected pig lines. Thus, although we expect that larger improvements in accuracy from the use of WGS are possible with a combination of larger training sets and optimised pipelines for generating and analysing such datasets, the use of WGS in the current implementations of genomic prediction should be carefully evaluated against the cost of large-scale WGS data on a case-by-case basis.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12711-022-00756-0.

Genomic Selection in Chinese Holsteins Using Regularized Regression Models for Feature Selection of Whole Genome Sequencing Data

Simple Summary

Genomic selection (GS) is increasingly widely used in animal breeding, owing to its high efficiency in the genetic improvement of economic traits. In China, GS has been implemented for genetic evaluation of young bulls in dairy cattle breeding programs since 2012. GS is commonly based on single nucleotide polymorphism (SNP) chips. The cost of whole genome sequencing (WGS) has decreased tremendously in recent years, allowing increased studies of WGS-based GS. In this study, based on the imputed WGS data of approximately 8000 Chinese Holsteins, we investigated the performance of GS of milk production traits using the feature selection method of regularized regression. The results showed that WGS-based GS using regularized regression models and the commonly used linear mixed models achieved comparable prediction accuracies. For milk and protein yields, GS using a combination of SNPs selected with a regularized regression model and 50K SNP chip data achieved the best prediction performance, and GS using SNPs selected with a linear mixed model combined with 50K SNP chip data performed best for fat yield. The proposed method of GS based on WGS data, i.e., feature selection using regularization regression models, provides a valuable novel strategy for genomic selection.

Abstract

Genomic selection (GS) is an efficient method to improve genetically economic traits. Feature selection is an important method for GS based on whole-genome sequencing (WGS) data. We investigated the prediction performance of GS of milk production traits using imputed WGS data on 7957 Chinese Holsteins. We used two regularized regression models, least absolute shrinkage and selection operator (LASSO) and elastic net (EN) for feature selection. For comparison, we performed genome-wide association studies based on a linear mixed model (LMM), and the N single nucleotide polymorphisms (SNPs) with the lowest p-values were selected (LMM_LASSO and LMM_EN), where N was the number of non-zero effect SNPs selected by LASSO or EN. GS was conducted using a genomic best linear unbiased prediction (GBLUP) model and several sets of SNPs: (1) selected WGS SNPs; (2) 50K SNP chip data; (3) WGS data; and (4) a combined set of selected WGS SNPs and 50K SNP chip data. The results showed that the prediction accuracies of GS with features selected using LASSO or EN were comparable to those using features selected with LMM_LASSO or LMM_EN. For milk and protein yields, GS using a combination of SNPs selected with LASSO and 50K SNP chip data achieved the best prediction performance, and GS using SNPs selected with LMM_LASSO combined with 50K SNP chip data performed best for fat yield. The proposed method, feature selection using regularization regression models, provides a valuable novel strategy for WGS-based GS.

Assessing the genomic diversity and relatedness in 10 Canadian heritage chicken lines using whole-genome sequence data

In the past 50 years, there has been a steep increase in the demand for poultry products, met by increasing production along with genetic selection for improved growth, efficiency, health and reproduction. The selection tends to reduce the number and type of genetic resources contributing to the majority of production. The University of Alberta maintains 10 heritage chicken lines (Brown Leghorn (BL), Light Sussex (LS), New Hampshire (NH), Saskatchewan Barred Rock (SaskBR), Shaver Barred Rock (ShaverBR), Shaver Rhode Island Red (RIR), White Leghorn (WL) and three commercial crosses called Alberta Meat Control strains 1957 (AMC-1957), 1978 sire line (AMC-1978-20S) and 1978 dam line (AMC-1978-30D), that played a large role in the evolution of the poultry industry in Canada. Since these lines have not been subjected to the same intensive selection pressures as commercial counterparts, they may contain unique genetic variants lost in commercial lines. Thus, for conservation management of these lines, the first step is to assess their genetic diversity. 71 male samples from across 10 lines were analysed using whole-genome sequencing and patterns of genetic diversity and relatedness among these lines were explored. AMC-1978-30D showed the highest genetic diversity as reflected in observed and expected heterozygosity (0.327 and 0.250), percentage of polymorphic markers (~ 65%) and average recent inbreeding coefficient (-0.039), followed by AMC-1978-20S and AMC-1957. BL showed the lowest genetic diversity as reflected in observed and expected heterozygosity (0.130 and 0.116), percentage of polymorphic markers (~31%) and average recent inbreeding coefficient (0.577), followed by LS, WL and NH. Our findings highlight the need for special attention for the populations of BL, WL, LS and NH, with the largest levels of inbreeding. Our results can be used to develop a breeding strategy to optimize and conserve the genetic variation present in heritage lines.

Genome-wide association studies for egg quality traits in White Leghorn layers using low-pass sequencing and SNP chip data

Low-pass sequencing data have been proposed as an alternative to single nucleotide polymorphism (SNP) chips in genome-wide association studies (GWAS) of several species. However, it has not been used in layer chickens yet. This study aims at comparing the GWAS results of White Leghorn chickens using low-pass sequencing data (1×) and 54 k SNP chip data. Ten commercially relevant egg quality traits including albumen height, shell strength, shell colour, egg weight and yolk weight collected from up to 1,420 White Leghorn chickens were analysed. The results showed that the genomic heritability estimates based on low-pass sequencing data were higher than those based on SNP chip data. Although two GWAS analyses showed similar overall landscape for most traits, low-pass sequencing captured some significant SNPs that were not on the SNP chip. In GWAS analysis using 54 k SNP chip data, after including more individuals (up to 5,700), additional significant SNPs not detected by low-pass sequencing data were found. In conclusion, GWAS using low-pass sequencing data showed similar results to those with SNP chip data and may require much larger sample sizes to show measurable advantages.

Improvement of prediction ability by integrating multi-omic datasets in barley

Background

Genomic prediction (GP) based on single nucleotide polymorphisms (SNP) has become a broadly used tool to increase the gain of selection in plant breeding. However, using predictors that are biologically closer to the phenotypes such as transcriptome and metabolome may increase the prediction ability in GP. The objectives of this study were to (i) assess the prediction ability for three yield-related phenotypic traits using different omic datasets as single predictors compared to a SNP array, where these omic datasets included different types of sequence variants (full-SV, deleterious-dSV, and tolerant-tSV), different types of transcriptome (expression presence/absence variation-ePAV, gene expression-GE, and transcript expression-TE) sampled from two tissues, leaf and seedling, and metabolites (M); (ii) investigate the improvement in prediction ability when combining multiple omic datasets information to predict phenotypic variation in barley breeding programs; (iii) explore the predictive performance when using SV, GE, and ePAV from simulated 3’end mRNA sequencing of different lengths as predictors.

Results

The prediction ability from genomic best linear unbiased prediction (GBLUP) for the three traits using dSV information was higher than when using tSV, all SV information, or the SNP array. Any predictors from the transcriptome (GE, TE, as well as ePAV) and metabolome provided higher prediction abilities compared to the SNP array and SV on average across the three traits. In addition, some (di)-similarity existed between different omic datasets, and therefore provided complementary biological perspectives to phenotypic variation. Optimal combining the information of dSV, TE, ePAV, as well as metabolites into GP models could improve the prediction ability over that of the single predictors alone.

Conclusions

The use of integrated omic datasets in GP model is highly recommended. Furthermore, we evaluated a cost-effective approach generating 3’end mRNA sequencing with transcriptome data extracted from seedling without losing prediction ability in comparison to the full-length mRNA sequencing, paving the path for the use of such prediction methods in commercial breeding programs.

Supplementary Information

The online version contains supplementary material available at (10.1186/s12864-022-08337-7).

