Combined Reference Population Research Articles

Integrating large-scale meta-analysis of genome-wide association studies improve the genomic prediction accuracy for combined pig populations.

The strategy of combining reference populations has been widely recognized as an effective way to enhance the accuracy of genomic prediction (GP). This study investigated the efficiency of genomic prediction using prior information and combined reference population. In total, prior information considering trait-associated single nucleotide polymorphisms (SNPs) obtained from meta-analysis of genome-wide association studies (GWAS meta-analysis) was incorporated into three models to assess the performance of GP using combined reference populations. Two different Yorkshire populations with imputed whole genome sequence (WGS) data (9,741,620 SNPs), named as P1 (1259 individuals) and P2 (1018 individuals), were used to predict genomic estimated breeding values for three live carcass traits, including backfat thickness, loin muscle area, and loin muscle depth. A 10 × 5 fold cross-validation was used to evaluate the prediction accuracy of 203 randomly selected candidate pigs from the P2 population and the reference population consisted of the remaining pigs from P2 and the stepwise added pigs from P1. By integrating SNPs with different p-value thresholds from GWAS meta-analysis downloaded from PigGTEx Project, the prediction accuracy of GBLUP, genomic feature BLUP (GFBLUP) and GBLUP given genetic architecture (BLUP|GA) were compared. Moreover, we explored effects of reference population size and heritability enrichment of genomic features on the prediction accuracy improvement of GFBLUP and BLUP|GA relative to GBLUP. The prediction accuracy of GBLUP using all WGS markers showed average improvement of 4.380% using the P1 + P2 reference population compared with the P2 reference population. Using the combined reference population, GFBLUP and BLUP|GA yielded 6.179% and 5.525% higher accuracies than GBLUP using all SNPs based on the single reference population, respectively. Positive regression coefficients were estimated in relation to the improvement in prediction accuracy (between GFBLUP/BLUP|GA and GBLUP) and the size of the reference as well as the heritability enrichment of genomic features. Compared to the classic GBLUP model, GFBLUP and BLUP|GA models integrating GWAS meta-analysis information increase the prediction accuracy and using combined populations with enlarged reference population size further enhances prediction accuracy of the two approaches. The heritability enrichment of genomic features can be used as an indicator to reflect weather prior information is accurately presented.

Multibreed genomic prediction using summary statistics and a breed-origin-of-alleles approach

Because of an increasing interest in crossbreeding between dairy breeds in dairy cattle herds, farmers are requesting breeding values for crossbred animals. However, genomically enhanced breeding values are difficult to predict in crossbred populations because the genetic make-up of crossbred individuals is unlikely to follow the same pattern as for purebreds. Furthermore, sharing genotype and phenotype information between breed populations are not always possible, which means that genetic merit (GM) for crossbred animals may be predicted without the information needed from some pure breeds, resulting in low prediction accuracy. This simulation study investigated the consequences of using summary statistics from single-breed genomic predictions for some or all pure breeds in two- and three-breed rotational crosses, rather than their raw data. A genomic prediction model taking into account the breed-origin of alleles (BOA) was considered. Because of a high genomic correlation between the breeds simulated (0.62–0.87), the prediction accuracies using the BOA approach were similar to a joint model, assuming homogeneous SNP effects for these breeds. Having a reference population with summary statistics available from all pure breeds and full phenotype and genotype information from crossbreds yielded almost as high prediction accuracies (0.720–0.768) as having a reference population with full information from all pure breeds and crossbreds (0.753–0.789). Lacking information from the pure breeds yielded much lower prediction accuracies (0.590–0.676). Furthermore, including crossbred animals in a combined reference population also benefitted prediction accuracies in the purebred animals, especially for the smallest breed population.

Accounting for differences in linkage disequilibrium in multi-breed genomic prediction

The lack of improvement in prediction accuracy in multi-breed genomic evaluation can be due to the (a) inconsistency in allele substitution effects across breeds, (b) the absence of close family relationships between breeds and (c) between-breed differences in linkage disequilibrium (LD) between single nucleotide polymorphisms (SNPs) and quantitative trait loci that influence a trait across breeds. The objective of this study was to investigate the possibility of improvement in prediction accuracy by accounting for the LD phase differences in a multi-breed reference population. Prediction accuracy was compared in three different scenarios. Scenarios had separate or combined reference population. In the proposed method, when reference population of two breeds were combined, subset of common SNPs between two breeds with high and similar LD phases were identified. Then, A Bayesian Ridge regression model was used to estimated effects for two set of SNPs, SNPs with similar LD phase in both breeds and the remaining SNPs. Results showed that simple pooling of two reference population into a single reference population did not improve prediction accuracy compared to a separate reference population of each breed. However, accounting for LD phase differences in two breeds, improved prediction accuracy compared to separate training and simple pooling of reference populations. This can be appealing for small populations (e.g., beef) that often have relatively small reference population and a combined genetic evaluation is usually performed on multiple breeds.

