Rohan L. Fernando

Deregressing estimated breeding values and weighting information for genomic regression analyses

BackgroundGenomic prediction of breeding values involves a so-called training analysis that predicts the influence of small genomic regions by regression of observed information on marker genotypes for a given population of individuals. Available observations may take the form of individual phenotypes, repeated observations, records on close family members such as progeny, estimated breeding values (EBV) or their deregressed counterparts from genetic evaluations. The literature indicates that researchers are inconsistent in their approach to using EBV or deregressed data, and as to using the appropriate methods for weighting some data sources to account for heterogeneous variance.MethodsA logical approach to using information for genomic prediction is introduced, which demonstrates the appropriate weights for analyzing observations with heterogeneous variance and explains the need for and the manner in which EBV should have parent average effects removed, be deregressed and weighted.ResultsAn appropriate deregression for genomic regression analyses is EBV/r2 where EBV excludes parent information and r2 is the reliability of that EBV. The appropriate weights for deregressed breeding values are neither the reliability nor the prediction error variance, two alternatives that have been used in published studies, but the ratio (1 - h2)/[(c + (1 - r2)/r2)h2] where c > 0 is the fraction of genetic variance not explained by markers.ConclusionsPhenotypic information on some individuals and deregressed data on others can be combined in genomic analyses using appropriate weighting.

Additive Genetic Variability and the Bayesian Alphabet

The use of all available molecular markers in statistical models for prediction of quantitative traits has led to what could be termed a genomic-assisted selection paradigm in animal and plant breeding. This article provides a critical review of some theoretical and statistical concepts in the context of genomic-assisted genetic evaluation of animals and crops. First, relationships between the (Bayesian) variance of marker effects in some regression models and additive genetic variance are examined under standard assumptions. Second, the connection between marker genotypes and resemblance between relatives is explored, and linkages between a marker-based model and the infinitesimal model are reviewed. Third, issues associated with the use of Bayesian models for marker-assisted selection, with a focus on the role of the priors, are examined from a theoretical angle. The sensitivity of a Bayesian specification that has been proposed (called “Bayes A”) with respect to priors is illustrated with a simulation. Methods that can solve potential shortcomings of some of these Bayesian regression procedures are discussed briefly.

Factors Affecting Accuracy From Genomic Selection in Populations Derived From Multiple Inbred Lines: A Barley Case Study

We compared the accuracies of four genomic-selection prediction methods as affected by marker density, level of linkage disequilibrium (LD), quantitative trait locus (QTL) number, sample size, and level of replication in populations generated from multiple inbred lines. Marker data on 42 two-row spring barley inbred lines were used to simulate high and low LD populations from multiple inbred line crosses: the first included many small full-sib families and the second was derived from five generations of random mating. True breeding values (TBV) were simulated on the basis of 20 or 80 additive QTL. Methods used to derive genomic estimated breeding values (GEBV) were random regression best linear unbiased prediction (RR–BLUP), Bayes-B, a Bayesian shrinkage regression method, and BLUP from a mixed model analysis using a relationship matrix calculated from marker data. Using the best methods, accuracies of GEBV were comparable to accuracies from phenotype for predicting TBV without requiring the time and expense of field evaluation. We identified a trade-off between a methods ability to capture marker-QTL LD vs. marker-based relatedness of individuals. The Bayesian shrinkage regression method primarily captured LD, the BLUP methods captured relationships, while Bayes-B captured both. Under most of the study scenarios, mixed-model analysis using a marker-derived relationship matrix (BLUP) was more accurate than methods that directly estimated marker effects, suggesting that relationship information was more valuable than LD information. When markers were in strong LD with large-effect QTL, or when predictions were made on individuals several generations removed from the training data set, however, the ranking of method performance was reversed and BLUP had the lowest accuracy.

Genomic Selection Using Low-Density Marker Panels

Genomic selection (GS) using high-density single-nucleotide polymorphisms (SNPs) is promising to improve response to selection in populations that are under artificial selection. High-density SNP genotyping of all selection candidates each generation, however, may not be cost effective. Smaller panels with SNPs that show strong associations with phenotype can be used, but this may require separate SNPs for each trait and each population. As an alternative, we propose to use a panel of evenly spaced low-density SNPs across the genome to estimate genome-assisted breeding values of selection candidates in pedigreed populations. The principle of this approach is to utilize cosegregation information from low-density SNPs to track effects of high-density SNP alleles within families. Simulations were used to analyze the loss of accuracy of estimated breeding values from using evenly spaced and selected SNP panels compared to using all high-density SNPs in a Bayesian analysis. Forward stepwise selection and a Bayesian approach were used to select SNPs. Loss of accuracy was nearly independent of the number of simulated quantitative trait loci (QTL) with evenly spaced SNPs, but increased with number of QTL for the selected SNP panels. Loss of accuracy with evenly spaced SNPs increased steadily over generations but was constant when the smaller number individuals that are selected for breeding each generation were also genotyped using the high-density SNP panel. With equal numbers of low-density SNPs, panels with SNPs selected on the basis of the Bayesian approach had the smallest loss in accuracy for a single trait, but a panel with evenly spaced SNPs at 10 cM was only slightly worse, whereas a panel with SNPs selected by forward stepwise selection was inferior. Panels with evenly spaced SNPs can, however, be used across traits and populations and their performance is independent of the number of QTL affecting the trait and of the methods used to estimate effects in the training data and are, therefore, preferred for broad applications in pedigreed populations under artificial selection.