The long-term effects of genomic selection: 1. Response to selection, additive genetic variance, and genetic architecture

Background

Genomic selection has revolutionized genetic improvement in animals and plants, but little is known about its long-term effects. Here, we investigated the long-term effects of genomic selection on response to selection, genetic variance, and the genetic architecture of traits using stochastic simulations. We defined the genetic architecture as the set of causal loci underlying each trait, their allele frequencies, and their statistical additive effects. We simulated a livestock population under 50 generations of phenotypic, pedigree, or genomic selection for a single trait, controlled by either only additive, additive and dominance, or additive, dominance, and epistatic effects. The simulated epistasis was based on yeast data.

Results

Short-term response was always greatest with genomic selection, while response after 50 generations was greater with phenotypic selection than with genomic selection when epistasis was present, and was always greater than with pedigree selection. This was mainly because loss of genetic variance and of segregating loci was much greater with genomic and pedigree selection than with phenotypic selection. Compared to pedigree selection, selection response was always greater with genomic selection. Pedigree and genomic selection lost a similar amount of genetic variance after 50 generations of selection, but genomic selection maintained more segregating loci, which on average had lower minor allele frequencies than with pedigree selection. Based on this result, genomic selection is expected to better maintain genetic gain after 50 generations than pedigree selection. The amount of change in the genetic architecture of traits was considerable across generations and was similar for genomic and pedigree selection, but slightly less for phenotypic selection. Presence of epistasis resulted in smaller changes in allele frequencies and less fixation of causal loci, but resulted in substantial changes in statistical additive effects across generations.

Conclusions

Our results show that genomic selection outperforms pedigree selection in terms of long-term genetic gain, but results in a similar reduction of genetic variance. The genetic architecture of traits changed considerably across generations, especially under selection and when non-additive effects were present. In conclusion, non-additive effects had a substantial impact on the accuracy of selection and long-term response to selection, especially when selection was accurate.

Functionally prioritised whole-genome sequence variants improve the accuracy of genomic prediction for heat tolerance

BACKGROUND

Heat tolerance is a trait of economic importance in the context of warm climates and the effects of global warming on livestock production, reproduction, health, and well-being. This study investigated the improvement in prediction accuracy for heat tolerance when selected sets of sequence variants from a large genome-wide association study (GWAS) were combined with a standard 50k single nucleotide polymorphism (SNP) panel used by the dairy industry.

METHODS

Over 40,000 dairy cattle with genotype and phenotype data were analysed. The phenotypes used to measure an individual's heat tolerance were defined as the rate of decline in milk production traits with rising temperature and humidity. We used Holstein and Jersey cows to select sequence variants linked to heat tolerance. The prioritised sequence variants were the most significant SNPs passing a GWAS p-value threshold selected based on sliding 100-kb windows along each chromosome. We used a bull reference set to develop the genomic prediction equations, which were then validated in an independent set of Holstein, Jersey, and crossbred cows. Prediction analyses were performed using the BayesR, BayesRC, and GBLUP methods.

RESULTS

The accuracy of genomic prediction for heat tolerance improved by up to 0.07, 0.05, and 0.10 units in Holstein, Jersey, and crossbred cows, respectively, when sets of selected sequence markers from Holstein cows were added to the 50k SNP panel. However, in some scenarios, the prediction accuracy decreased unexpectedly with the largest drop of - 0.10 units for the heat tolerance fat yield trait observed in Jersey cows when 50k plus pre-selected SNPs from Holstein cows were used. Using pre-selected SNPs discovered on a combined set of Holstein and Jersey cows generally improved the accuracy, especially in the Jersey validation. In addition, combining Holstein and Jersey bulls in the reference set generally improved prediction accuracy in most scenarios compared to using only Holstein bulls as the reference set.

CONCLUSIONS

Informative sequence markers can be prioritised to improve the genomic prediction of heat tolerance in different breeds. In addition to providing biological insight, these variants could also have a direct application for developing customized SNP arrays or can be used via imputation in current industry SNP panels.

The accuracy of genomic prediction for meat quality traits in Hanwoo cattle when using genotypes from different SNP densities and preselected variants from imputed whole genome sequence

Context Genomic prediction is the use of genomic data in the estimation of genomic breeding values (GEBV) in animal breeding. In beef cattle breeding programs, genomic prediction increases the rates of genetic gain by increasing the accuracy of selection at earlier ages. Aims The objectives of the study were to examine the effect of single-nucleotide polymorphism (SNP) density and to evaluate the effect of using SNPs preselected from imputed whole-genome sequence for genomic prediction. Methods Genomic and phenotypic data from 2110 Hanwoo steers were used to predict GEBV for marbling score (MS), meat texture (MT), and meat colour (MC) traits. Three types of SNP densities including 50k, high-density (HD), and whole-genome sequence data and preselected SNPs from genome-wide association study (GWAS) were used for genomic prediction analyses. Two scenarios (independent and dependent discovery populations) were used to select top significant SNPs. The accuracy of GEBV was assessed using random cross-validation. Genomic best linear unbiased prediction (GBLUP) was used to predict the breeding values for each trait. Key results Our result showed that very similar prediction accuracies were observed across all SNP densities used in the study. The prediction accuracy among traits ranged from 0.29±0.05 for MC to 0.46±0.04 for MS. Depending on the studied traits, up to 5% of prediction accuracy improvement was obtained when the preselected SNPs from GWAS analysis were included in the prediction analysis. Conclusions High SNP density such as HD and the whole-genome sequence data yielded a similar prediction accuracy in Hanwoo beef cattle. Therefore, the 50K SNP chip panel is sufficient to capture the relationships in a breed with a small effective population size such as the Hanwoo cattle population. Preselected variants improved prediction accuracy when they were included in the genomic prediction model. Implications The estimated genomic prediction accuracies are moderately accurate in Hanwoo cattle and for searching for SNPs that are more productive could increase the accuracy of estimated breeding values for the studied traits.

The construction of a haplotype reference panel using extremely low coverage whole genome sequences and its application in genome-wide association studies and genomic prediction in Duroc pigs

Extremely low coverage whole genome sequencing (lcWGS) is an economical technique to obtain high-density single nucleotide polymorphisms (SNPs). Here, we explored the feasibility of constructing a haplotype reference panel (lcHRP) using lcWGS and evaluated the effects of lcHRP through a genome-wide association study (GWAS) and genomic prediction in pigs. A total of 297 and 974 Duroc pigs were genotyped using lcWGS and a 50 K SNP array, respectively. We obtained 19,306,498 SNPs using lcWGS with an accuracy of 0.984. With the help of lcHRP, the accuracy of imputation from the SNP array to lcWGS was 0.922. Compared to the SNP array findings, those from the imputation-based GWAS identified more signals across four traits. With the integration of the top 1% imputation-based GWAS findings as genomic features, the accuracies of genomic prediction was improved by 6.0% to 13.2%. This study showed the great potential of lcWGS in pigs' molecular breeding.