Comparison of combined, crossbred, and purebred reference populations for genomic selection in small populations

Abstract The main purpose of the present study was to evaluate the accuracy of genomic predictions in two purebreds using three different ways for the reference population formation: (1) a separate purebred reference population (SI), (2) a combined reference population from two breeds (CO), and (3) use of crossbred data (CR). The effects of different reference population structure were compared under different marker densities, reference population sizes, divergence between breeds, and methods used for estimating marker effects. The results showed that at lower marker densities, the SI had the highest accuracy of breeding values. With the increase in the marker density, the use of the CO increased the accuracy relative to the SI in scenarios were generations since divergence was low. In the CO and CR, the BayesB method was more accurate than the rrBLUP, and decreasing the generations since divergence increased the accuracy of predictions at higher marker density. Increased population size, marker density, and genetic relationship between breeds and using the BayesB method retained accuracies higher after five generations. Interestingly, the CR had the lowest accuracy drop after five generations.

Genomic prediction for numerically small breeds, using models with pre-selected and differentially weighted markers

BackgroundGenomic prediction (GP) accuracy in numerically small breeds is limited by the small size of the reference population. Our objective was to test a multi-breed multiple genomic relationship matrices (GRM) GP model (MBMG) that weighs pre-selected markers separately, uses the remaining markers to explain the remaining genetic variance that can be explained by markers, and weighs information of breeds in the reference population by their genetic correlation with the validation breed.MethodsGenotype and phenotype data were used on 595 Jersey bulls from New Zealand and 5503 Holstein bulls from the Netherlands, all with deregressed proofs for stature. Different sets of markers were used, containing either pre-selected markers from a meta-genome-wide association analysis on stature, remaining markers or both. We implemented a multi-breed bivariate GREML model in which we fitted either a single multi-breed GRM (MBSG), or two distinct multi-breed GRM (MBMG), one made with pre-selected markers and the other with remaining markers. Accuracies of predicting stature for Jersey individuals using the multi-breed models (Holstein and Jersey combined reference population) was compared to those obtained using either the Jersey (within-breed) or Holstein (across-breed) reference population. All the models were subsequently fitted in the analysis of simulated phenotypes, with a simulated genetic correlation between breeds of 1, 0.5, and 0.25.ResultsThe MBMG model always gave better prediction accuracies for stature compared to MBSG, within-, and across-breed GP models. For example, with MBSG, accuracies obtained by fitting 48,912 unselected markers (0.43), 357 pre-selected markers (0.38) or a combination of both (0.43), were lower than accuracies obtained by fitting pre-selected and unselected markers in separate GRM in MBMG (0.49). This improvement was further confirmed by results from a simulation study, with MBMG performing on average 23% better than MBSG with all markers fitted.ConclusionsWith the MBMG model, it is possible to use information from numerically large breeds to improve prediction accuracy of numerically small breeds. The superiority of MBMG is mainly due to its ability to use information on pre-selected markers, explain the remaining genetic variance and weigh information from a different breed by the genetic correlation between breeds.

The genetic connectedness calculated from genomic information and its effect on the accuracy of genomic prediction.

