Integrating GWAS and Random Forest to Develop a Robust 5-SNP Panel for Genetic Sex Identification in Pacific White Shrimp (Penaeus vannamei)

The Pacific white shrimp (Penaeus vannamei) exhibits pronounced sexual dimorphism in growth, making early and accurate sex identification crucial for targeted monitoring of sexual development and in-depth research into monosex aquaculture. However, morphological sex identification is unreliable at early developmental stages, and single-marker molecular assays are prone to diagnostic failures caused by genotyping errors or missing data. In this study, we developed a multi-SNP framework for sex prediction by integrating genome-wide association studies (GWAS) with Random Forest (RF) modeling. Analyzing 1244 individuals with a 55K SNP array, we identified 18 genome-wide significant SNPs and 10 candidate genes, including SRRM1 and SPSB3, that may be associated with sexual differentiation. Subsequent multi-marker comparative analyses demonstrated that although the top-performing single marker yielded high accuracy, simulated genotyping errors and missing data caused significant drops in accuracy and elevated no-call rates, whereas multi-SNP panels exhibited substantially improved robustness. A stable performance plateau was reached when expanding the panel to 5 SNPs. The model achieved 100% classification accuracy in the two mixed-sex independent populations, ZJ and HN, and correctly identified all females in the all-female MK cohort. Overall, this study provides a robust multi-SNP tool for sex identification in P. vannamei and highlights the practical utility of machine learning in aquaculture.