Accuracies of genomic predictions for disease resistance of striped catfish to Edwardsiella ictaluri using artificial intelligence algorithms

Assessments of genomic prediction accuracies using artificial intelligent (AI) algorithms (i.e., machine and deep learning methods) are currently not available or very limited in aquaculture species. The principal aim of this study was to examine the predictive performance of these new methods for disease resistance to Edwardsiella ictaluri in a population of striped catfish Pangasianodon hypophthalmus and to make comparisons with four common methods, i.e., pedigree-based best linear unbiased prediction (PBLUP), genomic-based best linear unbiased prediction (GBLUP), single-step GBLUP (ssGBLUP) and a nonlinear Bayesian approach (notably BayesR). Our analyses using machine learning (i.e., ML-KAML) and deep learning (i.e., DL-MLP and DL-CNN) together with the four common methods (PBLUP, GBLUP, ssGBLUP, and BayesR) were conducted for two main disease resistance traits (i.e., survival status coded as 0 and 1 and survival time, i.e., days that the animals were still alive after the challenge test) in a pedigree consisting of 560 individual animals (490 offspring and 70 parents) genotyped for 14,154 single nucleotide polymorphism (SNPs). The results using 6,470 SNPs after quality control showed that machine learning methods outperformed PBLUP, GBLUP, and ssGBLUP, with the increases in the prediction accuracies for both traits by 9.1–15.4%. However, the prediction accuracies obtained from machine learning methods were comparable to those estimated using BayesR. Imputation of missing genotypes using AlphaFamImpute increased the prediction accuracies by 5.3–19.2% in all the methods and data used. On the other hand, there were insignificant decreases (0.3–5.6%) in the prediction accuracies for both survival status and survival time when multivariate models were used in comparison to univariate analyses. Interestingly, the genomic prediction accuracies based on only highly significant SNPs (P < 0.00001, 318–400 SNPs for survival status and 1,362–1,589 SNPs for survival time) were somewhat lower (0.3–15.6%) than those obtained from the whole set of 6,470 SNPs. In most of our analyses, the accuracies of genomic prediction were somewhat higher for survival time than survival status (0/1 data). It is concluded that although there are prospects for the application of genomic selection to increase disease resistance to E. ictaluri in striped catfish breeding programs, further evaluation of these methods should be made in independent families/populations when more data are accumulated in future generations to avoid possible biases in the genetic parameters estimates and prediction accuracies for the disease-resistant traits studied in this population of striped catfish P. hypophthalmus.

Dissection of the impact of prioritized QTL-linked and -unlinked SNP markers on the accuracy of genomic selection1

Background

Use of genomic information has resulted in an undeniable improvement in prediction accuracies and an increase in genetic gain in animal and plant genetic selection programs in spite of oversimplified assumptions about the true biological processes. Even for complex traits, a large portion of markers do not segregate with or effectively track genomic regions contributing to trait variation; yet it is not clear how genomic prediction accuracies are impacted by such potentially nonrelevant markers. In this study, a simulation was carried out to evaluate genomic predictions in the presence of markers unlinked with trait-relevant QTL. Further, we compared the ability of the population statistic F_ST and absolute estimated marker effect as preselection statistics to discriminate between linked and unlinked markers and the corresponding impact on accuracy.

Results

We found that the accuracy of genomic predictions decreased as the proportion of unlinked markers used to calculate the genomic relationships increased. Using all, only linked, and only unlinked marker sets yielded prediction accuracies of 0.62, 0.89, and 0.22, respectively. Furthermore, it was found that prediction accuracies are severely impacted by unlinked markers with large spurious associations. F_ST-preselected marker sets of 10 k and larger yielded accuracies 8.97 to 17.91% higher than those achieved using preselection by absolute estimated marker effects, despite selecting 5.1 to 37.7% more unlinked markers and explaining 2.4 to 5.0% less of the genetic variance. This was attributed to false positives selected by absolute estimated marker effects having a larger spurious association with the trait of interest and more negative impact on predictions. The Pearson correlation between F_ST scores and absolute estimated marker effects was 0.77 and 0.27 among only linked and only unlinked markers, respectively. The sensitivity of F_ST scores to detect truly linked markers is comparable to absolute estimated marker effects but the consistency between the two statistics regarding false positives is weak.

Conclusion

Identification and exclusion of markers that have little to no relevance to the trait of interest may significantly increase genomic prediction accuracies. The population statistic F_ST presents an efficient and effective tool for preselection of trait-relevant markers.

Genomic Prediction Using LD-Based Haplotypes Inferred From High-Density Chip and Imputed Sequence Variants in Chinese Simmental Beef Cattle

A haplotype is defined as a combination of alleles at adjacent loci belonging to the same chromosome that can be transmitted as a unit. In this study, we used both the Illumina BovineHD chip (HD chip) and imputed whole-genome sequence (WGS) data to explore haploblocks and assess haplotype effects, and the haploblocks were defined based on the different LD thresholds. The accuracies of genomic prediction (GP) for dressing percentage (DP), meat percentage (MP), and rib eye roll weight (RERW) based on haplotype were investigated and compared for both data sets in Chinese Simmental beef cattle. The accuracies of GP using the entire imputed WGS data were lower than those using the HD chip data in all cases. For DP and MP, the accuracy of GP using haploblock approaches outperformed the individual single nucleotide polymorphism (SNP) approach (GBLUP_In_Block) at specific LD levels. Hotelling’s test confirmed that GP using LD-based haplotypes from WGS data can significantly increase the accuracies of GP for RERW, compared with the individual SNP approach (∼1.4 and 1.9% for G_HBLUP and G_HBLUP+GBLUP, respectively). We found that the accuracies using haploblock approach varied with different LD thresholds. The LD thresholds (r² ≥ 0.5) were optimal for most scenarios. Our results suggested that LD-based haploblock approach can improve accuracy of genomic prediction for carcass traits using both HD chip and imputed WGS data under the optimal LD thresholds in Chinese Simmental beef cattle.

Perspectives on Applications of Hierarchical Gene-To-Phenotype (G2P) Maps to Capture Non-stationary Effects of Alleles in Genomic Prediction

Genomic prediction of complex traits across environments, breeding cycles, and populations remains a challenge for plant breeding. A potential explanation for this is that underlying non-additive genetic (GxG) and genotype-by-environment (GxE) interactions generate allele substitution effects that are non-stationary across different contexts. Such non-stationary effects of alleles are either ignored or assumed to be implicitly captured by most gene-to-phenotype (G2P) maps used in genomic prediction. The implicit capture of non-stationary effects of alleles requires the G2P map to be re-estimated across different contexts. We discuss the development and application of hierarchical G2P maps that explicitly capture non-stationary effects of alleles and have successfully increased short-term prediction accuracy in plant breeding. These hierarchical G2P maps achieve increases in prediction accuracy by allowing intermediate processes such as other traits and environmental factors and their interactions to contribute to complex trait variation. However, long-term prediction remains a challenge. The plant breeding community should undertake complementary simulation and empirical experiments to interrogate various hierarchical G2P maps that connect GxG and GxE interactions simultaneously. The existing genetic correlation framework can be used to assess the magnitude of non-stationary effects of alleles and the predictive ability of these hierarchical G2P maps in long-term, multi-context genomic predictions of complex traits in plant breeding.

Increased accuracy of genomic predictions for growth under chronic thermal stress in rainbow trout by prioritizing variants from GWAS using imputed sequence data

Through imputation of genotypes, genome‐wide association study (GWAS) and genomic prediction (GP) using whole‐genome sequencing (WGS) data are cost‐efficient and feasible in aquaculture breeding schemes. The objective was to dissect the genetic architecture of growth traits under chronic heat stress in rainbow trout (Oncorhynchus mykiss) and to assess the accuracy of GP based on imputed WGS and different preselected single nucleotide polymorphism (SNP) arrays. A total of 192 and 764 fish challenged to a heat stress experiment for 62 days were genotyped using a customized 1 K and 26 K SNP panels, respectively, and then, genotype imputation was performed from a low‐density chip to WGS using 102 parents (36 males and 66 females) as the reference population. Imputed WGS data were used to perform GWAS and test GP accuracy under different preselected SNP scenarios. Heritability was estimated for body weight (BW), body length (BL) and average daily gain (ADG). Estimates using imputed WGS data ranged from 0.33 ± 0.05 to 0.55 ± 0.05 for growth traits under chronic heat stress. GWAS revealed that the top five cumulatively SNPs explained a maximum of 0.94%, 0.86% and 0.51% of genetic variance for BW, BL and ADG, respectively. Some important functional candidate genes associated with growth‐related traits were found among the most important SNPs, including signal transducer and activator of transcription 5B and 3 (STAT5B and STAT3, respectively) and cytokine‐inducible SH2‐containing protein (CISH). WGS data resulted in a slight increase in prediction accuracy compared with pedigree‐based method, whereas preselected SNPs based on the top GWAS hits improved prediction accuracies, with values ranging from 1.2 to 13.3%. Our results support the evidence of the polygenic nature of growth traits when measured under heat stress. The accuracies of GP can be improved using preselected variants from GWAS, and the use of WGS marginally increases prediction accuracy.