The magnitude of connectedness among management units (e.g., flocks and herds) gives a reliable estimate of genetic evaluation across these units. Traditionally, pedigree-based methods have been used to evaluate the genetic connectedness in China. However, these methods have not been able to yield a substantial outcome due to the lack of accuracy and integrity of pedigree data. Therefore, it is necessary to ascertain genetic connectedness using genomic information (i.e., genome-based genetic connectedness). Moreover, the effects of various levels of genome-based genetic connectedness on the accuracy of genomic prediction still remain poorly understood. A simulation study was performed to evaluate the genome-based genetic connectedness across herds by applying prediction error variance of difference (PEVD), coefficient of determination (CD) and prediction error correlation (r). Genomic estimated breeding values (GEBV) were predicted using a GBLUP model from a single and joint reference population. Overall, a continued increase in CD and r with a corresponding decrease in PEVD was observed as the number of common sires varies from 0 to 19 regardless of heritability levels, indicating increasing genetic connectedness between herds. Higher heritability tends to obtain stronger genetic connectedness. Compared to pedigree information, genomic relatedness inferred from genomic information increased the estimates of genetic connectedness across herds. Genomic prediction using the joint versus single reference population increased the accuracy of genomic prediction by 25% and lower heritability benefited more. Moreover, the largest benefits were observed as the number of common sires equals 0, and the gain of accuracy decreased as the number of common sires increased. We confirmed that genome-based genetic connectedness enhanced the estimates of genetic connectedness across management units. Additionally, using the combined reference population substantially increased accuracy of genomic prediction. However, care should be taken when combining reference data for closely related populations, which may give less reliable prediction results.

Within- and across-breed genomic prediction using whole-genome sequence and single nucleotide polymorphism panels.

BackgroundCurrently, genomic prediction in cattle is largely based on panels of about 54k single nucleotide polymorphisms (SNPs). However with the decreasing costs of and current advances in next-generation sequencing technologies, whole-genome sequence (WGS) data on large numbers of individuals is within reach. Availability of such data provides new opportunities for genomic selection, which need to be explored.MethodsThis simulation study investigated how much predictive ability is gained by using WGS data under scenarios with QTL (quantitative trait loci) densities ranging from 45 to 132 QTL/Morgan and heritabilities ranging from 0.07 to 0.30, compared to different SNP densities, with emphasis on divergent dairy cattle breeds with small populations. The relative performances of best linear unbiased prediction (SNP-BLUP) and of a variable selection method with a mixture of two normal distributions (MixP) were also evaluated. Genomic predictions were based on within-population, across-population, and multi-breed reference populations.ResultsThe use of WGS data for within-population predictions resulted in small to large increases in accuracy for low to moderately heritable traits. Depending on heritability of the trait, and on SNP and QTL densities, accuracy increased by up to 31 %. The advantage of WGS data was more pronounced (7 to 92 % increase in accuracy depending on trait heritability, SNP and QTL densities, and time of divergence between populations) with a combined reference population and when using MixP. While MixP outperformed SNP-BLUP at 45 QTL/Morgan, SNP-BLUP was as good as MixP when QTL density increased to 132 QTL/Morgan.ConclusionsOur results show that, genomic predictions in numerically small cattle populations would benefit from a combination of WGS data, a multi-breed reference population, and a variable selection method.

Maximizing crossbred performance through purebred genomic selection.

BackgroundIn livestock production, many animals are crossbred, with two distinct advantages: heterosis and breed complementarity. Genomic selection (GS) can be used to select purebred parental lines for crossbred performance (CP). Dominance being the likely genetic basis of heterosis, explicitly including dominance in the GS model may be an advantage to select purebreds for CP. Estimated breeding values for CP can be calculated from additive and dominance effects of alleles that are estimated using pure line data. The objective of this simulation study was to investigate the benefits of applying GS to select purebred animals for CP, based on purebred phenotypic and genotypic information. A second objective was to compare the use of two separate pure line reference populations to that of a single reference population that combines both pure lines. These objectives were investigated under two conditions, i.e. either a low or a high correlation of linkage disequilibrium (LD) phase between the pure lines.ResultsThe results demonstrate that the gain in CP was higher when parental lines were selected for CP, rather than purebred performance, both with a low and a high correlation of LD phase. For a low correlation of LD phase between the pure lines, the use of two separate reference populations yielded a higher gain in CP than use of a single reference population that combines both pure lines. However, for a high correlation of LD phase, marker effects that were estimated using a single combined reference population increased the gain in CP.ConclusionsUnder the hypothesis that performance of crossbred animals differs from that of purebred animals due to dominance, a dominance model can be used for GS of purebred individuals for CP, without using crossbred data. Furthermore, if the correlation of LD phase between pure lines is high, accuracy of selection can be increased by combining the two pure lines into a single reference population to estimate marker effects.Electronic supplementary materialThe online version of this article (doi:10.1186/s12711-015-0099-3) contains supplementary material, which is available to authorized users.