Accuracy of Genomic Selection Methods in a Standard Data Set of Loblolly Pine (Pinus taeda L.)

Genomic selection can increase genetic gain per generation through early selection. Genomic selection is expected to be particularly valuable for traits that are costly to phenotype and expressed late in the life cycle of long-lived species. Alternative approaches to genomic selection prediction models may perform differently for traits with distinct genetic properties. Here the performance of four different original methods of genomic selection that differ with respect to assumptions regarding distribution of marker effects, including (i) ridge regression–best linear unbiased prediction (RR–BLUP), (ii) Bayes A, (iii) Bayes Cπ, and (iv) Bayesian LASSO are presented. In addition, a modified RR–BLUP (RR–BLUP B) that utilizes a selected subset of markers was evaluated. The accuracy of these methods was compared across 17 traits with distinct heritabilities and genetic architectures, including growth, development, and disease-resistance properties, measured in a Pinus taeda (loblolly pine) training population of 951 individuals genotyped with 4853 SNPs. The predictive ability of the methods was evaluated using a 10-fold, cross-validation approach, and differed only marginally for most method/trait combinations. Interestingly, for fusiform rust disease-resistance traits, Bayes Cπ, Bayes A, and RR–BLUB B had higher predictive ability than RR–BLUP and Bayesian LASSO. Fusiform rust is controlled by few genes of large effect. A limitation of RR–BLUP is the assumption of equal contribution of all markers to the observed variation. However, RR-BLUP B performed equally well as the Bayesian approaches.The genotypic and phenotypic data used in this study are publically available for comparative analysis of genomic selection prediction models.

Genomic prediction of simulated multibreed and purebred performance using observed fifty thousand single nucleotide polymorphism genotypes.

Genomic prediction involves characterization of chromosome fragments in a training population to predict merit. Confidence in the predictions relies on their evaluation in a validation population. Many commercial animals are multibreed (MB) or crossbred, but seedstock populations tend to be purebred (PB). Training in MB allows selection of PB for crossbred performance. Training in PB to predict MB performance quantifies the potential for across-breed genomic prediction. Efficiency of genomic selection was evaluated for a trait with heritability 0.5 simulated using actual SNP genotypes. The PB population had 1,086 Angus animals, and the MB population had 924 individuals from 8 sire breeds. Phenotypic values were simulated for scenarios including 50, 100, 250, or 500 additive QTL randomly selected from 50K SNP panels. Panels containing various numbers of SNP, including or excluding the QTL, were used in the analysis. A Bayesian model averaging method was used to simultaneously estimate the effects of all markers on the panels in MB (or PB) training populations. Estimated effects were utilized to predict genomic merit of animals in PB (or MB) validation populations. Correlations between predicted and simulated genomic merit in the validation population was used to reflect predictive ability. Panels that included QTL were able to account for 50% or more of the within-breed genetic variance when the trait was influenced by 50 QTL. The predictive power eroded as the number of QTL increased. Panels that composed the QTL and no other markers were able to account for 50% or more genetic variance even with 500 QTL. Panels that included genomic markers as well as QTL had less predictive power as the number of markers on the panel was increased. Panels that excluded the QTL and relied on markers in linkage disequilibrium (LD) to predict QTL effects performed more poorly than marker panels with QTL. Real-life situations with 50K panels that excluded the QTL could account for no more than 20% genetic variation for 50 QTL and less than 10% for 500 QTL. The difference between panels that included and excluded QTL indicates that the predictive ability of existing panels is limited by their LD. Training in PB to predict MB tended to be more predictive than training in MB to predict PB due to greater average levels of LD in PB than in MB populations. Improved across breed prediction of genomic merit will require panels with more than 50,000 markers.

Prediction of complex human traits using the genomic best linear unbiased predictor.

Despite important advances from Genome Wide Association Studies (GWAS), for most complex human traits and diseases, a sizable proportion of genetic variance remains unexplained and prediction accuracy (PA) is usually low. Evidence suggests that PA can be improved using Whole-Genome Regression (WGR) models where phenotypes are regressed on hundreds of thousands of variants simultaneously. The Genomic Best Linear Unbiased Prediction (G-BLUP, a ridge-regression type method) is a commonly used WGR method and has shown good predictive performance when applied to plant and animal breeding populations. However, breeding and human populations differ greatly in a number of factors that can affect the predictive performance of G-BLUP. Using theory, simulations, and real data analysis, we study the performance of G-BLUP when applied to data from related and unrelated human subjects. Under perfect linkage disequilibrium (LD) between markers and QTL, the prediction R-squared (R2) of G-BLUP reaches trait-heritability, asymptotically. However, under imperfect LD between markers and QTL, prediction R2 based on G-BLUP has a much lower upper bound. We show that the minimum decrease in prediction accuracy caused by imperfect LD between markers and QTL is given by (1−b)2, where b is the regression of marker-derived genomic relationships on those realized at causal loci. For pairs of related individuals, due to within-family disequilibrium, the patterns of realized genomic similarity are similar across the genome; therefore b is close to one inducing small decrease in R2. However, with distantly related individuals b reaches very low values imposing a very low upper bound on prediction R2. Our simulations suggest that for the analysis of data from unrelated individuals, the asymptotic upper bound on R2 may be of the order of 20% of the trait heritability. We show how PA can be enhanced with use of variable selection or differential shrinkage of estimates of marker effects.