How imputation can mitigate SNP ascertainment Bias

Background

Population genetic studies based on genotyped single nucleotide polymorphisms (SNPs) are influenced by a non-random selection of the SNPs included in the used genotyping arrays. The resulting bias in the estimation of allele frequency spectra and population genetics parameters like heterozygosity and genetic distances relative to whole genome sequencing (WGS) data is known as SNP ascertainment bias. Full correction for this bias requires detailed knowledge of the array design process, which is often not available in practice. This study suggests an alternative approach to mitigate ascertainment bias of a large set of genotyped individuals by using information of a small set of sequenced individuals via imputation without the need for prior knowledge on the array design.

Results

The strategy was first tested by simulating additional ascertainment bias with a set of 1566 chickens from 74 populations that were genotyped for the positions of the Affymetrix Axiom™ 580 k Genome-Wide Chicken Array. Imputation accuracy was shown to be consistently higher for populations used for SNP discovery during the simulated array design process. Reference sets of at least one individual per population in the study set led to a strong correction of ascertainment bias for estimates of expected and observed heterozygosity, Wright’s Fixation Index and Nei’s Standard Genetic Distance. In contrast, unbalanced reference sets (overrepresentation of populations compared to the study set) introduced a new bias towards the reference populations. Finally, the array genotypes were imputed to WGS by utilization of reference sets of 74 individuals (one per population) to 98 individuals (additional commercial chickens) and compared with a mixture of individually and pooled sequenced populations. The imputation reduced the slope between heterozygosity estimates of array data and WGS data from 1.94 to 1.26 when using the smaller balanced reference panel and to 1.44 when using the larger but unbalanced reference panel. This generally supported the results from simulation but was less favorable, advocating for a larger reference panel when imputing to WGS.

Conclusions

The results highlight the potential of using imputation for mitigation of SNP ascertainment bias but also underline the need for unbiased reference sets.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-021-07663-6.

Genomic Prediction Based on SNP Functional Annotation Using Imputed Whole-Genome Sequence Data in Korean Hanwoo Cattle

Whole-genome sequence (WGS) data are increasingly being applied into genomic predictions, offering a higher predictive ability by including causal mutations or single-nucleotide polymorphisms (SNPs) putatively in strong linkage disequilibrium with causal mutations affecting the trait. This study aimed to improve the predictive performance of the customized Hanwoo 50 k SNP panel for four carcass traits in commercial Hanwoo population by adding highly predictive variants from sequence data. A total of 16,892 Hanwoo cattle with phenotypes (i.e., backfat thickness, carcass weight, longissimus muscle area, and marbling score), 50 k genotypes, and WGS imputed genotypes were used. We partitioned imputed WGS data according to functional annotation [intergenic (IGR), intron (ITR), regulatory (REG), synonymous (SYN), and non-synonymous (NSY)] to characterize the genomic regions that will deliver higher predictive power for the traits investigated. Animals were assigned into two groups, the discovery set (7324 animals) used for predictive variant detection and the cross-validation set for genomic prediction. Genome-wide association studies were performed by trait to every genomic region and entire WGS data for the pre-selection of variants. Each set of pre-selected SNPs with different density (1000, 3000, 5000, or 10,000) were added to the 50 k genotypes separately and the predictive performance of each set of genotypes was assessed using the genomic best linear unbiased prediction (GBLUP). Results showed that the predictive performance of the customized Hanwoo 50 k SNP panel can be improved by the addition of pre-selected variants from the WGS data, particularly 3000 variants from each trait, which is then sufficient to improve the prediction accuracy for all traits. When 12,000 pre-selected variants (3000 variants from each trait) were added to the 50 k genotypes, the prediction accuracies increased by 9.9, 9.2, 6.4, and 4.7% for backfat thickness, carcass weight, longissimus muscle area, and marbling score compared to the regular 50 k SNP panel, respectively. In terms of prediction bias, regression coefficients for all sets of genotypes in all traits were close to 1, indicating an unbiased prediction. The strategy used to select variants based on functional annotation did not show a clear advantage compared to using whole-genome. Nonetheless, such pre-selected SNPs from the IGR region gave the highest improvement in prediction accuracy among genomic regions and the values were close to those obtained using the WGS data for all traits. We concluded that additional gain in prediction accuracy when using pre-selected variants appears to be trait-dependent, and using WGS data remained more accurate compared to using a specific genomic region.

Multi-omics-data-assisted genomic feature markers preselection improves the accuracy of genomic prediction

Background

Presently, multi-omics data (e.g., genomics, transcriptomics, proteomics, and metabolomics) are available to improve genomic predictors. Omics data not only offers new data layers for genomic prediction but also provides a bridge between organismal phenotypes and genome variation that cannot be readily captured at the genome sequence level. Therefore, using multi-omics data to select feature markers is a feasible strategy to improve the accuracy of genomic prediction. In this study, simultaneously using whole-genome sequencing (WGS) and gene expression level data, four strategies for single-nucleotide polymorphism (SNP) preselection were investigated for genomic predictions in the Drosophila Genetic Reference Panel.

Results

Using genomic best linear unbiased prediction (GBLUP) with complete WGS data, the prediction accuracies were 0.208 ± 0.020 (0.181 ± 0.022) for the startle response and 0.272 ± 0.017 (0.307 ± 0.015) for starvation resistance in the female (male) lines. Compared with GBLUP using complete WGS data, both GBLUP and the genomic feature BLUP (GFBLUP) did not improve the prediction accuracy using SNPs preselected from complete WGS data based on the results of genome-wide association studies (GWASs) or transcriptome-wide association studies (TWASs). Furthermore, by using SNPs preselected from the WGS data based on the results of the expression quantitative trait locus (eQTL) mapping of all genes, only the startle response had greater accuracy than GBLUP with the complete WGS data. The best accuracy values in the female and male lines were 0.243 ± 0.020 and 0.220 ± 0.022, respectively. Importantly, by using SNPs preselected based on the results of the eQTL mapping of significant genes from TWAS, both GBLUP and GFBLUP resulted in great accuracy and small bias of genomic prediction. Compared with the GBLUP using complete WGS data, the best accuracy values represented increases of 60.66% and 39.09% for the starvation resistance and 27.40% and 35.36% for startle response in the female and male lines, respectively.

Conclusions

Overall, multi-omics data can assist genomic feature preselection and improve the performance of genomic prediction. The new knowledge gained from this study will enrich the use of multi-omics in genomic prediction.

Use of whole-genome sequence data and novel genomic selection strategies to improve selection for age at puberty in tropically-adapted beef heifers

Background

In tropically-adapted beef heifers, application of genomic prediction for age at puberty has been limited due to low prediction accuracies. Our aim was to investigate novel methods of pre-selecting whole-genome sequence (WGS) variants and alternative analysis methodologies; including genomic best linear unbiased prediction (GBLUP) with multiple genomic relationship matrices (MGRM) and Bayesian (BayesR) analyses, to determine if prediction accuracy for age at puberty can be improved.

Methods

Genotypes and phenotypes were obtained from two research herds. In total, 868 Brahman and 960 Tropical Composite heifers were recorded in the first population and 3695 Brahman, Santa Gertrudis and Droughtmaster heifers were recorded in the second population. Genotypes were imputed to 23 million whole-genome sequence variants. Eight strategies were used to pre-select variants from genome-wide association study (GWAS) results using conditional or joint (COJO) analyses. Pre-selected variants were included in three models, GBLUP with a single genomic relationship matrix (SGRM), GBLUP MGRM and BayesR. Five-way cross-validation was used to test the effect of marker panel density (6 K, 50 K and 800 K), analysis model, and inclusion of pre-selected WGS variants on prediction accuracy.

