Best Linear Unbiased Prediction Selection Research Articles

Consequences of incorrect genetic parameter estimates for single-trait and multi-trait genetic evaluations in honeybees.

Genetic and residual variances of traits are important input parameters for best linear unbiased prediction (BLUP) breeding value estimation. In honeybees, estimates of these variances are often associated with large standard errors, entailing a risk to perform genetic evaluations under wrong premises. The consequences hereof have not been sufficiently studied. In particular, there are no adequate investigations on this topic accounting for multi-trait selection or genetic peculiarities of the honeybee. We performed simulation studies and explored the consequences of selection for honeybee populations with a broad range of true and assumed genetic parameters. We found that in single-trait evaluations, the response to selection was barely compromised by assuming erroneous parameters, so that reductions in genetic progress after 20 years never exceeded 21%. Phenotypic selection appeared inferior to BLUP selection, particularly under low heritabilities. Parameter choices for genetic evaluation had great effects on inbreeding development. By wrongly assuming high heritabilities, inbreeding rates were reduced by up to 74%. When parallel selection was performed for two traits, the right choice of genetic parameters appeared considerably more crucial as several incorrect premises yielded inadvertent negative selection for one of the traits. This phenomenon occurred in multiple constellations in which the selection traits expressed a negative genetic correlation. It was not reflected in the estimated breeding values. Our results indicate that breeding efforts heavily rely on detailed knowledge on genetic parameters, particularly when multi-trait selection is performed. Thus, considerable effort should be invested into precise parameter estimations.

Genetic evaluation of early egg production and maturation traits using two different approaches in Japanese quail

The objective of the current study was to evaluate a multi-trait selection program based on aggregated breeding values using an animal model Best Linear Unbiased Prediction (BLUP) in Japanese quail. The estimated genetic gain was compared by both mixed model and least squares methods. Data of 1,682 female Japanese quails were collected through four consecutive generations to estimate genetic gain, depending on aggregated breeding values, for age at first egg (AFE), body weight at sexual maturity (BWSM), and days needed to produce the first ten eggs (DN10). Estimates of cumulative selection response were favorable for all the studied traits and significant for AFE (-3.03) and BWSM(10.38), but not significant for DN10(-0.15). Estimates of direct heritability were moderate for AFE (0.21) and BWSM(0.25) but low for DN10(0.08), while estimates of maternal heritability were moderate for AFE (0.19) but low for BWSM(0.04) and DN10(0.01). High (0.45 to 0.56) genetic and low (-0.01 to -0.18) phenotypic correlations were observed among the studied traits. Negative (-0.23 to -0.95) correlations between additive genetic and maternal genetic effects were observed for all traits. Genetic trends were -0.76 (P=0.031), 2.54 (P=0.037), and -0.06 (P=0.052) with calculated product-moment correlations between breeding values, estimated by BLUP and phenotypic selection methods, of 0.78 (P=0.002), 0.77 (P=0.004), and 0.61 (P=0.007) for AFE, BWSM, and DN10, respectively. Aggregated breeding value estimation based on animal model BLUP could be an effective method of constructing a selection program to achieve a favorable selection response in egg production traits in Japanese quail.

Systematic differences in the response of genetic variation to pedigree and genome-based selection methods.

Genomic selection (GS) is a DNA-based method of selecting for quantitative traits in animal and plant breeding, and offers a potentially superior alternative to traditional breeding methods that rely on pedigree and phenotype information. Using a 60 K SNP chip with markers spaced throughout the entire chicken genome, we compared the impact of GS and traditional BLUP (best linear unbiased prediction) selection methods applied side-by-side in three different lines of egg-laying chickens. Differences were demonstrated between methods, both at the level and genomic distribution of allele frequency changes. In all three lines, the average allele frequency changes were larger with GS, 0.056 0.064 and 0.066, compared with BLUP, 0.044, 0.045 and 0.036 for lines B1, B2 and W1, respectively. With BLUP, 35 selected regions (empirical P < 0.05) were identified across the three lines. With GS, 70 selected regions were identified. Empirical thresholds for local allele frequency changes were determined from gene dropping, and differed considerably between GS (0.167-0.198) and BLUP (0.105-0.126). Between lines, the genomic regions with large changes in allele frequencies showed limited overlap. Our results show that GS applies selection pressure much more locally than BLUP, resulting in larger allele frequency changes. With these results, novel insights into the nature of selection on quantitative traits have been gained and important questions regarding the long-term impact of GS are raised. The rapid changes to a part of the genetic architecture, while another part may not be selected, at least in the short term, require careful consideration, especially when selection occurs before phenotypes are observed.

