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

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.

An effective hyper-parameter can increase the prediction accuracy in a single-step genetic evaluation

The H-matrix best linear unbiased prediction (HBLUP) method has been widely used in livestock breeding programs. It can integrate all information, including pedigree, genotypes, and phenotypes on both genotyped and non-genotyped individuals into one single evaluation that can provide reliable predictions of breeding values. The existing HBLUP method requires hyper-parameters that should be adequately optimised as otherwise the genomic prediction accuracy may decrease. In this study, we assess the performance of HBLUP using various hyper-parameters such as blending, tuning, and scale factor in simulated and real data on Hanwoo cattle. In both simulated and cattle data, we show that blending is not necessary, indicating that the prediction accuracy decreases when using a blending hyper-parameter <1. The tuning process (adjusting genomic relationships accounting for base allele frequencies) improves prediction accuracy in the simulated data, confirming previous studies, although the improvement is not statistically significant in the Hanwoo cattle data. We also demonstrate that a scale factor, α, which determines the relationship between allele frequency and per-allele effect size, can improve the HBLUP accuracy in both simulated and real data. Our findings suggest that an optimal scale factor should be considered to increase prediction accuracy, in addition to blending and tuning processes, when using HBLUP.

Genomic prediction in pigs using data from a commercial crossbred population: insights from the Duroc x (Landrace x Yorkshire) three-way crossbreeding system

BACKGROUND

Genomic selection is widely applied for genetic improvement in livestock crossbreeding systems to select excellent nucleus purebred (PB) animals and to improve the performance of commercial crossbred (CB) animals. Most current predictions are based solely on PB performance. Our objective was to explore the potential application of genomic selection of PB animals using genotypes of CB animals with extreme phenotypes in a three-way crossbreeding system as the reference population. Using real genotyped PB as ancestors, we simulated the production of 100,000 pigs for a Duroc x (Landrace x Yorkshire) DLY crossbreeding system. The predictive performance of breeding values of PB animals for CB performance using genotypes and phenotypes of (1) PB animals, (2) DLY animals with extreme phenotypes, and (3) random DLY animals for traits of different heritabilities ([Formula: see text] = 0.1, 0.3, and 0.5) was compared across different reference population sizes (500 to 6500) and prediction models (genomic best linear unbiased prediction (GBLUP) and Bayesian sparse linear mixed model (BSLMM)).

RESULTS

Using a reference population consisting of CB animals with extreme phenotypes showed a definite predictive advantage for medium- and low-heritability traits and, in combination with the BSLMM model, significantly improved selection response for CB performance. For high-heritability traits, the predictive performance of a reference population of extreme CB phenotypes was comparable to that of PB phenotypes when the effect of the genetic correlation between PB and CB performance ([Formula: see text]) on the accuracy obtained with a PB reference population was considered, and the former could exceed the latter if the reference size was large enough. For the selection of the first and terminal sires in a three-way crossbreeding system, prediction using extreme CB phenotypes outperformed the use of PB phenotypes, while the optimal design of the reference group for the first dam depended on the percentage of individuals from the corresponding breed that the PB reference data comprised and on the heritability of the target trait.

CONCLUSIONS

A commercial crossbred population is promising for the design of the reference population for genomic prediction, and selective genotyping of CB animals with extreme phenotypes has the potential for maximizing genetic improvement for CB performance in the pig industry.

Selective genotyping strategies for a sib test scheme of a broiler breeder program

BACKGROUND

In broiler breeding, genotype-by-environment interaction is known to result in a genetic correlation between body weight measured in bio-secure and commercial environments that is substantially less than 1. Thus, measuring body weights on sibs of selection candidates in a commercial environment and genotyping them could increase genetic progress. Using real data, the aim of this study was to evaluate which genotyping strategy and which proportion of sibs placed in the commercial environment should be genotyped to optimize a sib-testing breeding program in broilers. Phenotypic body weight and genomic information were collected on all sibs raised in a commercial environment, which allowed to retrospectively analyze different sampling strategies and genotyping proportions.

