Improvement of the accuracy of breeding value prediction for egg production traits in Muscovy duck using low-coverage whole-genome sequence data.

Low-coverage whole genome sequencing (lcWGS) is an effective low-cost genotyping technology when combined with genotype imputation approaches. It facilitates cost-effective genomic selection (GS) programs in agricultural animal populations. GS based on lcWGS data has been successfully applied to livestock such as pigs and donkeys. However, its effectiveness in poultry is poorly reported. Furthermore, due to the high linkage disequilibrium (LD) between markers and the high marker density in lcWGS data, it is necessary to explore how to effectively utilize lcWGS data for genomic prediction. Phenotypic data for egg production traits were collected from a population of 1491 Muscovy ducks, with 975 of them sequenced using low-coverage whole genomic sequencing at an average depth of ∼0.84x. In the prediction, we compared the pedigree-based best linear unbiased prediction (PBLUP) method, the genomic best linear unbiased prediction (GBLUP) method utilizing SNP marker data, and the single-step genomic best linear unbiased prediction (SSGBLUP) method, which integrates both pedigree and SNP marker information. Among the SNP-based approaches, we further extended our analysis by applying LD-based weighting of SNPs and employing a Gaussian kernel model to capture epistatic genetic effects. The result showed that the estimated heritability of egg production traits in Muscovy duck ranged from 0.071 to 0.573. Compared to the PBLUP, integrating lcWGS data and pedigree data through a single-step genetic evaluation improved the accuracy of genomic prediction for all traits in this study, with accuracy improvement ranging from 12.3 % to 43.9 % in random cross-validation. Additionally, compared to the GBLUP, the extended method of GBLUP that controls for LD heterogeneity and accounts for epistatic effects using lcWGS data showed a superior prediction performance, with accuracy improvement ranging from 0.6 %∼75.1 % in the optimal scenario. This study demonstrates that utilization of lcWGS data is a promising approach for genomic prediction of egg production traits in Muscovy duck. Our findings provide valuable strategies for optimizing genomic prediction methods using lcWGS data.