Sample size determination for training set optimization in genomic prediction.

Area of Science:

• Quantitative genetics
• Animal and plant breeding
• Statistical genomics

Background:

• Genomic prediction (GP) utilizes genomic estimated breeding values (GEBVs) for trait selection in breeding programs.
• Establishing an optimal training set size for GP models is critical but often unresolved.
• Current practices often overlook cost-effectiveness and resource constraints in training set determination.

Purpose of the Study:

• To develop a practical and cost-effective approach for determining the optimal training set size in genomic prediction studies.
• To provide a method for optimizing selective phenotyping strategies.
• To facilitate the efficient selection of genotypes with economical sample sizes.

Main Methods:

• Applied logistic growth curve analysis to model prediction accuracy in relation to training set size.
• Utilized three real genome datasets to validate the proposed approach.
• Developed an R function to enable widespread application of the sample size determination method.

Main Results:

• A method was established to identify a cost-effective optimal training set size for genomic prediction.
• The approach effectively balances prediction accuracy with the economic constraints of phenotyping.
• Demonstrated the utility of the method across diverse genomic datasets.

Conclusions:

• The developed approach provides a practical solution for optimizing training set sizes in genomic prediction.
• The accompanying R function simplifies the implementation for breeders seeking economical phenotyping strategies.
• This facilitates more efficient and cost-effective breeding programs through informed genotype selection.