Results

In all tested scenarios, prediction accuracies for age at puberty were highest in BayesR analyses. The addition of pre-selected WGS variants had little effect on the accuracy of prediction when BayesR was used. The inclusion of WGS variants that were pre-selected using a meta-analysis with COJO analyses by chromosome, fitted in a MGRM model, had the highest prediction accuracies in the GBLUP analyses, regardless of marker density. When the low-density (6 K) panel was used, the prediction accuracy of GBLUP was equal (0.42) to that with the high-density panel when only six additional sequence variants (identified using meta-analysis COJO by chromosome) were included.

Conclusions

While BayesR consistently outperforms other methods in terms of prediction accuracies, reasonable improvements in accuracy can be achieved when using GBLUP and low-density panels with the inclusion of a relatively small number of highly relevant WGS variants.

High-Throughput Sequencing With the Preselection of Markers Is a Good Alternative to SNP Chips for Genomic Prediction in Broilers

The choice of a genetic marker genotyping platform is important for genomic prediction in livestock and poultry. High-throughput sequencing can produce more genetic markers, but the genotype quality is lower than that obtained with single nucleotide polymorphism (SNP) chips. The aim of this study was to compare the accuracy of genomic prediction between high-throughput sequencing and SNP chips in broilers. In this study, we developed a new SNP marker screening method, the pre-marker-selection (PMS) method, to determine whether an SNP marker can be used for genomic prediction. We also compared a method which preselection marker based results from genome-wide association studies (GWAS). With the two methods, we analysed body weight at the12^th week (BW) and feed conversion ratio (FCR) in a local broiler population. A total of 395 birds were selected from the F2 generation of the population, and 10X specific-locus amplified fragment sequencing (SLAF-seq) and the Illumina Chicken 60K SNP Beadchip were used for genotyping. The genomic best linear unbiased prediction method (GBLUP) was used to predict the genomic breeding values. The accuracy of genomic prediction was validated by the leave-one-out cross-validation method. Without SNP marker screening, the accuracies of the genomic estimated breeding value (GEBV) of BW and FCR were 0.509 and 0.249, respectively, when using SLAF-seq, and the accuracies were 0.516 and 0.232, respectively, when using the SNP chip. With SNP marker screening by the PMS method, the accuracies of GEBV of the two traits were 0.671 and 0.499, respectively, when using SLAF-seq, and 0.605 and 0.422, respectively, when using the SNP chip. Our SNP marker screening method led to an increase of prediction accuracy by 0.089-0.250. With SNP marker screening by the GWAS method, the accuracies of genomic prediction for the two traits were also improved, but the gains of accuracy were less than the gains with PMS method for all traits. The results from this study indicate that our PMS method can improve the accuracy of GEBV, and that more accurate genomic prediction can be obtained from an increased number of genomic markers when using high-throughput sequencing in local broiler populations. Due to its lower genotyping cost, high-throughput sequencing could be a good alternative to SNP chips for genomic prediction in breeding programmes of local broiler populations.

A Guide for Using Deep Learning for Complex Trait Genomic Prediction

Deep learning (DL) has emerged as a powerful tool to make accurate predictions from complex data such as image, text, or video. However, its ability to predict phenotypic values from molecular data is less well studied. Here, we describe the theoretical foundations of DL and provide a generic code that can be easily modified to suit specific needs. DL comprises a wide variety of algorithms which depend on numerous hyperparameters. Careful optimization of hyperparameter values is critical to avoid overfitting. Among the DL architectures currently tested in genomic prediction, convolutional neural networks (CNNs) seem more promising than multilayer perceptrons (MLPs). A limitation of DL is in interpreting the results. This may not be relevant for genomic prediction in plant or animal breeding but can be critical when deciding the genetic risk to a disease. Although DL technologies are not "plug-and-play", they are easily implemented using Keras and TensorFlow public software. To illustrate the principles described here, we implemented a Keras-based code in GitHub.

Strategies for Obtaining and Pruning Imputed Whole-Genome Sequence Data for Genomic Prediction

Genomic prediction with imputed whole-genome sequencing (WGS) data is an attractive approach to improve predictive ability with low cost. However, high accuracy has not been realized using this method in livestock. In this study, we imputed 435 individuals from 600K single nucleotide polymorphism (SNP) chip data to WGS data using different reference panels. We also investigated the prediction accuracy of genomic best linear unbiased prediction (GBLUP) using imputed WGS data from different reference panels, linkage disequilibrium (LD)-based marker pruning, and pre-selected variants based on Genome-wide association society (GWAS) results. Results showed that the imputation accuracies from 600K to WGS data were 0.873 ± 0.038, 0.906 ± 0.036, and 0.979 ± 0.010 for the internal, external, and combined reference panels, respectively. In most traits of chickens, the prediction accuracy of imputed WGS data obtained from the internal reference panel was greater than or equal to that of the combined reference panel; the external reference panel had the lowest prediction accuracy. Compared with 600K chip data, GBLUP with imputed WGS data had only a small increase (1-3%) in prediction accuracy. Using only variants selected from imputed WGS data based on GWAS results resulted in almost no increase for most traits and even increased the bias of the regression coefficient. The impact of the degree of LD of selected and remaining variants on prediction accuracy was different. For average daily gain (ADG), residual feed intake (RFI), intestine length (IL), and body weight in 91 days (BW91), the accuracy of GBLUP increased as the degree of LD of selected variants decreased, but the opposite relationship occurred for the remaining variants. But for breast muscle weight (BMW) and average daily feed intake (ADFI), the accuracy of GBLUP increased as the degree of LD of selected variants increased, and the degree of LD of remaining variants had a small effect on prediction accuracy. Overall, the optimal imputation strategy to obtain WGS data for genomic prediction should consider the relationship between selected individuals and target population individuals to avoid heterogeneity of imputation. LD-based marker pruning can be used to improve the accuracy of genomic prediction using imputed WGS data.

Applications of the Soil, Plant and Rumen Microbiomes in Pastoral Agriculture

The production of dairy, meat, and fiber by ruminant animals relies on the biological processes occurring in soils, forage plants, and the animals' rumens. Each of these components has an associated microbiome, and these have traditionally been viewed as distinct ecosystems. However, these microbiomes operate under similar ecological principles and are connected via water, energy flows, and the carbon and nitrogen nutrient cycles. Here, we summarize the microbiome research that has been done in each of these three environments (soils, forage plants, animals' rumen) and investigate what additional benefits may be possible through understanding the interactions between the various microbiomes. The challenge for future research is to enhance microbiome function by appropriate matching of plant and animal genotypes with the environment to improve the output and environmental sustainability of pastoral agriculture.

Impact of rare and low-frequency sequence variants on reliability of genomic prediction in dairy cattle

Background

Availability of whole-genome sequence data for a large number of cattle and efficient imputation methodologies open a new opportunity to include rare and low-frequency variants (RLFV) in genomic prediction in dairy cattle. The objective of this study was to examine the impact of including RLFV that are within genes and selected from whole-genome sequence variants, on the reliability of genomic prediction for fertility, health and longevity in dairy cattle.

