Construction and Evaluation of a High-Quality Reference Panel for Dairy Goats

High-quality reference panels are important resources for genotype imputation and genomic selection in dairy goats; however, dairy-goat reference panels remain limited in sample size, population representation, and standardized workflow evaluation. In this study, 1092 dairy-goat samples from multiple populations, and public resequencing datasets, were integrated to construct and evaluate a dairy-goat reference panel for low-coverage whole-genome sequencing (lcWGS) and SNP-array data. Phasing and imputation strategies were compared using Beagle 5.4, SHAPEIT5, GLIMPSE2, and a BaseVar + Beagle pipeline, and the effects of reference-panel diversity, panel size, sequencing depth, and genotyping platform were evaluated using concordance, imputation quality score (IQS), and squared dosage correlation (r²). Beagle 5.4 phasing combined with GLIMPSE2 imputation reduced computational time by approximately 40% while maintaining high imputation accuracy. The largest evaluated panel (n = 1000) achieved concordance = 0.98, IQS = 0.94, and r² = 0.91, while a practical population-size threshold of 600-800 individuals balanced accuracy gains and resource costs. Reference panels containing two to three genetically similar dairy-goat populations achieved stable performance, with concordance values above 0.90 and chromosome-level r² values above 0.91. Low-coverage sequencing at 0.5× or above effectively reduced the loss of accuracy for low-frequency variants, whereas imputed SNP-array data increased marker density from 18 to 180 to 1850-18,500 SNPs per 1 Mb window in representative regions, corresponding to an approximately 103-fold increase, and produced more concentrated GWAS signal peaks on Chr5, Chr8, Chr17, and Chr19. These findings establish a technical framework and reference resource for genotype imputation, association analysis, and genomic selection in dairy goats.