Impact of QTL properties on the accuracy of multi-breed genomic prediction

Abstract

Background: Although simulation studies show that combining multiple breeds in one reference population increases accuracy of genomic prediction, this is not always confirmed in empirical studies. This discrepancy might be due to the assumptions on quantitative trait loci (QTL) properties applied in simulation studies, including number of QTL, spectrum of QTL allele frequencies across breeds, and distribution of allele substitution effects. We investigated the effects of QTL properties and of including a random across- and within-breed animal effect in a genomic best linear unbiased prediction (GBLUP) model on accuracy of multi-breed genomic prediction using genotypes of Holstein-Friesian and Jersey cows.

Methods: Genotypes of three classes of variants obtained from whole-genome sequence data, with moderately low, very low or extremely low average minor allele frequencies (MAF), were imputed in 3000 Holstein-Friesian and 3000 Jersey cows that had real high-density genotypes. Phenotypes of traits controlled by QTL with different properties were simulated by sampling 100 or 1000 QTL from one class of variants and their allele substitution effects either randomly from a gamma distribution, or computed such that each QTL explained the same variance, i.e. rare alleles had a large effect. Genomic breeding values for 1000 selection candidates per breed were estimated using GBLUP modelsincluding a random across- and a within-breed animal effect.

Results: For all three classes of QTL allele frequency spectra, accuracies of genomic prediction were not affected by the addition of 2000 individuals of the other breed to a reference population of the same breed as the selection candidates. Accuracies of both single- and multi-breed genomic prediction decreased as MAF of QTL decreased, especially when rare alleles had a large effect. Accuracies of genomic prediction were similar for the models with and without a random within-breed animal effect, probably because of insufficient power to separate across- and within-breed animal effects.

Conclusions: Accuracy of both single- and multi-breed genomic prediction depends on the properties of the QTL that underlie the trait. As QTL MAF decreased, accuracy decreased, especially when rare alleles had a large effect. This demonstrates that QTL properties are key parameters that determine the accuracy of genomic prediction.

Figure 1

Accuracies of genomic prediction for traits that are controlled by QTL with different properties when 100 QTL underlie the trait. Average accuracies of genomic prediction (± standard errors) for Holstein-Friesian (HF, solid fill) and Jersey (J, diagonal fill) animals using a model that included a random across-breed animal effect and a within-breed animal effect, 606 384 SNPs, seven different reference populations and using simulated allele substitution effects (A) randomly sampled from a gamma distribution or (B) with each QTL explaining an equal proportion of the genetic variance, when 100 QTL underlying the trait were sampled from variants with on average a moderately low allele frequency (black), very low minor allele frequency (dark grey) or extremely low minor allele frequency (light grey).

Figure 2

Accuracies of genomic prediction for traits that are controlled by QTL with different properties when 1000 QTL underlie the trait. Average accuracies of genomic prediction (± standard errors) for Holstein-Friesian (HF, solid fill) and Jersey (J, diagonal fill) animals using a model that included a random across-breed animal effect and a within-breed animal effect, 606 384 SNPs, seven different reference populations and using simulated allele substitution effects (A) randomly sampled from a gamma distribution or (B) with each QTL explaining an equal proportion of the genetic variance, when 1000 QTL underlying the trait were sampled from variants with on average a moderately low allele frequency (black), very low minor allele frequency (dark grey) or extremely low minor allele frequency (light grey).

Figure 3

Accuracies of genomic prediction using different marker densities to calculate the genomic relationship matrix. Average accuracies of genomic prediction (± standard errors) for Holstein-Friesian (HF, solid fill) and Jersey (J, diagonal fill) animals using a model that included a random across-breed animal effect and a within-breed animal effect, seven different reference populations and using simulated allele substitution effects (A) randomly sampled from a gamma distribution or (B) with each QTL explaining an equal proportion of the genetic variance, when 100 QTL underlying the trait were sampled from variants with on average a moderately low minor allele frequency. The genomic relationship matrices were calculated using 606 384 SNPs (black), 60 000 SNPs (dark grey), 606 384 SNPs plus all sampled QTL (grey), or 60 000 SNPs plus all sampled QTL (light grey).

Figure 4

Log likelihood comparison of models with fixed or estimated random across-breed and within-breed animal effects. Twice the difference in log-likelihood for each of the 10 replicates and 5% significance threshold (black dotted line) using models with fixed variance components for the random across-breed animal effect and a within-breed animal effect compared to a model that estimated both variance components. The genomic relationship matrix was calculated based on 606 384 SNPs, the reference population consisted of 2000 Holstein Friesian and 2000 Jersey animals, allele substitution effects were sampled from a gamma distribution, when 1000 QTL underlying the trait were sampled from variants with on average a (A) moderately low allele frequency, (B) very low minor allele frequency or (C) extremely low minor allele frequency.

Figure 5

Accuracy of across- and multi-breed genomic prediction versus average minor allele frequency of QTL. The average accuracy of across- and multi-breed genomic prediction for (A) Holstein-Friesian and (B) Jersey selection candidates versus the average minor allele frequency of the 100 simulated QTL. Black points represent the scenarios with allele substitution effects randomly sampled from a gamma distribution and grey points represent the scenario with each QTL explaining an equal proportion of the genetic variance. The circles represent the accuracy for the multi-breed reference population with 2000 Holstein-Friesian and 2000 Jersey animals, the triangles represent the accuracy of across-breed genomic prediction with a reference population of 2000 animals from the other breed.