Results

All genic RLFV with a minor allele frequency lower than 0.05 were extracted from imputed sequence data and subsets were created using different strategies. These subsets were subsequently combined with Illumina 50 k single nucleotide polymorphism (SNP) data and used for genomic prediction. Reliability of prediction obtained by using 50 k SNP data alone was used as reference value and absolute changes in reliabilities are referred to as changes in percentage points. Adding a component that included either all the genic or a subset of selected RLFV into the model in addition to the 50 k component changed the reliability of predictions by − 2.2 to 1.1%, i.e. hardly no change in reliability of prediction was found, regardless of how the RLFV were selected. In addition to these empirical analyses, a simulation study was performed to evaluate the potential impact of adding RLFV in the model on the reliability of prediction. Three sets of causal RLFV (containing 21,468, 1348 and 235 RLFV) that were randomly selected from different numbers of genes were generated and accounted for 10% additional genetic variance of the estimated variance explained by the 50 k SNPs. When genic RLFV based on mapping results were included in the prediction model, reliabilities improved by up to 4.0% and when the causal RLFV were included they improved by up to 6.8%.

Conclusions

Using selected RLFV from whole-genome sequence data had only a small impact on the empirical reliability of genomic prediction in dairy cattle. Our simulations revealed that for sequence data to bring a benefit, the key is to identify causal RLFV.

Electronic supplementary material

The online version of this article (10.1186/s12711-018-0432-8) contains supplementary material, which is available to authorized users.

Accuracy of Genomic Prediction for Foliar Terpene Traits in Eucalyptus polybractea

Unlike agricultural crops, most forest species have not had millennia of improvement through phenotypic selection, but can contribute energy and material resources and possibly help alleviate climate change. Yield gains similar to those achieved in agricultural crops over millennia could be made in forestry species with the use of genomic methods in a much shorter time frame. Here we compare various methods of genomic prediction for eight traits related to foliar terpene yield in Eucalyptus polybractea, a tree grown predominantly for the production of Eucalyptus oil. The genomic markers used in this study are derived from shallow whole genome sequencing of a population of 480 trees. We compare the traditional pedigree-based additive best linear unbiased predictors (ABLUP), genomic BLUP (GBLUP), BayesB genomic prediction model, and a form of GBLUP based on weighting markers according to their influence on traits (BLUP|GA). Predictive ability is assessed under varying marker densities of 10,000, 100,000 and 500,000 SNPs. Our results show that BayesB and BLUP|GA perform best across the eight traits. Predictive ability was higher for individual terpene traits, such as foliar α-pinene and 1,8-cineole concentration (0.59 and 0.73, respectively), than aggregate traits such as total foliar oil concentration (0.38). This is likely a function of the trait architecture and markers used. BLUP|GA was the best model for the two biomass related traits, height and 1 year change in height (0.25 and 0.19, respectively). Predictive ability increased with marker density for most traits, but with diminishing returns. The results of this study are a solid foundation for yield improvement of essential oil producing eucalypts. New markets such as biopolymers and terpene-derived biofuels could benefit from rapid yield increases in undomesticated oil-producing species.

Genomic evaluation of feed efficiency component traits in Duroc pigs using 80K, 650K and whole-genome sequence variants

BACKGROUND

Increasing marker density was proposed to have potential to improve the accuracy of genomic prediction for quantitative traits; whole-sequence data is expected to give the best accuracy of prediction, since all causal mutations that underlie a trait are expected to be included. However, in cattle and chicken, this assumption is not supported by empirical studies. Our objective was to compare the accuracy of genomic prediction of feed efficiency component traits in Duroc pigs using single nucleotide polymorphism (SNP) panels of 80K, imputed 650K, and whole-genome sequence variants using GBLUP, BayesB and BayesRC methods, with the ultimate purpose to determine the optimal method to increase genetic gain for feed efficiency in pigs.

RESULTS

Phenotypes of average daily feed intake (ADFI), average daily gain (ADG), ultrasound backfat depth (FAT), and loin muscle depth (LMD) were available for 1363 Duroc boars from a commercial breeding program. Genotype imputation accuracies reached 92.1% from 80K to 650K and 85.6% from 650K to whole-genome sequence variants. Average accuracies across methods and marker densities of genomic prediction of ADFI, FAT, LMD and ADG were 0.40, 0.65, 0.30 and 0.15, respectively. For ADFI and FAT, BayesB outperformed GBLUP, but increasing marker density had little advantage for genomic prediction. For ADG and LMD, GBLUP outperformed BayesB, while BayesRC based on whole-genome sequence data gave the best accuracies and reached up to 0.35 for LMD and 0.25 for ADG.

CONCLUSIONS

Use of genomic information was beneficial for prediction of ADFI and FAT but not for that of ADG and LMD compared to pedigree-based estimates. BayesB based on 80K SNPs gave the best genomic prediction accuracy for ADFI and FAT, while BayesRC based on whole-genome sequence data performed best for ADG and LMD. We suggest that these differences between traits in the effect of marker density and method on accuracy of genomic prediction are mainly due to the underlying genetic architecture of the traits.

Estimation of linkage disequilibrium and analysis of genetic diversity in Korean chicken lines

The development of genetic markers for animal breeding is an effective strategy to reduce the time and cost required to improve economically important traits. To implement genomic selection in the multibreed chicken population of Korea, an understanding of the linkage disequilibrium (LD) status of the target population is essential. In this study, we performed population genetic analyses to investigate LD decay, the effective population size, and breed diversity using 600K high-density single nucleotide polymorphism genotypes of 189 native chickens in 14 lines (including Korean native chicken, imported and adapted purebred and commercial chickens). The results indicated that commercial native chickens have less calculated LD (average, r2 = 0.13-0.26) and purebred native chickens have more calculated LD (average, r2 = 0.24-0.37) across the entire genome. The effective population sizes of the examined lines showed patterns opposite to those of population LD. The phylogeny and admixture analyses showed that commercial and purebred chickens were well distinguished, except for Rhode Island Red (RIR) purebred lines of NC (NIAS_RIR_C) and ND (NIAS_RIR_D). These lines are difficult to distinguish clearly because they originated from the same respective breeds. The results of this study may provide important information for the development of genetic markers that can be used in breeding to improve the economic traits of native chickens.

Estimated allele substitution effects underlying genomic evaluation models depend on the scaling of allele counts

Background

Genomic evaluation is used to predict direct genomic values (DGV) for selection candidates in breeding programs, but also to estimate allele substitution effects (ASE) of single nucleotide polymorphisms (SNPs). Scaling of allele counts influences the estimated ASE, because scaling of allele counts results in less shrinkage towards the mean for low minor allele frequency (MAF) variants. Scaling may become relevant for estimating ASE as more low MAF variants will be used in genomic evaluations. We show the impact of scaling on estimates of ASE using real data and a theoretical framework, and in terms of power, model fit and predictive performance.

Results

In a dairy cattle dataset with 630 K SNP genotypes, the correlation between DGV for stature from a random regression model using centered allele counts (RRc) and centered and scaled allele counts (RRcs) was 0.9988, whereas the overall correlation between ASE using RRc and RRcs was 0.27. The main difference in ASE between both methods was found for SNPs with a MAF lower than 0.01. Both the ratio (ASE from RRcs/ASE from RRc) and the regression coefficient (regression of ASE from RRcs on ASE from RRc) were much higher than 1 for low MAF SNPs. Derived equations showed that scenarios with a high heritability, a large number of individuals and a small number of variants have lower ratios between ASE from RRc and RRcs. We also investigated the optimal scaling parameter [from − 1 (RRcs) to 0 (RRc) in steps of 0.1] in the bovine stature dataset. We found that the log-likelihood was maximized with a scaling parameter of − 0.8, while the mean squared error of prediction was minimized with a scaling parameter of − 1, i.e., RRcs.

Conclusions

Large differences in estimated ASE were observed for low MAF SNPs when allele counts were scaled or not scaled because there is less shrinkage towards the mean for scaled allele counts. We derived a theoretical framework that shows that the difference in ASE due to shrinkage is heavily influenced by the power of the data. Increasing the power results in smaller differences in ASE whether allele counts are scaled or not.

