Advancing selective breeding in leopard coral grouper (P. leopardus) through development of a high-throughput image-based growth trait

Utilizing image-based computer vision techniques, many high-throughput phenotyping methods have been employed to capture intricate growth trait characteristics, offering reliable estimates of phenotypic traits crucial for breeding programs. In this study, we explored the application of partial differential equation (PDE)-based level set approaches to introduce image-based body area percentage (IBAP) as a novel growth trait in Plectropomus leopardus, as a substitution for the traditional growth trait body weight (BW). Assessing the genetic parameters essential for robust growth trait improvement in P. leopardus breeding programs, we estimated SNP-based heritability for IBAP and BW using a comprehensive set of SNPs (673,039 SNPs with MAF >2%). Results revealed heritability estimates of 0.515 (S.E. 0.06) for IBAP and 0.542 (S.E. 0.06) for BW. Moreover, strong phenotypic and genetic correlations of 0.812 (S.E. 0.001) and 0.903 (S.E 0.021) between IBAP and BW, respectively, underscored the potential for IBAP as a surrogate trait of BW for the genetic improvement of P. leopardus. We established a linear regression model of IBAP and BW (y ​= ​−730 ​+ ​1700×, (R2 ​= ​0.71)), after rigorous assessments of linearity, normality, and homoscedasticity, to confirm model fit. Evaluation of breeding value prediction accuracies using two linear models (rr-GBLUP and Bayes B) and a non-linear (RKHS) model demonstrated the superior performance of RKHS across IBAP and BW. Exploring the impact of varied marker densities for SNP selection on genomic prediction accuracy for IBAP and BW demonstrated a threshold of 10,000 SNPs for maximal model accuracy. These findings provide essential reference information and methodological groundwork for leveraging image-based traits in P. leopardus breeding endeavors, facilitating more efficient and precise genetic improvement programs.