Indirect genetic effects and inbreeding: consequences of BLUP selection for socially affected traits on rate of inbreeding.

BackgroundSocial interactions often occur among living organisms, including aquatic animals. There is empirical evidence showing that social interactions may genetically affect phenotypes of individuals and their group mates. In this context, the heritable effect of an individual on the phenotype of another individual is known as an Indirect Genetic Effect (IGE). Selection for socially affected traits may increase response to artificial selection, but also affect rate of inbreeding.MethodsA simulation study was conducted to examine the effect of Best Linear Unbiased Prediction (BLUP) selection for socially affected traits on the rate of inbreeding. A base scenario without IGE and three alternative scenarios with different magnitudes of IGE were simulated. In each generation, 25 sires and 50 dams were mated, producing eight progeny per dam. The population was selected for 20 generations using BLUP. Individuals were randomly assigned to groups of eight members in each generation, with two families per group, each contributing four individuals. “Heritabilities” (for both direct and indirect genetic effects) were equal to 0.1, 0.3 or 0.5, and direct–indirect genetic correlations were −0.8, −0.4, 0, 0.4, or 0.8. The rate of inbreeding was calculated from generation 10 to 20.ResultsFor the base scenario, the rates of inbreeding were 4.09, 2.80 and 1.95% for “heritabilities” of 0.1, 0.3 and 0.5, respectively. Overall, rates of inbreeding for the three scenarios with IGE ranged from 2.21 to 5.76% and were greater than for the base scenarios. The results show that social interaction within groups of two families increases the resemblance between estimated breeding values of relatives, which, in turn, increases the rate of inbreeding.ConclusionBLUP selection for socially affected traits increased the rate of inbreeding. To maintain inbreeding at an acceptable rate, a selection algorithm that restricts the increase in mean kinship, such as optimum contribution selection, is required.

Sustainable plant breeding

AbstractPlant breeders disrupt Hardy–Weinberg equilibrium through selection, non‐random mating, drift, migration and mutation. Sustainable plant breeding can be defined as productive and competitive breeding that is achieved without loss of genetic diversity in the elite breeding population during the professional career of the breeder. Breeding is often productive but not sustainable. From 1974 to 2000, the animal breeding programme Meatlinc in the United Kingdom had effective population size of 95, population inbreeding of 0.19% per year and generation interval of 2.15 years. Genetic progress in Meatlinc tripled in the 8 years following introduction of best linear unbiased prediction (BLUP) selection (based on the information from relatives) in 1992. Canola breeding in Australia from 1970 to 2000 had longer generation interval (6 years), smaller effective population size (&lt;11) and higher rates of inbreeding (&gt;0.7% per year). BLUP selection in canola was first reported in 2010. Neither programme replaced genetic diversity lost through selection and drift. Most breeding programmes violate conditions of the infinitesimal model, thereby reducing predictability of selection, but breeders can minimize these limitations to sustainable plant breeding.

Efficient selection against categorically scored hip dysplasia in dogs is possible using best linear unbiased prediction and optimum contribution selection: a simulation study

Breeding to reduce the prevalence of categorically scored hip dysplasia (HD), based on phenotypic assessment of radiographic hip status, has had limited success. The aim of this study was to evaluate two selection strategies for improved hip status: truncation selection based on phenotypic record versus best linear unbiased prediction (BLUP), using stochastic simulation and selection scenarios resembling those in real dog populations. In addition, optimum contribution selection (OCS) was evaluated. Two traits were considered: HD (as a categorical trait with five classes and a heritability of 0.45 on the liability scale) and a continuous trait (with a heritability of 0.25) intended to represent other characteristics in the breeding goal. A population structure mimicking that in real dog populations was modelled. The categorical nature of HD caused a considerably lower genetic gain compared to simulating HD as a continuous trait. Genetic gain was larger for BLUP selection than for phenotypic selection in all scenarios. However, BLUP selection resulted in higher rates of inbreeding. By applying OCS, the rate of inbreeding was lowered to about the same level as phenotypic selection but with increased genetic improvement. For efficient selection against HD, use of BLUP breeding values should be prioritized. In small populations, BLUP should be used together with OCS or similar strategy to maintain genetic variation.

