Updating the reference population to achieve constant genomic prediction reliability across generations

Abstract

The reliability of genomic breeding values (DGV) decays over generations. To keep the DGV reliability at a constant level, the reference population (RP) has to be continuously updated with animals from new generations. Updating RP may be challenging due to economic reasons, especially for novel traits involving expensive phenotyping. Therefore, the goal of this study was to investigate a minimal RP update size to keep the reliability at a constant level across generations. We used a simulated dataset resembling a dairy cattle population. The trait of interest was not included itself in the selection index, but it was affected by selection pressure by being correlated with an index trait that represented the overall breeding goal. The heritability of the index trait was assumed to be 0.25 and for the novel trait the heritability equalled 0.2. The genetic correlation between the two traits was 0.25. The initial RP ( n = 2000) was composed of cows only with a single observation per animal. Reliability of DGV using the initial RP was computed by evaluating contemporary animals. Thereafter, the RP was used to evaluate animals which were one generation younger from the reference individuals. The drop in the reliability when evaluating younger animals was then assessed and the RP was updated to re-gain the initial reliability. The update animals were contemporaries of evaluated animals (EVA). The RP was updated in batches of 100 animals/update. First, the animals most closely related to the EVA were chosen to update RP. The results showed that, approximately, 600 animals were needed every generation to maintain the DGV reliability at a constant level across generations. The sum of squared relationships between RP and EVA and the sum of off-diagonal coefficients of the inverse of the genomic relationship matrix for RP, separately explained 31% and 34%, respectively, of the variation in the reliability across generations. Combined, these parameters explained 53% of the variation in the reliability across generations. Thus, for an optimal RP update an algorithm considering both relationships between reference and evaluated animals, as well as relationships among reference animals, is required.

Impact of Mislabeling on Genomic Selection in Cassava Breeding

“…Ly et al (2013) used real phenotypic and marker genotype data of cassava to suggest that training populations with close relatives to selection candidates attained higher prediction accuracy. Other simulation studies (Jannink, 2010; Iwata et al, 2011; Yabe et al, 2014; Pszczola and Calus, 2016) also suggested updating the prediction model during selection cycles, with the idea of using individuals close to the selection candidates as the training population. In animal breeding studies (Calus, 2010; Habier et al, 2010; Wolc et al, 2011), the decay of the prediction accuracy across generations was especially strong when the prediction methods depended on kinship information (or realized kinship calculated from use of markers).…”

Review: Opportunities and challenges for small populations of dairy cattle in the era of genomics

“…Since the reliability of genomic prediction decreases as the genetic distance between reference population and selection candidates increases (Habier et al, 2010), reference populations need to be continuously updated to avoid a decline in relationships between the reference and the predicted populations and thus a decrease in accuracy of genomic prediction (Pryce et al, 2012a). Pszczola et al (2014) calculated a loss of accuracy per generation of 7% in a simulated reference population of 2000 cows in which no animals were added over the years. If the goal is to maintain the original level of accuracy, it is necessary to include new animals, whereby the selection of new individuals to enter the reference population is apparently more important than their actual quantity (Pszczola et al, 2014).…”

“…Pszczola et al (2014) calculated a loss of accuracy per generation of 7% in a simulated reference population of 2000 cows in which no animals were added over the years. If the goal is to maintain the original level of accuracy, it is necessary to include new animals, whereby the selection of new individuals to enter the reference population is apparently more important than their actual quantity (Pszczola et al, 2014). The opposite situation, removing data from former generations, is a practicable way to reduce computation requirements without loss in reliability.…”

Selection of trait-specific markers and multi-environment models improve genomic predictive ability in rice

“…In the present study, we showed that (i) reasonably high PA of genomic prediction can be their GEBV and to update the multi-environment prediction model for the next breeding cycle [67][68]. The process can be implemented in the framework of the pedigree-breeding of progeny of biparental crosses between members of the reference population of the breeding program that is used as the training population [36; 42].…”