Statistical considerations for genomic selection

Genomic selection is becoming increasingly important in animal and plant breeding, and is attracting greater attention for human disease risk prediction. This review covers the most commonly used statistical methods and some extensions of them, i.e., ridge regression and genomic best linear unbiased prediction, Bayesian alphabet, and least absolute shrinkage and selection operator. Then it discusses the measurement of the performance of genomic selection and factors affecting the prediction of performance. Among the measurements of prediction performance, the most important and commonly used measurement is prediction accuracy. In simulation studies where true breeding values are available, accuracy of genomic estimated breeding value can be calculated directly. In real or industrial data studies, either training-testing approach or k-fold cross-validation is commonly employed to validate methods. Factors influencing the accuracy of genomic selection include linkage disequilibrium between markers and quantitative trait loci, genetic architecture of the trait, and size and composition of the training population. Genomic selection has been implemented in the breeding programs of dairy cattle, beef cattle, pigs and poultry. Genomic selection in other species has also been intensively researched, and is likely to be implemented in the near future