Application Of Genomic Selection Research Articles

The development and application of genomic selection as a new breeding paradigm

The development and application of genomic selection as a new breeding paradigm

Genomic selection using a realized genomic relationship matrix in a Pinus taeda L. cloned population

Genetic merit can be considered the finite sum of thousands of allelic effects, each physically located at some place on the genome, whose transmission can be traced through molecular markers. Traditionally, best linear unbiased prediction (BLUP) of breeding values relies on average additive genetic covariances (the numerator relationship matrix A) derived from pedigrees to utilize information from relatives. For example, all pairs of full-sib offspring of a cross are assumed to share 50% of alleles in common. Such assumptions ignore variation in Mendelian segregation of alleles among progeny within family. With advances in marker genotyping technology and reduction in genotyping cost, it is now feasible to estimate genetic covariances from markers. Linear mixed models that utilize realized genomic relationship matrices could predict genomic estimated breeding values (GEBV) more accurately than those that use expected average genetic covariances derived from pedigrees. Dense markers can be used to trace identity by descent probabilities at each locus, and those probabilities used to construct an incidence matrix. The incidence matrix is used to estimate the genomic relationship matrix (G), which is used in place of the A matrix in solving the mixed model equations. This may allow more accurate estimation of individual breeding values than the traditional model based on average genetic covariances. We estimated realized genetic covariances between cloned progeny of a P. taeda population. There were 165 cloned progeny derived from nine full-sib families. The realized genomic relationships were based on a set of 3,461 biallelic SNP markers. We used the following linear mixed model y = Xb + Zu + e to estimate GEBV. In the model X and Z are incidence matrices, b is the vector of fixed mean, u is the vector of additive genetic effects that correspond to allele substitution effects for each marker with Var (u) = Iσ2m; where σ2m is the marker variance and I is the identity matrix. The term e is the vector of residuals. The dimension of Z is the number of individuals (n) by the number of loci (m). The regression method used to construct our G matrix did not require allele frequencies; instead, the inverse of the G matrix was generated by regressing ZZ’ as a dependent variable on the A matrix as the independent variable. Therefore, the expected value of G is A plus a constant matrix. Different cross-validation methods were used to test performance of the G matrix. Clones were divided into a training group with both marker and phenotypic information and a validation group for which only marker genotypes were used. In one scenario ~90% of the clones (148) were sampled for training, either randomly selecting within each of the nine families or at random without family consideration. The remaining ~10% were used for validation (17 clones). In the second scenario, ~50% of clones (84) were sampled either within family or randomly from the whole population for training, and the remaining ~50% were used for validation (81 clones). The model parameters estimated in the training set were used to predict GEBV in the validation set. For each scenario, six independent samplings were carried out. The mean correlation between the GEBV based on G-BLUP and breeding values based on A-BLUP were determined for each scenario, along with the accuracy of the BLUP predictions for both G and A based models. The mean correlation varied from 0.37 to 0.61 across the four validation methods. The accuracies of the predictions for any validation scenario were always higher for G-BLUP (range of 0.65 to 0.75) than A-BLUP (0.60 to 0.62), which is related to the smaller standard error of the predicted G-BLUP for the validation clones (17 or 81) under the different scenarios. Estimating realized genetic covariances based on the genotypes of biallelic markers and incorporating those estimates into G-BLUP helps to more accurately characterize Mendelian segregation of alleles, and could allow more accurate selection within family. Such a method would increase genetic gains in forest tree breeding. The major impact would be on reducing the need for expensive field testing, but it may also be possible to shorten the breeding cycle and thus increase genetic gain per unit time and cost. The impact of genomic selection applications in forest tree breeding may be greater than for other crop or animal species, because of the biology of trees and their much longer breeding cycles.

The impact of the rank of marker variance–covariance matrix in principal component evaluation for genomic selection applications

In genomic selection (GS) programmes, direct genomic values (DGV) are evaluated using information provided by high-density SNP chip. Being DGV accuracy strictly dependent on SNP density, it is likely that an increase in the number of markers per chip will result in severe computational consequences. Aim of present work was to test the effectiveness of principal component analysis (PCA) carried out by chromosome in reducing the marker dimensionality for GS purposes. A simulated data set of 5700 individuals with an equal number of SNP distributed over six chromosomes was used. PCs were extracted both genome-wide (ALL) and separately by chromosome (CHR) and used to predict DGVs. In the ALL scenario, the SNP variance-covariance matrix (S) was singular, positive semi-definite and contained null information which introduces 'spuriousness' in the derived results. On the contrary, the S matrix for each chromosome (CHR scenario) had a full rank. Obtained DGV accuracies were always better for CHR than ALL. Moreover, in the latter scenario, DGV accuracies became soon unsettled as the number of animals decreases, whereas in CHR, they remain stable till 900-1000 individuals. In real applications where a 54k SNP chip is used, the largest number of markers per chromosome is approximately 2500. Thus, a number of around 3000 genotyped animals could lead to reliable results when the original SNP variables are replaced by a reduced number of PCs.