Multi-breed genomic prediction using Bayes R with sequence data and dropping variants with a small effect

Background

The increasing availability of whole-genome sequence data is expected to increase the accuracy of genomic prediction. However, results from simulation studies and analysis of real data do not always show an increase in accuracy from sequence data compared to high-density (HD) single nucleotide polymorphism (SNP) chip genotypes. In addition, the sheer number of variants makes analysis of all variants and accurate estimation of all effects computationally challenging. Our objective was to find a strategy to approximate the analysis of whole-sequence data with a Bayesian variable selection model. Using a simulated dataset, we applied a Bayes R hybrid model to analyse whole-sequence data, test the effect of dropping a proportion of variants during the analysis, and test how the analysis can be split into separate analyses per chromosome to reduce the elapsed computing time. We also investigated the effect of imputation errors on prediction accuracy. Subsequently, we applied the approach to a dataset that contained imputed sequences and records for production and fertility traits for 38,492 Holstein, Jersey, Australian Red and crossbred bulls and cows.

Results

With the simulated dataset, we found that prediction accuracy was highly increased for a breed that was not represented in the training population for sequence data compared to HD SNP data. Either dropping part of the variants during the analysis or splitting the analysis into separate analyses per chromosome decreased accuracy compared to analysing whole-sequence data. First, dropping variants from each chromosome and reanalysing the retained variants together resulted in an accuracy similar to that obtained when analysing whole-sequence data. Adding imputation errors decreased prediction accuracy, especially for errors in the validation population. With real data, using sequence variants resulted in accuracies that were similar to those obtained with the HD SNPs.

Conclusions

We present an efficient approach to approximate analysis of whole-sequence data with a Bayesian variable selection model. The lack of increase in prediction accuracy when applied to real data could be due to imputation errors, which demonstrates the importance of developing more accurate methods of imputation or directly genotyping sequence variants that have a major effect in the prediction equation.

Electronic supplementary material

The online version of this article (doi:10.1186/s12711-017-0347-9) contains supplementary material, which is available to authorized users.

Influence of epistasis on response to genomic selection using complete sequence data

BACKGROUND

The effect of epistasis on response to selection is a highly debated topic. Here, we investigated the impact of epistasis on response to sequence-based selection via genomic best linear prediction (GBLUP) in a regime of strong non-symmetrical epistasis under divergent selection, using real Drosophila sequence data. We also explored the possible advantage of including epistasis in the evaluation model and/or of knowing the causal mutations.

RESULTS

Response to selection was almost exclusively due to changes in allele frequency at a few loci with a large effect. Response was highly asymmetric (about four phenotypic standard deviations higher for upward than downward selection) due to the highly skewed site frequency spectrum. Epistasis accentuated this asymmetry and affected response to selection by modulating the additive genetic variance, which was sustained for longer under upward selection whereas it eroded rapidly under downward selection. Response to selection was quite insensitive to the evaluation model, especially under an additive scenario. Nevertheless, including epistasis in the model when there was none eventually led to lower accuracies as selection proceeded. Accounting for epistasis in the model, if it existed, was beneficial but only in the medium term. There was not much gain in response if causal mutations were known, compared to using sequence data, which is likely due to strong linkage disequilibrium, high heritability and availability of phenotypes on candidates.

CONCLUSIONS

Epistatic interactions affect the response to genomic selection by modulating the additive genetic variance used for selection. Epistasis releases additive variance that may increase response to selection compared to a pure additive genetic action. Furthermore, genomic evaluation models and, in particular, GBLUP are robust, i.e. adding complexity to the model did not modify substantially the response (for a given architecture).

Fixed-length haplotypes can improve genomic prediction accuracy in an admixed dairy cattle population

Background

Fitting covariates representing the number of haplotype alleles rather than single nucleotide polymorphism (SNP) alleles may increase genomic prediction accuracy if linkage disequilibrium between quantitative trait loci and SNPs is inadequate. The objectives of this study were to evaluate the accuracy, bias and computation time of Bayesian genomic prediction methods that fit fixed-length haplotypes or SNPs. Genotypes at 37,740 SNPs that were common to Illumina BovineSNP50 and high-density panels were phased for ~58,000 New Zealand dairy cattle. Females born before 1 June 2008 were used for training, and genomic predictions for milk fat yield (n = 24,823), liveweight (n = 13,283) and somatic cell score (n = 24,864) were validated within breed (predominantly Holstein–Friesian, predominantly Jersey, or admixed KiwiCross) in later-born females. Covariates for haplotype alleles of five lengths (125, 250, 500 kb, 1 or 2 Mb) were generated and rare haplotypes were removed at four thresholds (1, 2, 5 or 10%), resulting in 20 scenarios tested. Genomic predictions fitting covariates for either SNPs or haplotypes were calculated by using BayesA, BayesB or BayesN. This is the first study to quantify the accuracy of genomic prediction using haplotypes across the whole genome in an admixed population.

Results

A correlation of 0.349 ± 0.016 between yield deviation and genomic breeding values was obtained for milk fat yield in Holstein–Friesians using BayesA fitting covariates. Genomic predictions were more accurate with short haplotypes than with SNPs but less accurate with longer haplotypes than with SNPs. Fitting only the most frequent haplotype alleles reduced computation time with little decrease in prediction accuracy for short haplotypes. Trends were similar for all traits and breeds and there was little difference between Bayesian methods.

Conclusions

Fitting covariates for haplotype alleles rather than SNPs can increase prediction accuracy, although it decreased drastically for long (>500 kb) haplotypes. In this population, fitting 250 kb haplotypes with a 1% frequency threshold resulted in the highest genomic prediction accuracy and fitting 125 kb haplotypes with a 10% frequency threshold improved genomic prediction accuracy with comparable computation time to fitting SNPs. This increased accuracy is likely to increase genetic gain by changing the ranking of selection candidates.

Electronic supplementary material

The online version of this article (doi:10.1186/s12711-017-0329-y) contains supplementary material, which is available to authorized users.

Whole-genome sequence-based genomic prediction in laying chickens with different genomic relationship matrices to account for genetic architecture

Background

With the availability of next-generation sequencing technologies, genomic prediction based on whole-genome sequencing (WGS) data is now feasible in animal breeding schemes and was expected to lead to higher predictive ability, since such data may contain all genomic variants including causal mutations. Our objective was to compare prediction ability with high-density (HD) array data and WGS data in a commercial brown layer line with genomic best linear unbiased prediction (GBLUP) models using various approaches to weight single nucleotide polymorphisms (SNPs).

Methods

A total of 892 chickens from a commercial brown layer line were genotyped with 336 K segregating SNPs (array data) that included 157 K genic SNPs (i.e. SNPs in or around a gene). For these individuals, genome-wide sequence information was imputed based on data from re-sequencing runs of 25 individuals, leading to 5.2 million (M) imputed SNPs (WGS data), including 2.6 M genic SNPs. De-regressed proofs (DRP) for eggshell strength, feed intake and laying rate were used as quasi-phenotypic data in genomic prediction analyses. Four weighting factors for building a trait-specific genomic relationship matrix were investigated: identical weights, −(log₁₀P) from genome-wide association study results, squares of SNP effects from random regression BLUP, and variable selection based weights (known as BLUP|GA). Predictive ability was measured as the correlation between DRP and direct genomic breeding values in five replications of a fivefold cross-validation.

Results

Averaged over the three traits, the highest predictive ability (0.366 ± 0.075) was obtained when only genic SNPs from WGS data were used. Predictive abilities with genic SNPs and all SNPs from HD array data were 0.361 ± 0.072 and 0.353 ± 0.074, respectively. Prediction with −(log₁₀P) or squares of SNP effects as weighting factors for building a genomic relationship matrix or BLUP|GA did not increase accuracy, compared to that with identical weights, regardless of the SNP set used.

