Genomic prediction accuracy for five economically important traits in cashmere goats across single-nucleotide polymorphism densities and methods

The production and quality improvement of cashmere from Inner Mongolia's cashmere goats face numerous challenges, including high costs associated with measuring cashmere diameter and the low accuracy of cashmere yield measurements. Genetic improvement of the breed relies on traditional breeding methods, resulting in low selection accuracy and slow genetic progress. The high cost of typing using high-density single-nucleotide polymorphism (SNP) chip constrains the large-scale application of genomic selection techniques. This study employed seven randomly selected SNP densities-5, 10, 20, 30, 40, 50, and 60 K-and four genomic prediction methods-Genomic Best Linear Unbiased Prediction (GBLUP), Single-step Genomic Best Linear Unbiased Prediction (ssGBLUP), BayesA, and BayesB-to evaluate the prediction accuracy of five economically important traits in Inner Mongolia cashmere goats using five-fold cross-validation: cashmere diameter, cashmere yield, cashmere thickness, fleece length, and BW. The optimal "trait-SNP density-prediction method" combination was defined as the pairing that achieves the highest prediction accuracy for each trait, with prediction accuracy being the primary criterion for optimality. Research has found that the relationship between SNP density and prediction accuracy is not a simple linear one. For cashmere diameter and BW, the best prediction accuracy using the BayesB method was achieved at a SNP density of 60 K, with values of 0.7210 and 0.5302, respectively. The accuracy of predicting cashmere yield, fleece length, and cashmere thickness was highest when using the BayesA method at SNP densities of 50, 30, and 10 K, respectively, yielding accuracies of 0.4243, 0.4071, and 0.4661. For each of these traits, the accuracy of GBLUP and ssGBLUP remained relatively stable across increasing SNP densities; however, their predictive performance was significantly lower than that of the Bayesian method. Importantly, low-density chips combined with appropriate Bayesian methods can achieve prediction accuracy comparable to that of high-density chips, offering a cost-effective strategy for genomic selection in cashmere goat breeding programmes. Although Bayesian methods require longer computation time, this is manageable under current sample sizes and does not impede practical implementation.