Gene-interactions reduce genomic prediction accuracy across generations

Published on
January 30, 2020

The accuracy of genomic prediction partly depends on the genetic correlation between the reference population and the population of selection candidates. This correlation can be lower than 1 due to interactions between genes, such as dominance and epistasis. Wageningen researchers investigated the relationship between gene-interactions and the genetic correlation between populations. They published their results in the journal G3.

Genomic prediction and importance of genetic correlations between populations

Animal breeders aim to improve the average performance of their populations. They achieve this by selecting individuals with the highest breeding values as parents of the next generation. The breeding values can be estimated with a technique called genomic prediction. This technique makes use of a reference population that consists of animals that have both phenotypes and marker genotypes available. The data of the reference population is used to estimate breeding values of selection candidates that only have marker genotypes available. The accuracy of estimated breeding values is lower when the genetic correlation between the reference population and the population of selection candidates is lower than 1.

Publication: Gene-interactions and genetic correlations

Interactions among genes, such as dominance and epistasis, can cause the genetic correlation to be lower than 1, thereby reducing the accuracy of the genomic prediction. Researchers from Wageningen University & Research investigated how gene-interactions affect the genetic correlation between populations. They simulated two diverging populations with different types of gene-interaction. Their results showed that the genetic correlation decreased when populations had diverged for a longer time, and when interaction effects were larger. With dominance, genetic correlations did not drop below 80%, whereas with epistasis, values dropped to as low as 45%. These insights may contribute to the understanding of genetic differences in trait expression between populations, and may help in explaining the inefficiency of genomic prediction across populations and generations.