Optimal strategies for the use of genomic selection in dairy cattle breeding programs

The objective of the present study was to conduct a stochastic simulation study on the possible benefits of an application of genomic selection in dairy cattle breeding programs according to a variety of selection schemes. In addition, the heritability of the trait in question, the accuracy of genomic breeding values, and the number of animals to be genotyped were varied. Specifically, the question of genotyping males, females, or both, was addressed. Selection schemes were compared with a young bull breeding program. The main criterion for comparison was the average of true breeding values of selected young males to be used as replacements for artificial insemination bulls. Stochastic simulations were run with 50 repetitions each to generate individuals with phenotypes, breeding values estimated by BLUP, and true breeding values. Genomic breeding values were generated from true breeding values with defined accuracy. Examined scenarios included a group of selection schemes that featured genotyping of parents of future bulls only. Such schemes can be viewed as improvements of young bull programs, and they were found to be competitive with or superior to a classical young bull program. However, a genomic breeding program usually involves at least genotyping young male candidates. A second group of selection schemes reflected this requirement. Scenarios in this group were found to be superior over the young bull program by 1.0 to 1.2 standard deviations of the average true breeding value of young male candidates. Within this group of scenarios, one scheme referred to an ideal situation under which genotypes for male calves were available without limitation. Using the average of true breeding values as the criterion for comparison, this idealistic scenario was competitive with other scenarios only if the reliability of genomic breeding values was larger than 0.50. Conventionally, not all males available will have genotypes, and the 2 most promising scenarios included a preselection step for dams of future bulls. This preselection step can be based on conventional BLUP estimated breeding values for bull dams, because differences with a scheme under which both parents and the resulting male offspring are genotyped were marginal. Genotyping of young male candidates should be the focus of activities of today's breeding organizations.

Review. Promises, pitfalls and challenges of genomic selection in breeding programs

The aim of this work was to review the main challenges and pitfalls of the implementation of genomic selection in the breeding programs of different livestock species. Genomic selection is now one of the main challenges in animal breeding and genetics. Its application could considerably increase the genetic gain in traits of interest. However, the success of its practical implementation depends on the selection scheme characteristics, and these must be studied for each particular case. In dairy cattle, especially in Holsteins, genomic selection is a reality. However, in other livestock species (beef cattle, small ruminants, monogastrics and fish) genomic selection has mainly been used experimentally.The main limitation for its implementation in the mentioned livestock species is the high genotyping costs compared to the low selection value of the candidate. Nevertheless, nowadays the possibility of using single-nucleotide polymorphism (SNP) chips of low density to make genomic selection applications economically feasible is under study. Economic studies may optimize the benefits of genomic selection (GS) to include new traits in the breeding goals. It is evident that genomic selection offers great potential; however, a suitable genotyping strategy and recording system for each case is needed in order to properly exploit it.

L2-Boosting algorithm applied to high-dimensional problems in genomic selection

The L(2)-Boosting algorithm is one of the most promising machine-learning techniques that has appeared in recent decades. It may be applied to high-dimensional problems such as whole-genome studies, and it is relatively simple from a computational point of view. In this study, we used this algorithm in a genomic selection context to make predictions of yet to be observed outcomes. Two data sets were used: (1) productive lifetime predicted transmitting abilities from 4702 Holstein sires genotyped for 32 611 single nucleotide polymorphisms (SNPs) derived from the Illumina BovineSNP50 BeadChip, and (2) progeny averages of food conversion rate, pre-corrected by environmental and mate effects, in 394 broilers genotyped for 3481 SNPs. Each of these data sets was split into training and testing sets, the latter comprising dairy or broiler sires whose ancestors were in the training set. Two weak learners, ordinary least squares (OLS) and non-parametric (NP) regression were used for the L2-Boosting algorithm, to provide a stringent evaluation of the procedure. This algorithm was compared with BL [Bayesian LASSO (least absolute shrinkage and selection operator)] and BayesA regression. Learning tasks were carried out in the training set, whereas validation of the models was performed in the testing set. Pearson correlations between predicted and observed responses in the dairy cattle (broiler) data set were 0.65 (0.33), 0.53 (0.37), 0.66 (0.26) and 0.63 (0.27) for OLS-Boosting, NP-Boosting, BL and BayesA, respectively. The smallest bias and mean-squared errors (MSEs) were obtained with OLS-Boosting in both the dairy cattle (0.08 and 1.08, respectively) and broiler (-0.011 and 0.006) data sets, respectively. In the dairy cattle data set, the BL was more accurate (bias=0.10 and MSE=1.10) than BayesA (bias=1.26 and MSE=2.81), whereas no differences between these two methods were found in the broiler data set. L2-Boosting with a suitable learner was found to be a competitive alternative for genomic selection applications, providing high accuracy and low bias in genomic-assisted evaluations with a relatively short computational time.