A comparison of collinearity mitigation techniques used in predicting BLUP breeding values and genetic gains over generations

Collinearity potentially has a negative impact on the prediction of genetic gains in tree breeding programs. This study investigated the reliability and impact of best linear unbiased prediction (BLUP) using various collinearity mitigation techniques and of two computational numerical precisions on the genetic gains in breeding populations. Multiple-trait, multiple-trial BLUP selection scenarios were run on Eucalyptus grandis (F1 F2 and F3) and Pinus patula (F1 and F2) data, comparing predicted breeding values of parents (forward prediction) with those realised in progeny (backward prediction of parents). Numeric precision had an impact on intergenerational correlations of BLUPs of some scenarios, indicating that it may not always be optimal to use higher precision when there is collinearity in the data. The relative difference in genetic gains between techniques varied by up to 0.38 standard deviation units in the less-stable pine population. This highlights the potentially large impact that instability can have on the efficiency of a breeding programme. BLUP performed close to expected in the relatively stable (less collinear) population (eucalypt F.,), and performed poorly in the other two populations. In the unstable pine data, some of the techniques resulted in improved intergenerational correlations coming in line with expected performance. This study indicates that BLUP can perform as expected and also confirms the potential problem of instability and consequences thereof. BLUP users should examine the nature of the population of predicted values and should these be outside expectation, various mitigation techniques should be explored.

Best linear unbiased prediction and optimum allocation of test resources in maize breeding with doubled haploids

With best linear unbiased prediction (BLUP), information from genetically related candidates is combined to obtain more precise estimates of genotypic values of test candidates and thereby increase progress from selection. We developed and applied theory and Monte Carlo simulations implementing BLUP in 2 two-stage maize breeding schemes and various selection strategies. Our objectives were to (1) derive analytical solutions of the mixed model equations under two breeding schemes, (2) determine the optimum allocation of test resources with BLUP under different assumptions regarding the variance component ratios for grain yield in maize, (3) compare the progress from selection using BLUP and conventional phenotypic selection based on mean performance solely of the candidates, and (4) analyze the potential of BLUP for further improving the progress from selection. The breeding schemes involved selection for testcross performance either of DH lines at both stages (DHTC) or of S(1) families at the first stage and DH lines at the second stage (S(1)TC-DHTC). Our analytical solutions allowed much faster calculations of the optimum allocations and superseded matrix inversions to solve the mixed model equations. Compared to conventional phenotypic selection, the progress from selection was slightly higher with BLUP for both optimization criteria, namely the selection gain and the probability to select the best genotypes. The optimum allocation of test resources in S(1)TC-DHTC involved ≥ 10 test locations at both stages, a low number of crosses (≤ 6) each with 100-300 S(1) families at the first stage, and 500-1,000 DH lines at the second stage. In breeding scheme DHTC, the optimum number of test candidates at the first stage was 5-10 times larger, whereas the number of test locations at the first stage and the number of test candidates at the second stage were strongly reduced compared to S(1)TC-DHTC.

Inbreeding in genome‐wide selection

Traditional selection methods, such as sib and best linear unbiased prediction (BLUP) selection, which increased genetic gain by increasing accuracy of evaluation have also led to an increased rate of inbreeding per generation (DeltaFG). This is not necessarily the case with genome-wide selection, which also increases genetic gain by increasing accuracy. This paper explains why genome-wide selection reduces DeltaFG when compared with sib and BLUP selection. Genome-wide selection achieves high accuracies of estimated breeding values through better prediction of the Mendelian sampling term component of breeding values. This increases differentiation between sibs and reduces coselection of sibs and DeltaFG. The high accuracy of genome-wide selection is expected to reduce the between family variance and reweigh the emphasis of estimated breeding values of individuals towards the Mendelian sampling term. Moreover, estimation induced intraclass correlations of sibs are expected to be lower in genome-wide selection leading to a further decrease of coselection of sibs when compared with BLUP. Genome-wide prediction of breeding values, therefore, enables increased genetic gain while at the same time reducing DeltaFG when compared with sib and BLUP selection.

Comparison of genomic and traditional BLUP‐estimated breeding value accuracy and selection response under alternative trait and genomic parameters

Accuracy of prediction of estimated breeding values based on genome-wide markers (GEBV) and selection based on GEBV as compared with traditional Best Linear Unbiased Prediction (BLUP) was examined for a number of alternatives, including low heritability, number of generations of training, marker density, initial distributions, and effective population size (Ne). Results show that the more the generations of data in which both genotypes and phenotypes were collected, termed training generations (TG), the better the accuracy and persistency of accuracy based on GEBV. GEBV excelled for traits of low heritability regardless of initial equilibrium conditions, as opposed to traditional marker-assisted selection, which is not useful for traits of low heritability. Effective population size is critical for populations starting in Hardy-Weinberg equilibrium but not for populations started from mutation-drift equilibrium. In comparison with traditional BLUP, GEBV can exceed the accuracy of BLUP provided enough TG are included. Unfortunately selection rapidly reduces the accuracy of GEBV. In all cases examined, classic BLUP selection exceeds what was possible for GEBV selection. Even still, GEBV could have an advantage over traditional BLUP in cases such as sex-limited traits, traits that are expensive to measure, or can only be measured on relatives. A combined approach, utilizing a mixed model with a second random effect to account for quantitative trait loci in linkage equilibrium (the polygenic effect) was suggested as a way to capitalize on both methodologies.

