Exploring the areas of applicability of whole-genome prediction methods for Asian rice (Oryza sativa L.)

Abstract

Our simulation results clarify the areas of applicability of nine prediction methods and suggest the factors that affect their accuracy at predicting empirical traits. Whole-genome prediction is used to predict genetic value from genome-wide markers. The choice of method is important for successful prediction. We compared nine methods using empirical data for eight phenological and morphological traits of Asian rice cultivars (Oryza sativa L.) and data simulated from real marker genotype data. The methods were genomic BLUP (GBLUP), reproducing kernel Hilbert spaces regression (RKHS), Lasso, elastic net, random forest (RForest), Bayesian lasso (Blasso), extended Bayesian lasso (EBlasso), weighted Bayesian shrinkage regression (wBSR), and the average of all methods (Ave). The objectives were to evaluate the predictive ability of these methods in a cultivar population, to characterize them by exploring the area of applicability of each method using simulation, and to investigate the causes of their different accuracies for empirical traits. GBLUP was the most accurate for one trait, RKHS and Ave for two, and RForest for three traits. In the simulation, Blasso, EBlasso, and Ave showed stable performance across the simulated scenarios, whereas the other methods, except wBSR, had specific areas of applicability; wBSR performed poorly in most scenarios. For each method, the accuracy ranking for the empirical traits was largely consistent with that in one of the simulated scenarios, suggesting that the simulation conditions reflected the factors that affected the method accuracy for the empirical results. This study will be useful for genomic prediction not only in Asian rice, but also in populations from other crops with relatively small training sets and strong linkage disequilibrium structures.

Evaluating the accuracy of genomic prediction of growth and wood traits in twoEucalyptusspecies and their F1hybrids

“…However, for pulp yield, RKHS was instead the worst performing method, and it is possible that the definition of a kernel simply was not suitable for this particular trait [17]. Our results corroborate previous reports from both crops and animals [18,50,51], as well as forest trees. In loblolly pine, for example, the performance of rrBLUP and three Bayesian methods were only marginally different when compared across 17 traits with distinct heritabilities, with a small improvement seen for BayesA only for fusiform rust resistance where loci of relatively larger effect have been described [44].…”

An experimental validation of genomic selection in octoploid strawberry

“…28,51 In a simulation study in Asian rice, RKHS performed better than GBLUP, Bayesian methods and other non-parametric methods when traits had high heritability, presence of epistasis and were controlled by a large number of QTLs. 52 In a simulation study comparing PBLUP, Bayes B and GBLUP under different genetic models (that is, major QTL model, rare variant model and the infinitesimal model), Bayes B showed higher accuracy of breeding values for the QTL and rare variant models and similar accuracy for the infinitesimal model. 53 In a simulation study in barley, under a high-density marker scenario, two Bayes B models with different priors ( π ) had better prediction accuracies when the trait was controlled by less than 20 QTL, than RR-BLUP and a BLUP model based on the marker relationship matrix.…”

Next generation breeding

“…A number of statistical models have been proposed, mainly based on corrected linear regression, best linear unbiased prediction (BLUP) and Bayesian regression methods [81,82]. The prediction accuracy of the different methods is debated; while in some cases all the models gave similar accuracy in estimation [83], other studies evidenced how different population features (LD structure, presence of epistasis and relationship between the training and validation sets) as well as trait characteristics (genetic architecture, heritability) may influence the relative performance of the prediction methods [82,84].…”

“…Once the model providing the highest accuracy is identified, GS would allow selection of lines without utilization of phenotypic data through the model predicting the individual GEBVs [79,81,82,85].…”