RESULTS

Accuracies of genomic estimated breeding values (GEBV) obtained with the different genotyping strategies were assessed by computing their correlation with GEBV obtained when all sibs in the commercial environment were genotyped. Results showed that, compared to random sampling (RND), genotyping sibs with extreme phenotypes (EXT) resulted in higher GEBV accuracy across all genotyping proportions, especially for genotyping proportions of 12.5% or 25%, which resulted in correlations of 0.91 vs 0.88 for 12.5% and 0.94 vs 0.91 for 25% genotyped. Including pedigree on birds with phenotype in the commercial environment that were not genotyped increased accuracy at lower genotyping proportions, especially for the RND strategy (correlations of 0.88 vs 0.65 at 12.5% and 0.91 vs 0.80 at 25%), and a smaller but still substantial increase in accuracy for the EXT strategy (0.91 vs 0.79 for 12.5% and 0.94 vs 0.88 for 25% genotyped). Dispersion bias was virtually absent for RND if 25% or more birds were genotyped. However, GEBV were considerably inflated for EXT, especially when the proportion genotyped was low, which was further exacerbated if the pedigree of non-genotyped sibs was excluded.

CONCLUSIONS

When less than 75% of all animals placed in a commercial environment are genotyped, it is recommended to use the EXT strategy, because it yields the highest accuracy. However, caution should be taken when interpreting the resulting GEBV because they will be over-dispersed. When 75% or more of the animals are genotyped, random sampling is recommended because it yields virtually no bias of GEBV and results in similar accuracies as the EXT strategy.

First large-scale genomic prediction in the honey bee

Genomic selection has increased genetic gain in several livestock species, but due to the complicated genetics and reproduction biology not yet in honey bees. Recently, 2970 queens were genotyped to gather a reference population. For the application of genomic selection in honey bees, this study analyzes the accuracy and bias of pedigree-based and genomic breeding values for honey yield, three workability traits, and two traits for resistance against the parasite Varroa destructor. For breeding value estimation, we use a honey bee-specific model with maternal and direct effects, to account for the contributions of the workers and the queen of a colony to the phenotypes. We conducted a validation for the last generation and a five-fold cross-validation. In the validation for the last generation, the accuracy of pedigree-based estimated breeding values was 0.12 for honey yield, and ranged from 0.42 to 0.61 for the workability traits. The inclusion of genomic marker data improved these accuracies to 0.23 for honey yield, and a range from 0.44 to 0.65 for the workability traits. The inclusion of genomic data did not improve the accuracy of the disease-related traits. Traits with high heritability for maternal effects compared to the heritability for direct effects showed the most promising results. For all traits except the Varroa resistance traits, the bias with genomic methods was on a similar level compared to the bias with pedigree-based BLUP. The results show that genomic selection can successfully be applied to honey bees.

The effect of high-density genotypic data and different methods on joint genomic prediction: A case study in large white pigs

Joint genomic prediction (GP) is an attractive method to improve the accuracy of GP by combining information from multiple populations. However, many factors can negatively influence the accuracy of joint GP, such as differences in linkage disequilibrium phasing between single nucleotide polymorphisms (SNPs) and causal variants, minor allele frequencies and causal variants' effect sizes across different populations. The objective of this study was to investigate whether the imputed high-density genotype data can improve the accuracy of joint GP using genomic best linear unbiased prediction (GBLUP), single-step GBLUP (ssGBLUP), multi-trait GBLUP (MT-GBLUP) and GBLUP based on genomic relationship matrix considering heterogenous minor allele frequencies across different populations (wGBLUP). Three traits, including days taken to reach slaughter weight, backfat thickness and loin muscle area, were measured on 67 276 Large White pigs from two different populations, for which 3334 were genotyped by SNP array. The results showed that a combined population could substantially improve the accuracy of GP compared with a single-population GP, especially for the population with a smaller size. The imputed SNP data had no effect for single population GP but helped to yield higher accuracy than the medium-density array data for joint GP. Of the four methods, ssGLBUP performed the best, but the advantage of ssGBLUP decreased as more individuals were genotyped. In some cases, MT-GBLUP and wGBLUP performed better than GBLUP. In conclusion, our results confirmed that joint GP could be beneficial from imputed high-density genotype data, and the wGBLUP and MT-GBLUP methods are promising for joint GP in pig 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.