Optimizing Selection on Multiple Identified Quantitative Trait Loci in Population with Overlapping Generations

A method was developed to model and optimize selection on multiple identified quantitative trait loci (QTLs) and polygenic estimated breeding value, in order to maximize a weighted sum of cumulative response to selection over multiple years in a population with overlapping generations. The model allows for a population with multiple sex-age classes, different number of age class between sires and dams, and varied genetic contribution of the age class. The optimization problem was formulated as a multiple-stage optimal control problem and solved by a forward and backward iteration loop. The practical utility of this method was illustrated in an example of pig breeding population with overlapping generations. The selection response of this method was compared with standard QTL selection and conventional best linear unbiased prediction (BLUP) selection. Simulation results show that optimal selection achieved greater selection response than either standard QTL or conventional BLUP selections. The influence of population structure on optimal selection was significant. Optimal QTL selection and standard QTL selection were more favorable in a population with overlapping generations than discrete generations, and obtained more benefits relative to conventional BLUP selection in a population with overlapping generations. Optimal QTL selection relative to conventional BLUP selection is also more favorable following increase of genetic contribution of two-year-old boars and sows in a population with overlapping generations.

Expected increases in genetic merit in the UK Aberdeen Angus beef cattle breed from applying optimised selection

Best Linear Unbiased Prediction (BLUP) estimates of breeding values (EBVs) for economically relevant traits have been used for selection decisions in the UK Aberdeen Angus (AA) population since the early nineteen nineties. Selection exclusively based on BLUP-EBVs is expected to give higher gains than less accurate selection but can also lead to increased rates of inbreeding (ΔF). Dynamic rules using BLUP-EBVs to maximise genetic merit while DF is constrained to a pre-defined level are currently available (e.g. Grundy et al 1998). They showed that the use of these rules gives higher gains than standard BLUP selection at the same level of ΔF. The objective of this study was to investigate the potential of these procedures for optimising selection decisions in the UK AA population.

Optimisation of selection decisions in the UK Meatlinc breed of sheep

Best Linear Unbiased Prediction (BLUP) estimates of breeding values (EBVs) have been routinely used for selection decisions in the UK Meatlinc (ML) population since the early nineteen nineties. This has enabled accurate selection and has allowed higher genetic gains for traits of economic relevance than in other terminal sheep breeds (MLC, 1999). However, concerns regarding increased rates of inbreeding (ΔF) by selecting exclusively on BLUP-EBVs have arisen in this small population. Dynamic rules to maximise genetic merit while ΔF is constrained to a pre-defined level using BLUP EBVs are currently available (e.g. Grundy et al 1998). They found higher gains than standard BLUP selection at the same ΔF by using these rules. The objective of this study was to investigate the potential of these procedures for optimising selection decisions under constrained inbreeding in the UK ML sheep population.

Use of marker haplotypes to refine covariances among relatives for breeding value estimation

Markers flanking DNA regions, where quantitative trait loci (QTL) have been previously spotted, can be used to trace the common inheritance of major genes for a better definition of covariances among animals. A practical approach to the use of marker data to refine the additive relationship matrix used in the traditional best linear unbiased prediction (BLUP) methodology is presented. The technique allows the number of the mixed model equations to be kept to an animal level, blending polygenic pedigree data with marker haplotype information. The advantage of this marker‐assisted selection (MAS) approach over BLUP selection has been assessed through a stochastic simulation. A finite locus model with 32 independent biallelic loci was generated with normally distributed allelic effects. The heritability of the trait, measured on both sexes and on females only, was set to 0.2 and 0.5. Five‐allelic markers 2, 10 and 20 cM apart, bracketed the QTL with the largest effect on the trait, accounting for 17% of the genetic variance. The bracketed QTL had two or eight alleles and its position was undefined within the bracket. Results show a moderate 2% advantage of MAS over BLUP in terms of higher genetic response when trait was recorded on both sexes and heritability was 0.2. The benefit is in the short term, but it lasts longer with polyallelic QTL. When the trait was recorded on females only, MAS produced only a small and insignificant genetic gain, but reduced the overall inbreeding in the population. MAS was also inefficient when heritability was 0.5.