Conclusions

Our results show that little or no benefit was gained when using all imputed WGS data to perform genomic prediction compared to using HD array data regardless of the weighting factors tested. However, using only genic SNPs from WGS data had a positive effect on prediction ability.

Electronic supplementary material

The online version of this article (doi:10.1186/s12711-016-0277-y) contains supplementary material, which is available to authorized users.

Predictive performance of genomic selection methods for carcass traits in Hanwoo beef cattle: impacts of the genetic architecture

Background

Hanwoo beef is known for its marbled fat, tenderness, juiciness and characteristic flavor, as well as for its low cholesterol and high omega 3 fatty acid contents. As yet, there has been no comprehensive investigation to estimate genomic selection accuracy for carcass traits in Hanwoo cattle using dense markers. This study aimed at evaluating the accuracy of alternative statistical methods that differed in assumptions about the underlying genetic model for various carcass traits: backfat thickness (BT), carcass weight (CW), eye muscle area (EMA), and marbling score (MS).

Methods

Accuracies of direct genomic breeding values (DGV) for carcass traits were estimated by applying fivefold cross-validation to a dataset including 1183 animals and approximately 34,000 single nucleotide polymorphisms (SNPs).

Results

Accuracies of BayesC, Bayesian LASSO (BayesL) and genomic best linear unbiased prediction (GBLUP) methods were similar for BT, EMA and MS. However, for CW, DGV accuracy was 7% higher with BayesC than with BayesL and GBLUP. The increased accuracy of BayesC, compared to GBLUP and BayesL, was maintained for CW, regardless of the training sample size, but not for BT, EMA, and MS. Genome-wide association studies detected consistent large effects for SNPs on chromosomes 6 and 14 for CW.

Conclusions

The predictive performance of the models depended on the trait analyzed. For CW, the results showed a clear superiority of BayesC compared to GBLUP and BayesL. These findings indicate the importance of using a proper variable selection method for genomic selection of traits and also suggest that the genetic architecture that underlies CW differs from that of the other carcass traits analyzed. Thus, our study provides significant new insights into the carcass traits of Hanwoo cattle.

Electronic supplementary material

The online version of this article (doi:10.1186/s12711-016-0283-0) contains supplementary material, which is available to authorized users.

Evaluating Sequence-Based Genomic Prediction with an Efficient New Simulator

The vast amount of sequence data generated to analyze complex traits is posing new challenges in terms of the analysis and interpretation of the results. Although simulation is a fundamental tool to investigate the reliability of genomic analyses and to optimize experimental design, existing software cannot realistically simulate complete genomes. To remedy this, we have developed a new strategy (Sequence-Based Virtual Breeding, SBVB) that uses real sequence data and simulates new offspring genomes and phenotypes in a very efficient and flexible manner. Using this tool, we studied the efficiency of full sequence in genomic prediction compared to SNP arrays. We used real porcine sequences from three breeds as founder genomes of a 2500-animal pedigree and two genetic architectures: "neutral" and "selective." In the neutral architecture, frequencies and allele effects were sampled independently whereas, in the selective case, SNPs were sites putatively under selection after domestication and a negative correlation between effect and frequency was induced. We compared the effectiveness of different genotyping strategies for genomic selection, including the use of full sequence commercial arrays or randomly chosen SNP sets in both outbred and crossbred experimental designs. We found that accuracy increases using sequence instead of commercial chips but modestly, perhaps by ≤ 4%. This result was robust to extreme genetic architectures. We conclude that full sequence is unlikely to offset commercial arrays for predicting genetic value when the number of loci is relatively large and the prior given to each SNP is uniform. Using sequence to improve selection thus requires optimized prior information and, likely, increased population sizes. The code and manual for SBVB are available at https://github.com/mperezenciso/sbvb0.

Genomic prediction using preselected DNA variants from a GWAS with whole-genome sequence data in Holstein-Friesian cattle

Background

Whole-genome sequence data is expected to capture genetic variation more completely than common genotyping panels. Our objective was to compare the proportion of variance explained and the accuracy of genomic prediction by using imputed sequence data or preselected SNPs from a genome-wide association study (GWAS) with imputed whole-genome sequence data.

Methods

Phenotypes were available for 5503 Holstein–Friesian bulls. Genotypes were imputed up to whole-genome sequence (13,789,029 segregating DNA variants) by using run 4 of the 1000 bull genomes project. The program GCTA was used to perform GWAS for protein yield (PY), somatic cell score (SCS) and interval from first to last insemination (IFL). From the GWAS, subsets of variants were selected and genomic relationship matrices (GRM) were used to estimate the variance explained in 2087 validation animals and to evaluate the genomic prediction ability. Finally, two GRM were fitted together in several models to evaluate the effect of selected variants that were in competition with all the other variants.

Results

The GRM based on full sequence data explained only marginally more genetic variation than that based on common SNP panels: for PY, SCS and IFL, genomic heritability improved from 0.81 to 0.83, 0.83 to 0.87 and 0.69 to 0.72, respectively. Sequence data also helped to identify more variants linked to quantitative trait loci and resulted in clearer GWAS peaks across the genome. The proportion of total variance explained by the selected variants combined in a GRM was considerably smaller than that explained by all variants (less than 0.31 for all traits). When selected variants were used, accuracy of genomic predictions decreased and bias increased.

Conclusions

Although 35 to 42 variants were detected that together explained 13 to 19% of the total variance (18 to 23% of the genetic variance) when fitted alone, there was no advantage in using dense sequence information for genomic prediction in the Holstein data used in our study. Detection and selection of variants within a single breed are difficult due to long-range linkage disequilibrium. Stringent selection of variants resulted in more biased genomic predictions, although this might be due to the training population being the same dataset from which the selected variants were identified.

Electronic supplementary material

The online version of this article (doi:10.1186/s12711-016-0274-1) contains supplementary material, which is available to authorized users.

Efficient genomic prediction based on whole-genome sequence data using split-and-merge Bayesian variable selection

BACKGROUND

Use of whole-genome sequence data is expected to increase persistency of genomic prediction across generations and breeds but affects model performance and requires increased computing time. In this study, we investigated whether the split-and-merge Bayesian stochastic search variable selection (BSSVS) model could overcome these issues. BSSVS is performed first on subsets of sequence-based variants and then on a merged dataset containing variants selected in the first step.

RESULTS

We used a dataset that included 4,154,064 variants after editing and de-regressed proofs for 3415 reference and 2138 validation bulls for somatic cell score, protein yield and interval first to last insemination. In the first step, BSSVS was performed on 106 subsets each containing ~39,189 variants. In the second step, 1060 up to 472,492 variants, selected from the first step, were included to estimate the accuracy of genomic prediction. Accuracies were at best equal to those achieved with the commonly used Bovine 50k-SNP chip, although the number of variants within a few well-known quantitative trait loci regions was considerably enriched. When variant selection and the final genomic prediction were performed on the same data, predictions were biased. Predictions computed as the average of the predictions computed for each subset achieved the highest accuracies, i.e. 0.5 to 1.1 % higher than the accuracies obtained with the 50k-SNP chip, and yielded the least biased predictions. Finally, the accuracy of genomic predictions obtained when all sequence-based variants were included was similar or up to 1.4 % lower compared to that based on the average predictions across the subsets. By applying parallelization, the split-and-merge procedure was completed in 5 days, while the standard analysis including all sequence-based variants took more than three months.

CONCLUSIONS

The split-and-merge approach splits one large computational task into many much smaller ones, which allows the use of parallel processing and thus efficient genomic prediction based on whole-genome sequence data. The split-and-merge approach did not improve prediction accuracy, probably because we used data on a single breed for which relationships between individuals were high. Nevertheless, the split-and-merge approach may have potential for applications on data from multiple breeds.