Simulation studies to optimize genomic selection in honey bees

Background

With the completion of a single nucleotide polymorphism (SNP) chip for honey bees, the technical basis of genomic selection is laid. However, for its application in practice, methods to estimate genomic breeding values need to be adapted to the specificities of the genetics and breeding infrastructure of this species. Drone-producing queens (DPQ) are used for mating control, and usually, they head non-phenotyped colonies that will be placed on mating stations. Breeding queens (BQ) head colonies that are intended to be phenotyped and used to produce new queens. Our aim was to evaluate different breeding program designs for the initiation of genomic selection in honey bees.

Methods

Stochastic simulations were conducted to evaluate the quality of the estimated breeding values. We developed a variation of the genomic relationship matrix to include genotypes of DPQ and tested different sizes of the reference population. The results were used to estimate genetic gain in the initial selection cycle of a genomic breeding program. This program was run over six years, and different numbers of genotyped queens per year were considered. Resources could be allocated to increase the reference population, or to perform genomic preselection of BQ and/or DPQ.

Results

Including the genotypes of 5000 phenotyped BQ increased the accuracy of predictions of breeding values by up to 173%, depending on the size of the reference population and the trait considered. To initiate a breeding program, genotyping a minimum number of 1000 queens per year is required. In this case, genetic gain was highest when genomic preselection of DPQ was coupled with the genotyping of 10–20% of the phenotyped BQ. For maximum genetic gain per used genotype, more than 2500 genotyped queens per year and preselection of all BQ and DPQ are required.

Conclusions

This study shows that the first priority in a breeding program is to genotype phenotyped BQ to obtain a sufficiently large reference population, which allows successful genomic preselection of queens. To maximize genetic gain, DPQ should be preselected, and their genotypes included in the genomic relationship matrix. We suggest, that the developed methods for genomic prediction are suitable for implementation in genomic honey bee breeding programs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12711-021-00654-x.

Accuracy of GEBV of sires based on pooled allele frequency of their progeny

Despite decreasing genotyping costs, in some cases individually genotyping animals is not economically feasible (e.g., in small ruminants). An alternative is to pool DNA, using the pooled allele frequency (PAF) to garner information on performance. Still, the use of PAF for prediction (estimation of genomic breeding values; GEBVs) has been limited. Two potential sources of error on accuracy of GEBV of sires, obtained from PAF of their progeny themselves lacking pedigree information, were tested: (i) pool construction error (unequal contribution of DNA from animals in pools), and (ii) technical error (variability when reading the array). Pooling design (random, extremes, K-means), pool size (5, 10, 25, 50, and 100 individuals), and selection scenario (random, phenotypic) also were considered. These factors were tested by simulating a sheep population. Accuracy of GEBV-the correlation between true and estimated values-was not substantially affected by pool construction or technical error, or selection scenario. A significant interaction, however, between pool size and design was found. Still, regardless of design, mean accuracy was higher for pools of 10 or less individuals. Mean accuracy of GEBV was 0.174 (SE 0.001) for random pooling, and 0.704 (SE 0.004) and 0.696 (SE 0.004) for extreme and K-means pooling, respectively. Non-random pooling resulted in moderate accuracy of GEBV. Overall, pooled genotypes can be used in conjunction with individual genotypes of sires for moderately accurate predictions of their genetic merit with little effect of pool construction or technical error.

