# Efficiency of Recurrent Genomic Selection in Panmictic Populations

## TL;DR

This study uses simulations to show how effective genomic selection is in breeding animals, especially when considering genetic interactions and population structure.

## Contribution

The study introduces a simulation-based evaluation of recurrent genomic selection in panmictic populations under different genetic models.

## Key findings

- Genomic selection efficacy is proportional to linkage disequilibrium and genotypic variance.
- Model updating increases genomic selection efficacy, while non-allelic gene interaction decreases it.
- A minimum of 10% genotyped and phenotyped individuals is sufficient to maintain reasonable efficacy.

## Abstract

Simulation-based studies offer knowledge that can only be indirectly provided by empirical investigations. The objective of this simulation-based study was to assess the efficiency of recurrent genomic selection in a panmictic population under additive–dominance and additive–dominance with epistasis models. Genomic selection efficacy is proportional to the population linkage disequilibrium and genotypic variance. Genomic selection efficacy can be increased by model updating. Genomic selection efficacy can be kept at a reasonable level, with a minimum of 10% of genotyped and phenotyped individuals. Non-allelic gene interaction decreases genomic selection efficacy. The inbreeding that occurs over generations of genomic selection should be minimized by avoiding sub-crosses. We demonstrated that genomic selection is an efficient method for animal breeding, especially for complex traits.

Simulation-based studies can support breeders’ decisions inexpensively, since there is no need to perform a new procedure. The objective was to assess the efficiency of recurrent genomic selection in panmictic population under additive–dominance and additive–dominance with epistasis models. We assumed two broiler chicken populations with contrasting linkage disequilibrium (LD) levels, 38,500 SNPs, and 1000 genes controlling feed conversion ratio. We applied recurrent genomic selection over seven cycles. The genomic selection efficacy, expressed as realized total genetic gain, was proportional to the LD level and genotypic variance. Genomic selection required model updating to achieve a higher efficacy. The training set size required by genomic selection can be as low as 10%/generation. Under this low-cost scenario, the genomic selection efficacy was slightly lower than the maximum efficacy. There is no difference between genetic evaluation methods regarding the decrease in the genotypic variance due to selection. In general, additive value prediction accuracies and realized genetic gains were highly correlated. The accumulated inbreeding level was not high due to avoidance of sib cross. The genomic inbreeding coefficient over generations was close to zero. Except for dominant epistasis, the efficacy of genomic selection was 4.1 to 46.2% lower than the efficacy under no epistasis.