Bias in genomic predictions by mating practices for linear type traits in a large-scale genomic evaluation

Abstract

2000 to 2014. Traits with more genetic progress tended to have more "inflated" genomic predictions (i.e., "inflation" means here that genomic predictions are larger in absolute values than expected, whereas "deflation" means smaller than expected). Heritability estimates for 14 out of 18 traits declined in the last 16 yr, and Δh 2 ranged from −0.09 to 0.04. Traits with a greater decline in heritability tended to have more deflated genomic predictions. Biases (inflation or deflation) in genomic predictions were not improved by using the latest genetic parameters, implying that bias in genomic predictions due to preselection was not substantial for a large-scale genomic evaluation. Moreover, the strong selection intensity was not fully responsible for bias in genomic predictions. The directional selection can decrease heritability; however, positive assortative mating, which was strongly associated with large genetic gains, could minimize the decline in heritability for a trait under strong selection and could affect bias in genomic predictions.

Reducing computational cost of large-scale genomic evaluation by using indirect genomic prediction

“…The sum of the b 0 value and b 1 × IGP must be reported together to avoid bias with the interpretation of IGP. The correlations of MEAN, MAX, and b 0 with standardized genetic progress (ΔG) from Tsuruta et al (2021) were high (0.96 to 0.98) for all genomic data sets. To compare b 0 and ΔG on the same scale, MEAN, MAX, and b 0 were also standardized by dividing by each genetic standard deviation.…”

“…Therefore, it may be more practical to predict their genomic performance separately and indirectly, rather than to include them in ssGBLUP evaluations to obtain GEBV. About 90% of the genotyped animals included in the genomic evaluation for type traits in US Holsteins are young females (heifers), which may have neither phenotypes for type traits nor progeny in the future, although they may have other phenotypes (e.g., production traits; Tsuruta et al, 2021). Including all of these animals in the main routine evaluation may not be reasonable because of the computing cost.…”

Accounting for Population Structure and Phenotypes From Relatives in Association Mapping for Farm Animals: A Simulation Study

“…This algorithm enables the computation of genomic predictions for millions of genotyped individuals with much less memory usage and computing time. Indeed, a successful computation of genomic predictions for 13.5 million animals in the pedigree, of which 2.3 million were genotyped, using the BLUPF90 software suite has recently been shown to be feasible ( Tsuruta et al, 2021 ). Although the computation of genomic predictions (GEBV), SNP effects, and variance explained by SNPs can be done efficiently with APY in ssGBLUP, the same does not apply to the computation of p -values in ssGWAS.…”

Single-Step Genomic Evaluations from Theory to Practice: Using SNP Chips and Sequence Data in BLUPF90

“…Computing time for constructing A −1 22 for 570,000 genotyped animals extracted from a population of 10 M animals was around 11 min [88]. Single-step GBLUP with G −1 APY and efficient A −1 22 was successfully applied to over 10.9 M cows with milking records, 13.5M animals in the pedigree, and about 2.3 M genotyped Holsteins [89]; using 15,000 core animals, the complete evaluation for a model with 18 type traits took four and a half days to converge. Within this time stamp, the construction of G −1 APY and A −1 22 took one day.…”