Predictive ability of genomic selection models for breeding value estimation on growth traits of Pacific white shrimp Litopenaeus vannamei

• Published in: Chinese journal of oceanology and limnology Vol. 35; no. 5; pp. 1221 - 1229
• Main Author: 王全超 于洋 李富花 张晓军 相建海
• Format: Journal Article
• Language: English
• Published: Heidelberg Science Press 01.09.2017; Springer Nature B.V
• Subjects: Aquaculture and Fisheries; Bayesian analysis; Bias; Body length; Body weight; Breeding; breeding value; Coefficients; Crustaceans; DNA sequences; Earth and Environmental Science; Earth Sciences; Estimating techniques; Gene mapping; Gene polymorphism; Genetic improvement; Genomics; genotyping; growth traits; Litopenaeus vannamei; marker-assisted selection; Nucleotides; Oceanography; phenotype; Phenotypes; Polymorphism; Population studies; prediction; Predictions; Probability theory; Regression analysis; Regression coefficients; Reliability; Reliability analysis; Saline water; Shellfish; Single-nucleotide polymorphism; Training; 单核苷酸多态性; 南美白对虾; 基因组; 太平洋; 生长性状; 育种值估计; 选择模型; 预测能力; Pacific Ocean; genomic selection; model prediction; growth traits; penaeid shrimp
• ISSN: 0254-4059; 2096-5508; 1993-5005; 2523-3521
• DOI: 10.1007/s00343-017-6038-0

Genomic selection （GS） can be used to accelerate genetic improvement by shortening the selection interval. The successful application of GS depends largely on the accuracy of the prediction of genomic estimated breeding value （GEBV）. This study is a first attempt to understand the practicality of GS in Litopenaeus vannamei and aims to evaluate models for GS on growth traits. The performance of GS models in L. vannamei was evaluated in a population consisting of 205 individuals, which were genotyped for 6 359 single nucleotide polymorphism （SNP） markers by specific length amplified fragment sequencing （SLAF-seq） and phenotyped for body length and body weight. Three GS models （RR-BLUP, BayesA, and Bayesian LASSO） were used to obtain the GEBV, and their predictive ability was assessed by the reliability of the GEBV and the bias of the predicted phenotypes. The mean reliability of the GEBVs for body length and body weight predicted by the different models was 0.296 and 0.411, respectively. For each trait, the performances of the three models were very similar to each other with respect to predictability. The regression coefficients estimated by the three models were close to one, suggesting near to zero bias for the predictions. Therefore, when GS was applied in a L. vannamei population for the studied scenarios, all three models appeared practicable. Further analyses suggested that improved estimation of the genomic prediction could be realized by increasing the size of the training population as well as the density of SNPs.