How Population Structure Impacts Genomic Selection Accuracy in Cross-Validation: Implications for Practical Breeding.

Area of Science:

• Plant breeding
• Quantitative genetics
• Genomic selection

Background:

• Genomic selection is widely used in crop improvement, with prediction accuracy commonly reported via cross-validation.
• Population and family structure can inflate genomic prediction accuracies from random cross-validation, impacting breeding program success.
• Understanding these structural effects is vital for effective genomic selection implementation.

Purpose of the Study:

• To clarify the impact of population and family structure on genomic prediction accuracy for breeders and scientists.
• To compare genomic prediction accuracies from different cross-validation approaches and within-family prediction scenarios.
• To illustrate the distinct measures of prediction accuracy captured by among-family and within-family predictions.

Main Methods:

• Compared genomic prediction accuracies using random cross-validation and within-family prediction.
• Utilized a structured population of 940 *Brassica napus* hybrids from 46 testcross families and two subpopulations.
• Analyzed three distinct prediction scenarios to evaluate prediction accuracy measures.

Main Results:

• Genomic prediction accuracies from among-family predictions (in random cross-validation) and within-family predictions capture different aspects of accuracy.
• Among-family prediction accuracy reflects both the parent average component and the Mendelian sampling term.
• Within-family prediction accuracy solely assesses the predictability of the Mendelian sampling term.

Conclusions:

• Different genomic prediction accuracy measures provide distinct insights into prediction performance.
• A critical evaluation of prediction accuracy values in genomic selection studies is essential.
• This study promotes careful analysis of genomic prediction accuracy metrics in *Brassica napus* and other crops.