Increasing imputation and prediction accuracy for Chinese Holsteins using joint Chinese-Nordic reference population

This study investigated the effect of including Nordic Holsteins in the reference population on the imputation accuracy and prediction accuracy for Chinese Holsteins. The data used in this study include 85 Chinese Holstein bulls genotyped with both 54K chip and 777K (HD) chip, 2862 Chinese cows genotyped with 54K chip, 510 Nordic Holstein bulls genotyped with HD chip, and 4398 Nordic Holstein bulls genotyped with 54K chip and with deregressed proofs for five milk production traits. Based on these data, the accuracy of imputation from 54K to HD marker data and the accuracy of genomic predictions in Chinese Holstein were assessed. The allele correct rate increased around 2.7 and 1.7% in imputation from the 54K to the HD marker data for Chinese Holstein bulls and cows, respectively, when the Nordic HD-genotyped bulls were included in the reference data for imputation. However, the prediction accuracy was improved slightly when using the marker data imputed based on the combined HD reference data, compared with using the marker data imputed based on the Chinese HD reference data only. On the other hand, when using the combined reference population including 4398 Nordic Holstein bulls, the accuracy of genomic predictions increased 6.5 percentage points together with a reduction of prediction bias. The HD markers did not outperform the 54K markers in genomic prediction based on the present data. The results indicate that for Chinese Holsteins, it is necessary to genotype more individuals with 54K chip to increase reference population rather than increasing marker density.

Imputation of genotypes from low density (50,000 markers) to high density (700,000 markers) of cows from research herds in Europe, North America, and Australasia using 2 reference populations

Combining data from research herds may be advantageous, especially for difficult or expensive-to-measure traits (such as dry matter intake). Cows in research herds are often genotyped using low-density single nucleotide polymorphism (SNP) panels. However, the precision of quantitative trait loci detection in genome-wide association studies and the accuracy of genomic selection may increase when the low-density genotypes are imputed to higher density. Genotype data were available from 10 research herds: 5 from Europe [Denmark, Germany, Ireland, the Netherlands, and the United Kingdom (UK)], 2 from Australasia (Australia and New Zealand), and 3 from North America (Canada and the United States). Heifers from the Australian and New Zealand research herds were already genotyped at high density (approximately 700,000 SNP). The remaining genotypes were imputed from around 50,000 SNP to 700,000 using 2 reference populations. Although it was not possible to use a combined reference population, which would probably result in the highest accuracies of imputation, differences arising from using 2 high-density reference populations on imputing 50,000-marker genotypes of 583 animals (from the UK) were quantified. The European genotypes (n=4,097) were imputed as 1 data set, using a reference population of 3,150 that included genotypes from 835 Australian and 1,053 New Zealand females, with the remainder being males. Imputation was undertaken using population-wide linkage disequilibrium with no family information exploited. The UK animals were also included in the North American data set (n=1,579) that was imputed to high density using a reference population of 2,018 bulls. After editing, 591,213 genotypes on 5,999 animals from 10 research herds remained. The correlation between imputed allele frequencies of the 2 imputed data sets was high (>0.98) and even stronger (>0.99) for the UK animals that were part of each imputation data set. For the UK genotypes, 2.2% were imputed differently in the 2 high-density reference data sets used. Only 0.025% of these were homozygous switches. The number of discordant SNP was lower for animals that had sires that were genotyped. Discordant imputed SNP genotypes were most common when a large difference existed in allele frequency between the 2 imputed genotype data sets. For SNP that had ≥20% discordant genotypes, the difference between imputed data sets of allele frequencies of the UK (imputed) genotypes was 0.07, whereas the difference in allele frequencies of the (reference) high-density genotypes was 0.30. In fact, regions existed across the genome where the frequency of discordant SNP was higher. For example, on chromosome 10 (centered on 520,948 bp), 52 SNP (out of a total of 103 SNP) had ≥20% discordant SNP. Four hundred and eight SNP had more than 20% discordant genotypes and were removed from the final set of imputed genotypes. We concluded that both discordance of imputed SNP genotypes and differences in allele frequencies, after imputation using different reference data sets, may be used to identify and remove poorly imputed SNP.

Strategies for imputation to whole genome sequence using a single or multi-breed reference population in cattle.