Genomic Prediction in Local Breeds: The Rendena Cattle as a Case Study

Simple Summary

Although genomic selection is being used in many livestock species, it has not yet been considered in local breeds due to the lower population size and the potential less effective impact on the genetic evaluation of these breeds. The current research aims to investigate how genomic data can impact the accuracy of genetic predictions for beef traits in Rendena, a small local cattle breed of the North-East of Italy selected for a dual purpose. Classical animal models using only phenotypic information were compared with two models that integrated genomic data with pedigree information. The genomic models presented better accuracy in estimated breeding values of the animals than the ‘classical’ animal model, especially the ‘simpler’ one assuming homogeneous variances of single nucleotide polymorphisms. Our results show that the inclusion of genomic information can be successfully applied to breeding selection scenarios even in small local cattle breeds such as Rendena.

Abstract

The maintenance of local cattle breeds is key to selecting for efficient food production, landscape protection, and conservation of biodiversity and local cultural heritage. Rendena is an indigenous cattle breed from the alpine North-East of Italy, selected for dual purpose, but with lesser emphasis given to beef traits. In this situation, increasing accuracy for beef traits could prevent detrimental effects due to the antagonism with milk production. Our study assessed the impact of genomic information on estimated breeding values (EBVs) in Rendena performance-tested bulls. Traits considered were average daily gain, in vivo EUROP score, and in vivo estimate of dressing percentage. The final dataset contained 1691 individuals with phenotypes and 8372 animals in pedigree, 1743 of which were genotyped. Using the cross-validation method, three models were compared: (i) Pedigree-BLUP (PBLUP); (ii) single-step GBLUP (ssGBLUP), and (iii) weighted single-step GBLUP (WssGBLUP). Models including genomic information presented higher accuracy, especially WssGBLUP. However, the model with the best overall properties was the ssGBLUP, showing higher accuracy than PBLUP and optimal values of bias and dispersion parameters. Our study demonstrated that integrating phenotypes for beef traits with genomic data can be helpful to estimate EBVs, even in a small local breed.

Selective genotyping and phenotypic data inclusion strategies of crossbred progeny for combined crossbred and purebred selection in swine breeding

Inclusion of crossbred (CB) data into traditionally purebred (PB) genetic evaluations has been shown to increase the response in CB performance. Currently, it is unrealistic to collect data on all CB animals in swine production systems, thus, a subset of CB animals must be selected to contribute genomic/phenotypic information. The aim of this study was to evaluate selective genotyping strategies in a simulated 3-way swine crossbreeding scheme. The swine crossbreeding scheme was simulated and produced 3-way CB animals for 6 generations with 3 distinct PB breeds each with 25 and 175 mating males and females, respectively. F1 crosses (400 mating females) produced 4,000 terminal CB progeny which were subjected to selective genotyping. The genome consisted of 18 chromosomes with 1,800 QTL and 72k SNP markers. Selection was performed using estimated breeding values (EBV) for CB performance. It was assumed that both PB and CB performance was moderately heritable (h2=0.4). Several scenarios altering the genetic correlation between PB and CB performance (rpc=0.1, 0.3, 0.5, 0.7,or 0.9) were considered. CB animals were chosen based on phenotypes to select 200, 400, or 800 CB animals to genotype per generation. Selection strategies included: (1) Random: random selection, (2) Top: highest phenotype, (3) Bottom: lowest phenotype, (4) Extreme: half highest and half lowest phenotypes, and (5) Middle: average phenotype. Each selective genotyping strategy, except for Random, was considered by selecting animals in half-sib (HS) or full-sib (FS) families. The number of PB animals with genotypes and phenotypes each generation was fixed at 1,680. Each unique genotyping strategy and rpc scenario was replicated 10 times. Selection of CB animals based on the Extreme strategy resulted in the highest (P < 0.05) rates of genetic gain in CB performance (ΔG) when rpc<0.9. For highly correlated traits (rpc=0.9) selective genotyping did not impact (P > 0.05) ΔG. No differences (P > 0.05) were observed in ΔG between top, bottom, or middle when rpc>0.1. Higher correlations between true breeding values (TBV) and EBV were observed using Extreme when rpc<0.9. In general, family sampling method did not impact ΔG or the correlation between TBV and EBV. Overall, the Extreme genotyping strategy produced the greatest genetic gain and the highest correlations between TBV and EBV, suggesting that 2-tailed sampling of CB animals is the most informative when CB performance is the selection goal.