Genomic selection in admixed and crossbred populations

In livestock, genomic selection (GS) has primarily been investigated by simulation of purebred populations. Traits of interest are, however, often measured in crossbred or mixed populations with uncertain breed composition. If such data are used as the training data for GS without accounting for breed composition, estimates of marker effects may be biased due to population stratification and admixture. To investigate this, a genome of 100 cM was simulated with varying marker densities (5 to 40 segregating markers per cM). After 1,000 generations of random mating in a population of effective size 500, 4 lines with effective size 100 were isolated and mated for another 50 generations to create 4 pure breeds. These breeds were used to generate combined, F(1), F(2), 3- and 4-way crosses, and admixed training data sets of 1,000 individuals with phenotypes for an additive trait controlled by 100 segregating QTL and heritability of 0.30. The validation data set was a sample of 1,000 genotyped individuals from one pure breed. Method Bayes-B was used to simultaneously estimate the effects of all markers for breeding value estimation. With 5 (40) markers per cM, the correlation of true with estimated breeding value of selection candidates (accuracy) was greatest, 0.79 (0.85), when data from the same pure breed were used for training. When the training data set consisted of crossbreds, the accuracy ranged from 0.66 (0.79) to 0.74 (0.83) for the 2 marker densities, respectively. The admixed training data set resulted in nearly the same accuracies as when training was in the breed to which selection candidates belonged. However, accuracy was greatly reduced when genes from the target pure breed were not included in the admixed or crossbred population. This implies that, with high-density markers, admixed and crossbred populations can be used to develop GS prediction equations for all pure breeds that contributed to the population, without a substantial loss of accuracy compared with training on purebred data, even if breed origin has not been explicitly taken into account. In addition, using GS based on high-density marker data, purebreds can be accurately selected for crossbred performance without the need for pedigree or breed information. Results also showed that haplotype segments with strong linkage disequilibrium are shorter in crossbred and admixed populations than in purebreds, providing opportunities for QTL fine mapping.

Controlling the proportion of false positives in multiple dependent tests.

Genome scan mapping experiments involve multiple tests of significance. Thus, controlling the error rate in such experiments is important. Simple extension of classical concepts results in attempts to control the genomewise error rate (GWER), i.e., the probability of even a single false positive among all tests. This results in very stringent comparisonwise error rates (CWER) and, consequently, low experimental power. We here present an approach based on controlling the proportion of false positives (PFP) among all positive test results. The CWER needed to attain a desired PFP level does not depend on the correlation among the tests or on the number of tests as in other approaches. To estimate the PFP it is necessary to estimate the proportion of true null hypotheses. Here we show how this can be estimated directly from experimental results. The PFP approach is similar to the false discovery rate (FDR) and positive false discovery rate (pFDR) approaches. For a fixed CWER, we have estimated PFP, FDR, pFDR, and GWER through simulation under a variety of models to illustrate practical and philosophical similarities and differences among the methods.

Genomic-BLUP Decoded: A Look into the Black Box of Genomic Prediction

Genomic best linear unbiased prediction (BLUP) is a statistical method that uses relationships between individuals calculated from single-nucleotide polymorphisms (SNPs) to capture relationships at quantitative trait loci (QTL). We show that genomic BLUP exploits not only linkage disequilibrium (LD) and additive-genetic relationships, but also cosegregation to capture relationships at QTL. Simulations were used to study the contributions of those types of information to accuracy of genomic estimated breeding values (GEBVs), their persistence over generations without retraining, and their effect on the correlation of GEBVs within families. We show that accuracy of GEBVs based on additive-genetic relationships can decline with increasing training data size and speculate that modeling polygenic effects via pedigree relationships jointly with genomic breeding values using Bayesian methods may prevent that decline. Cosegregation information from half sibs contributes little to accuracy of GEBVs in current dairy cattle breeding schemes but from full sibs it contributes considerably to accuracy within family in corn breeding. Cosegregation information also declines with increasing training data size, and its persistence over generations is lower than that of LD, suggesting the need to model LD and cosegregation explicitly. The correlation between GEBVs within families depends largely on additive-genetic relationship information, which is determined by the effective number of SNPs and training data size. As genomic BLUP cannot capture short-range LD information well, we recommend Bayesian methods with t-distributed priors.