Increase in accuracy using multi-trait genomic breeding value estimation

Introduction Genomic selection is becoming common practice in animal breeding. It uses genome-wide dense marker maps, to accurately predict the genetic ability of animals, without the need to record phenotypic performance from the animal itself or from close relatives. Presented applications of genomic selection have mainly been limited to implementations where genomic breeding values are estimated using single trait models. A major breakthrough in traditional breeding value estimation was the application of multi-trait breeding value estimation, for instance to combine mastitis and somatic cell count information. Therefore, our objective was to develop multi-trait genomic breeding value estimation methods.

Whole genome marker-assisted selection.

Abstract Most applications of genomic selection (GS) have so far been in animals, especially dairy cattle, although theoretical studies have also been conducted for maize. For the last 50 years, commercial breeding programmes for dairy cattle have been based on the progeny test scheme, which results in an average generation interval of approximately 7 years for the sire to dam path. It should be possible to double rates of genetic gain by the application of GS, if generation intervals are reduced to close to the biological minimum. During the last decade, methods were developed for high-throughput genotyping of thousands of single nucleotide polymorphisms per individuals. The number of potential polymorphic markers per species was increased from about 1000 to tens of thousands, and costs per individual marker genotype were reduced from several dollars to less than one cent. The application of marker-assisted selection based on genome-wide association studies requires solutions to new statistical problems. Specifically how should information from pedigree, phenotypic records and genotypes be combined to optimally rank candidates for selection? Various linear and Bayesian methods have been proposed and tested to compute genomic estimated breeding values. Linear models require significantly less computing time, and perform nearly as well as Bayesian methodologies. Methods are generally evaluated by comparing genomic evaluations based only on pedigree and genotype to genetic evaluations based on daughter records of the same bulls. Genomic selection programmes for dairy cattle have been implemented in the USA, Canada, Australia, New Zealand and the Netherlands. With declining genotyping costs it becomes economically viable to genotype more individuals, including candidate bull dams. More emphasis can be placed on low heritability traits, such as fertility and disease traits, and it will be easier to control the increase in inbreeding in commercial animal populations.

Accuracy of breeding values of 'unrelated' individuals predicted by dense SNP genotyping

BackgroundRecent developments in SNP discovery and high throughput genotyping technology have made the use of high-density SNP markers to predict breeding values feasible. This involves estimation of the SNP effects in a training data set, and use of these estimates to evaluate the breeding values of other 'evaluation' individuals. Simulation studies have shown that these predictions of breeding values can be accurate, when training and evaluation individuals are (closely) related. However, many general applications of genomic selection require the prediction of breeding values of 'unrelated' individuals, i.e. individuals from the same population, but not particularly closely related to the training individuals.MethodsAccuracy of selection was investigated by computer simulation of small populations. Using scaling arguments, the results were extended to different populations, training data sets and genome sizes, and different trait heritabilities.ResultsPrediction of breeding values of unrelated individuals required a substantially higher marker density and number of training records than when prediction individuals were offspring of training individuals. However, when the number of records was 2*Ne*L and the number of markers was 10*Ne*L, the breeding values of unrelated individuals could be predicted with accuracies of 0.88 – 0.93, where Ne is the effective population size and L the genome size in Morgan. Reducing this requirement to 1*Ne*L individuals, reduced prediction accuracies to 0.73–0.83.ConclusionFor livestock populations, 1NeL requires about ~30,000 training records, but this may be reduced if training and evaluation animals are related. A prediction equation is presented, that predicts accuracy when training and evaluation individuals are related. For humans, 1NeL requires ~350,000 individuals, which means that human disease risk prediction is possible only for diseases that are determined by a limited number of genes. Otherwise, genotyping and phenotypic recording need to become very common in the future.

Performance of Genomic Selection in Mice

Selection plans in plant and animal breeding are driven by genetic evaluation. Recent developments suggest using massive genetic marker information, known as "genomic selection." There is little evidence of its performance, though. We empirically compared three strategies for selection: (1) use of pedigree and phenotypic information, (2) use of genomewide markers and phenotypic information, and (3) the combination of both. We analyzed four traits from a heterogeneous mouse population (http://gscan.well.ox.ac.uk/), including 1884 individuals and 10,946 SNP markers. We used linear mixed models, using extensions of association analysis. Cross-validation techniques were used, providing assumption-free estimates of predictive ability. Sampling of validation and training data sets was carried out across and within families, which allows comparing across- and within-family information. Use of genomewide genetic markers increased predictive ability up to 0.22 across families and up to 0.03 within families. The latter is not statistically significant. These values are roughly comparable to increases of up to 0.57 (across family) and 0.14 (within family) in accuracy of prediction of genetic value. In this data set, within-family information was more accurate than across-family information, and populational linkage disequilibrium was not a completely accurate source of information for genetic evaluation. This fact questions some applications of genomic selection.