Bias in genomic predictions by mating practices for linear type traits in a large-scale genomic evaluation

The objective of this study was to clarify how bias in genomic predictions is created by investigating a relationship among selection intensity, a change in heritability (Δh²), and assortative mating (ASM). A change in heritability, resulting from selection, reflects the impact that the Bulmer effect has on the reduction in between-family variation, whereas assortative mating impacts the within-family variance or Mendelian sampling variation. A partial data set up to 2014, including 841K genotyped animals, was used to calculate genomic predictions with a single-step genomic model for 18 linear type traits in US Holsteins. A full data set up to 2018, including 2.3 million genotyped animals, was used to calculate benchmark genomic predictions. Inbreeding and unknown parent groups for missing parents of animals were included in the model. Genomic evaluation was performed using 2 different genetic parameters: those estimated 14 yr ago, which have been used in the national genetic evaluation for linear type traits in the United States, and those newly estimated with recent records from 2015 to 2018 and those corresponding pedigrees. Genetic trends for 18 type traits were estimated for bulls with daughters and cows with phenotypes in 2018. Based on selection intensity and mating decisions, these traits can be categorized into 3 groups: (a) high directional selection, (b) moderate selection, and (c) intermediate optimum selection. The first 2 categories can be explained by positive assortative mating, and the last can be explained by negative assortative or disassortative mating. Genetic progress was defined by genetic gain per year based on average standardized genomic predictions for cows from 2000 to 2014. Traits with more genetic progress tended to have more "inflated" genomic predictions (i.e., "inflation" means here that genomic predictions are larger in absolute values than expected, whereas "deflation" means smaller than expected). Heritability estimates for 14 out of 18 traits declined in the last 16 yr, and Δh² ranged from -0.09 to 0.04. Traits with a greater decline in heritability tended to have more deflated genomic predictions. Biases (inflation or deflation) in genomic predictions were not improved by using the latest genetic parameters, implying that bias in genomic predictions due to preselection was not substantial for a large-scale genomic evaluation. Moreover, the strong selection intensity was not fully responsible for bias in genomic predictions. The directional selection can decrease heritability; however, positive assortative mating, which was strongly associated with large genetic gains, could minimize the decline in heritability for a trait under strong selection and could affect bias in genomic predictions.

Phenotypically Selective Genotyping Realizes More Genetic Gains in a Rainbow Trout Breeding Program in the Presence of Genotype-by-Environment Interactions

Selective genotyping of phenotypically superior animals may lead to bias and less accurate genomic breeding values (GEBV). Performing selective genotyping based on phenotypes measured in the breeding environment (B) is not necessarily a good strategy when the aim of a breeding program is to improve animals’ performance in the commercial environment (C). Our simulation study compared different genotyping strategies for selection candidates and for fish in C in a breeding program for rainbow trout in the presence of genotype-by-environment interactions when the program had limited genotyping resources and unregistered pedigrees of individuals. For the reference population, selective genotyping of top and bottom individuals in C based on phenotypes measured in C led to the highest genetic gains, followed by random genotyping and then selective genotyping of top individuals in C. For selection candidates, selective genotyping of top individuals in B based on phenotypes measured in B led to the highest genetic gains, followed by selective genotyping of top and bottom individuals and then random genotyping. Selective genotyping led to bias in predicting GEBV. However, in scenarios that used selective genotyping of top fish in B and random genotyping of fish in C, predictions of GEBV were unbiased, with genetic correlations of 0.2 and 0.5 between traits measured in B and C. Estimates of variance components were sensitive to genotyping strategy, with an overestimation of the variance with selective genotyping of top and bottom fish and an underestimation of the variance with selective genotyping of top fish. Unbiased estimates of variance components were obtained when fish in B and C were genotyped at random. In conclusion, we recommend phenotypic genotyping of top and bottom fish in C and top fish in B for the purpose of selecting breeding animals and random genotyping of individuals in B and C for the purpose of estimating variance components when a genomic breeding program for rainbow trout aims to improve animals’ performance in C.