Genomic selection has been utilized for the last decade to accelerate genetic improvement in various species. However, field data are insufficient to quantify the long-term effect of genomic selection for a variety of genetic architectures . Simulations based on either genomic selection or traditional genetic selection based on phenotypes and pedigree information could provide a better direction for more efficient improvement in farm animal populations. The objective of this study was to compare genetic progresses by genomic selection and traditional genetic selection for a quantitative trait over the generations, assuming different numbers of QTLs and animals.
Simulations for a population with an effective size of 4 0 included 21 overlapping generations with phenotypes for a trait with a heritability of 0.5 and an additive genetic variance of 50. G enotypes simulated for the last 10 generation comprised 30K SNPs across 30 autosomes including 1, 10, and 100 QTLs per chromosome for 220, 1100, 11,000, and 33,000 genotyped animals with phenotypes (Table 1) . The QTLs explained all the genetic variation for the trait ( i.e., no extra polygenic effects ). Selection was conducted by mating the top 10% males and 50% females for 20 generations. We estimated and compared genetic trends when selection was based exclusively on either a single-step genomic (ssG)BLUP, a traditional pedigree-based (A )BLUP, or a phenotypic BLUP without considering relationships among animals (IBLUP).
When the number of genotyped animals is smaller, the genetic gain by genomic selection (ssGBLUP ) wa s smaller than that by ABLUP and IBLUP (Table 1). With fewer QTLs , the genetic gain by genomic selection was even smaller than that by ABLUP and IBLUP. However, when the number of QTLs increased , the genetic gain became larger. When weighting SNPs by the percentage of genetic variance explained on the trait , the genetic gain was slightly bett er for a small genotyped population but not significant. t he percentage of additive genetic variance explained by top SNPs was lower (< 15%) for more polygenic traits, indicating that selection based on a few SNPs or QTLs chosen from genome-wide association would lead to less genetic gain. Maximizing the rate of genetic gain under genomic selection would require a large number of genotyped animals that explain the genomic variation significantly, especially for a trait with polygenic effects. The long-term selection study on genetic trend with field data should be conducted with caution.