The impact of genetic architecture on genome-wide evaluation methods

Abstract

The rapid increase in high-throughput single-nucleotide polymorphism data has led to a great interest in applying genome-wide evaluation methods to identify an individual's genetic merit. Genome-wide evaluation combines statistical methods with genomic data to predict genetic values for complex traits. Considerable uncertainty currently exists in determining which genome-wide evaluation method is the most appropriate. We hypothesize that genome-wide methods deal differently with the genetic architecture of quantitative traits and genomes. A genomic linear method (GBLUP), and a genomic nonlinear Bayesian variable selection method (BayesB) are compared using stochastic simulation across three effective population sizes and a wide range of numbers of quantitative trait loci (N(QTL)). GBLUP had a constant accuracy, for a given heritability and sample size, regardless of N(QTL). BayesB had a higher accuracy than GBLUP when N(QTL) was low, but this advantage diminished as N(QTL) increased and when N(QTL) became large, GBLUP slightly outperformed BayesB. In addition, deterministic equations are extended to predict the accuracy of both methods and to estimate the number of independent chromosome segments (M(e)) and N(QTL). The predictions of accuracy and estimates of M(e) and N(QTL) were generally in good agreement with results from simulated data. We conclude that the relative accuracy of GBLUP and BayesB for a given number of records and heritability are highly dependent on M(e,) which is a property of the target genome, as well as the architecture of the trait (N(QTL)).

Figure 1.—

Accuracy of GBLUP and BayesB (informed priors for π) in validation individuals for different numbers of QTL and heritabilities (h²) when the effective population size is 1000 and the number of individuals in the training set is 1000. SE < 0.018 in all scenarios.

Figure 2.—

Accuracy of BayesB in validation individuals with informed priors on π (BayesB Prior Informed) and a low prior of 57 QTL regardless of the actual number of QTL (BayesB Prior 57 QTL) when the effective population size is 1000, the number of individuals in the training set is 1000, and heritability is 0.3. Accuracy of GBLUP is included for reference. SE < 0.009 in all scenarios.

Figure 3.—

Predicted (solid bars) and simulated (shaded bars) accuracy of GBLUP and BayesB for (A) a heritability (h²) of 0.3 and varying effective population size (N_e) and number of individuals in the training set (N_P) and for (B) N_e = 1000 and varying h² and N_P. Different numbers of QTL expressed as proportions of M_e were considered for BayesB.

Figure 4.—

Actual number of QTL simulated and number of QTL predicted with Equation 4 using BayesB accuracy when the effective population size is 1000, the number of individuals in the training set is 2000, and the heritability is 0.3.