Abstract

Understanding the genetic architecture of economically important traits is essential for the design and implementation of efficient breeding programs. Here, we analysed 3,653 Eastern oysters (Crassostrea virginica) from 62 crossing groups and 307 full-sib families, all genotyped with a 200 K SNP array, to (i) estimate heritability and genetic correlations for six growth-related traits, (ii) detect quantitative trait loci (QTL) and candidate genes, and (iii) compare the accuracy of pedigree- (PBLUP) versus genomic-based (GBLUP) predictions across marker densities, selection strategies (physical distance (PD) versus linkage disequilibrium (LD) and phenotype-based genotyping strategies. SNP-based Heritabilities ranged from 0.33 to 0.56, and genetic correlations among traits exceeded 0.61, indicating strong pleiotropy. GWAS revealed a polygenic architecture with a shared QTL on chromosome 2; the top SNP explained ? 1.1% of phenotypic variance. Except for the 0.5 K LD panel—where PBLUP was more accurate—and the 1 K LD panel—where both methods performed similarly—GBLUP outperformed PBLUP in all scenarios, with the greatest advantage observed for physically spaced (PD) panels. The highest accuracy (0.80) was achieved with the full 100 K SNP set. GBLUP models required panels containing ? 0.5 K PD (accuracy ? 0.55) or ? 2 K LD (accuracy ? 0.56) to significatively surpass PBLUP predictions. Training sets built from extreme phenotypes boosted accuracy in small to intermediate sample sizes (e.g., 0.74 at n = 1,500 with 5 K PD SNPs versus 0.53 under random sampling). These results provide novel insights into not only the genomic regions controlling growth in this species but also about the utility of PD-based SNP panels and balanced sampling designs for enhancing GS accuracy in C. virginica. © The Author(s) 2025.