Comparative analysis of genomic prediction models based on body weight trait in large yellow croaker (Larimichthys crocea)

In recent years, genomic selection (GS) has been increasingly utilized in the genetic improvement of aquaculture species, exhibiting superior prediction accuracy relative to conventional breeding methods based on pedigree information. Nevertheless, there is a paucity of studies comparing the prediction accuracy of GS models specifically for aquatic species. This study aims to evaluate the feasibility of the genomic selection for the body weight trait in large yellow croaker and to analyze the prediction accuracy among various models, including GBLUP, Bayes, and machine learning. We sequenced 564 samples from reference populations using the NingXinIII liquid SNP array and subsequently evaluated the genomic heritability of body weight trait, as well as the prediction accuracy of models in large yellow croaker. The findings indicated that the estimated heritability values for the body weight trait in large yellow croaker ranged from 0.57 to 0.69. While machine learning is most suitable for binary trait, GBLUP demonstrated greater efficacy for continuous trait. A small number of SNPs derived from genome-wide association study (GWAS) exhibited higher predictive abilities compared to the whole-genome level in GS. These findings demonstrated that implementing GS to enhance body weight traits in breeding programs for large yellow croaker is feasible, with the potential to achieve high accuracy in genomic predictions.