BackgroundThe advent of low cost next generation sequencing has made it possible to sequence a large number of dairy and beef bulls which can be used as a reference for imputation of whole genome sequence data. The aim of this study was to investigate the accuracy and speed of imputation from a high density SNP marker panel to whole genome sequence level. Data contained 132 Holstein, 42 Jersey, 52 Nordic Red and 16 Brown Swiss bulls with whole genome sequence data; 16 Holstein, 27 Jersey and 29 Nordic Reds had previously been typed with the bovine high density SNP panel and were used for validation. We investigated the effect of enlarging the reference population by combining data across breeds on the accuracy of imputation, and the accuracy and speed of both IMPUTE2 and BEAGLE using either genotype probability reference data or pre-phased reference data. All analyses were done on Bovine autosome 29 using 387,436 bi-allelic variants and 13,612 SNP markers from the bovine HD panel.ResultsA combined breed reference population led to higher imputation accuracies than did a single breed reference. The highest accuracy of imputation for all three test breeds was achieved when using BEAGLE with un-phased reference data (mean genotype correlations of 0.90, 0.89 and 0.87 for Holstein, Jersey and Nordic Red respectively) but IMPUTE2 with un-phased reference data gave similar accuracies for Holsteins and Nordic Red. Pre-phasing the reference data only lead to a minor decrease in the imputation accuracy, but gave a large improvement in computation time. Pre-phasing with BEAGLE was substantially faster than pre-phasing with SHAPEIT2 (2.5 hours vs. 52 hours for 242 individuals), and imputation with pre-phased data was faster in IMPUTE2 than in BEAGLE (5 minutes vs. 50 minutes per individual).ConclusionCombining reference populations across breeds is a good option to increase the size of the reference data and in turn the accuracy of imputation when only few animals are available. Pre-phasing the reference data only slightly decreases the accuracy but gives substantial improvements in speed. Using BEAGLE for pre-phasing and IMPUTE2 for imputation is a fast and accurate strategy.

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

Using a combined multi-breed reference population, this study explored the influence of model specification and the effect of including a polygenic effect on the reliability of genomic breeding values (DGV and GEBV). The combined reference population consisted of 2986 Swedish Red Breed (SRB) and Finnish Ayrshire (FAY) dairy cattle. Bayesian methodology (common prior and mixture models with different prior distribution settings for the marker effects) as well as a best linear unbiased prediction with a genomic relationship matrix [genomic best linear unbiased predictor (GBLUP)] was used in the prediction of DGV. Mixture models including a polygenic effect were used to predict GEBV. In total, five traits with low, high and medium heritability were analysed. For the models using a mixture prior distribution, reliabilities of DGV tended to decrease with an increasing proportion of markers with small effects. The influence of the inclusion of a polygenic effect on the reliability of DGV varied across traits and model specifications. Average correlation between DGV with the Mendelian sampling term, across traits, was highest (R(2) = 0.25) for the GBLUP model and decreased with increasing proportion of markers with large effects. Reliabilities increased when DGV and parent average information were combined in an index. The GBLUP model with the largest gain across traits in the reliability of the index achieved the highest DGV mean reliability. However, the polygenic models showed to be less biased and more consistent in the estimation of DGV regardless of the model specifications compared with the mixture models without the polygenic effect.

Reliabilities of genomic prediction using combined reference data of the Nordic Red dairy cattle populations

This study investigated the possibility of increasing the reliability of direct genomic values (DGV) by combining reference populations. The data were from 3,735 bulls from Danish, Swedish, and Finnish Red dairy cattle populations. Single nucleotide polymorphism markers were fitted as random variables in a Bayesian model, using published estimated breeding values as response variables. In total, 17 index traits were analyzed. Reliabilities were estimated using a 5-fold cross validation, and calculated as the within-year squared correlation between estimated breeding values and DGV. Marker effects were estimated using reference populations from individual countries, as well as using a combined reference population from all 3 countries. Single-country reference populations gave mean reliabilities across 17 traits of 0.19 to 0.23, whereas the combined reference gave mean reliabilities of 0.26 for all populations. Using marker effects from 1 population to predict the other 2 gave a loss in mean reliability of 0.14 to 0.21 when predicting Swedish or Finnish animals with Danish marker effects, or vice versa. Using Swedish or Finnish marker effects to predict each other only showed a loss in mean reliability of 0.03 to 0.05. A combined Swedish-Finnish reference population led to an average reliability as high as that from the 3-country reference population, but somewhat different for individual traits. The results from this study show that it is possible to increase the reliability of DGV by combining reference populations from related populations.