# Genotyping strategies for single-step genomic predictions in a simulated sheep population under different scenarios of pedigree error types

## TL;DR

This study explores how different genotyping strategies and pedigree errors affect the accuracy of genomic predictions in sheep, finding that random genotyping is more effective than selective methods.

## Contribution

The study introduces a simulation-based comparison of genotyping strategies and pedigree error impacts on genomic prediction accuracy in sheep.

## Key findings

- Random genotyping improved genomic prediction accuracy by up to 19% compared to selective strategies.
- Missing pedigree information had a greater negative impact on genomic predictions than misidentified sires.
- Prioritizing male genotyping up to 10% of the population improved prediction accuracy.

## Abstract

Genomic predictions provide more accurate estimated breeding values (EBV) in younger animals. However, sheep reference populations are still small and if the animals included in the reference populations are not chosen carefully, genomic predictions may be biased. In this context, we compared genotyping strategies varying in the proportion of animals genotyped (using a 50K SNP panel) and the extent of pedigree errors (misidentified sires or missing information) on accuracy, bias, and dispersion of genomically-enhanced EBV (GEBV). We simulated a composite sheep population mimicking the formation and flock structure of the Katahdin breed using the AlphaSimR package. Sixteen flocks with an effective population size of 103 were simulated for two traits with heritabilities of 0.35 and 0.10. Breeding values were predicted with Best Linear Unbiased Prediction (BLUP) and Single-step Genomic BLUP (ssGBLUP). Scenarios included combinations of 0%–100% males or females genotyped, 0%–20% pedigree errors, and three genotyping strategies (random, highest EBV, or highest phenotypic values). The final population (18,717 animals) was divided into training and validation sets for calculating validation statistics of GEBV. Genomic prediction accuracy significantly improved with random genotyping, outperforming phenotype and EBV-based strategies by up to 19%. Pedigree errors reduced GEBV accuracy while increasing bias and dispersion. Missing pedigree information impacted results more than misidentified sires. Increasing the proportion of animals genotyped improved GEBV prediction metrics, with random genotyping yielding higher accuracies, lower biases, and dispersion closer to 1 (desirable). Prioritizing the genotyping of males up to 10% of the population before incorporating females enhanced the accuracy of GEBV. Genomic information mitigated some pedigree error effects. However, selective genotyping increased GEBV bias and dispersion, and reduced prediction accuracy. Compared to random genotyping, selective genotyping captured less genomic diversity, limiting the effectiveness of the reference population. Similar conclusions were obtained for both trait heritability levels. These findings highlight the importance of genotyping strategies when implementing genomic selection in sheep and the usefulness of genomic information for minimizing the impact of pedigree errors.