Genetic gains and levels of relatedness from best linear unbiased prediction selection ofEucalyptus urophyllafor pulp production in southeastern China

Breeding values for diameter at breast height (DBH), tree height (HT), relative bark thickness (BKR), and pilodyn penetration (PP) in Eucalyptus urophylla St. Blake plantations were predicted with best linear unbiased prediction (BLUP) approach. These values along with their economic weights derived from a previous study were then used to estimate economic genetic gains for three breeding objectives (pulp, woodchips, and wood volume) in southeastern China. The results showed substantial gain can be expected from selecting top 5% trees, with a reduction of up to US$35 for producing a tonne of ovendry pulp. However, actual gains can be strongly influenced by how the breeding objectives have been defined and whether the key traits have been included in the selection criteria. This study also showed that problem in the increase of coancestry associated with selection on BLUP would not be serious, with average coancestry amongst the selected population was less than 1%. More importantly, an unrestricted multiple-trait BLUP selection did not result in the same increase in relatedness in the selected population than it does for the single trait situation.

The effect of using different culling regimens on genetic response with two-trait, two-stage selection in a nucleus broiler stock

Stochastic simulation was used to study the effect on genetic response and inbreeding of various two-stage two-trait culling strategies. Four different parameter sets were considered for the two traits, BW and egg number. Selection of replacement animals was based on animal model best linear unbiased prediction (BLUP) to obtain estimated breeding values (EBV) at the second stage. Culling at Stage 1 was based on either animal model BLUP or phenotypes, and information from culled animals was either available or not available for calculation of second stage EBV. Besides founder individuals, six discrete generations were considered. Culling based on BLUP of two traits at Stage 1 produced higher response than culling on phenotypic evaluations. It was found that culling based on phenotypic evaluation and not carrying information to the second stage reduce rates of response by 9 to 17% and produced inbreeding higher than or close to that of BLUP selection. This study clearly shows that a double penalty of less response and higher inbreeding is generally paid for not using all information. Optimum selection schemes will depend on relative costs and benefits of collecting and processing the extra information required for full BLUP selection schemes.

Systems of mating to reduce inbreeding in selected populations

AbstractStochastic simulation is used to compare different systems of mating to reduce rates of inbreeding in selection programmes with phenotypic or animal model best linear unbiased prediction (BLUP) evaluation. Compensatory mating (the mating between individuals from the largest selected families to individuals from the smallest) turns out to be proportionately about 0-30 more effective than minimum coancestry matings for situations with low rates of inbreeding, such as phenotypic selection or high population size, although the advantage is less apparent if common environmental effects are important. A modification of this system of mating is proposed which can be applied for overlapping generations, and this is shown to reduce rates of inbreeding proportionately by about 0-50 more than for discrete generations. Under high inbreeding, however, such as for BLUP selection and small population size, minimum coancestry matings, or even avoidance of sib matings are more effective. A procedure combining compensatory and minimum coancestry matings is also simulated and gives the largest reductions in the rate of inbreeding. The effects of these and other systems of mating on the rate of inbreeding are shown to occur through a reduction in the cumulative effect of selection and a deviation from Hardy-Weinberg proportions.

Computer simulation of marker-assisted selection utilizing linkage disequilibrium

The value of marker-assisted selection (MAS) using linkage disequilibrium between genetic markers and quantitative trait loci (QTL) was examined. To simulate the disequilibrium, four base populations were created, F2, F5, F10 and F20, by random mating from a cross between two inbred lines. Selections were on breeding values estimated from: (1) marker QTL (MQTL) associations (MAS); (2) conventional best linear unbiased prediction (BLUP) methods; and (3) a combination of 1 and 2 (COMB). Alternative cases were studied by varying the parameters (heritability, initial linkage disequilibrium, and distribution of QTL effects). A genome with 100 QTL and 100 markers randomly (but equally) spread over 20 chromosomes, each 100 centiMorgans (cM) in length, was generated. Linkage disequilibrium (over 30 replicates) of QTLs with their nearest marker averaged 0.153, 0.104, 0.068, and 0.047 for the four base populations, and fell to 0.035, 0.025, 0.021, and 0.018, respectively, after ten generations of MAS selection (heritability 0.25). The initial linkage disequilibrium had the greatest effect on the genetic gain by MAS with the responses for the base populations F2>F5>F10>F20. Genetic gains by conventional BLUP selection were usually greater than by MAS. However, MAS contributed to the combined selection (COMB) to give appreciably higher genetic responses. Hybridization of selected lines after several generations of selection contributed little to generating further linkage disequilibrium. Detection of markers closer to the QTL will increase the linkage disequilibrium available for selection. Eventually with very close linkage each QTL allele can be uniquely identified in selection, and selection will then be equivalent to selection on the QTLs themselves.