First Genomic Prediction of Single-Step Models in Large Yellow Croaker

Genome selection is mainly used in disease-resistant traits of aquatic species; however, its implementation is hindered by a high cost of genotype and phenotype data collection. Single-step genomic best linear unbiased prediction (SSGBLUP) can integrate phenotypes, genetic markers, and pedigree records into simultaneous prediction without increasing genotyping costs. The objective of this study is to investigate the performance of SSGBLUP in large yellow croaker and to evaluate the effects of the number of phenotypic records and genotyping per family on the predictive ability of SSGBLUP. A large yellow croaker population consists of 6898 individuals from 14 families with survival time resistant against Cryptocaryon irritans (C. irritans), body weight (BW), and body length (BL) traits were collected, of which 669 individuals were genotyped. Results showed that the mean predictive ability of all traits in the individuals randomly sampling for SSGBLUP, GBLUP, and BLUP was 0.738, 0.738, and 0.736, respectively. Moreover, the predictive ability of SSGBLUP and BLUP models did not increase with the extra phenotypic records per family, in which the predictive ability of SSGBLUP and BLUP in survival time was 0.853 and 0.851 for only genotyped data (N = 0) used, and 0.852, 0.845 for all phenotypic records (N = 600) used, respectively. However, with the increase in the genotype number of training set, the prediction ability of SSGBLUP and GBLUP model was increased and the highest predictive ability was gained when the genotype number per family was 40 or 45. In addition, the prediction ability of SSGBLUP model was higher than that of GBLUP. Our study showed that the SSGBLUP model still has great potential and advantages in genomic breeding of large yellow croakers. It is recommended that each family provide 100 phenotypic individuals, of which 40 individuals with genotyping data for SSGBLUP model prediction and family resistance evaluation.