A comprehensive comparison of high-density SNP panels and an alternative ultra-high-density panel for genomic analyses in Nellore cattle

There is evidence that some genotyping platforms might not work very well for Zebu cattle when compared with Taurine breeds. In addition, the availability of panels with low to moderate number of overlapping markers is a limitation for combining datasets for genomic evaluations, especially when animals are genotyped using different SNP panels. In the present study, we compared the performance of medium- and high-density (HD) commercially available panels and investigated the feasibility of developing an ultra-HD panel (SP) containing markers from an Illumina (HD\_I) and an Affymetrix (HD\_A) panels. The SP panel contained 1 123 442 SNPs. After performing SNP pruning on the basis of linkage disequilibrium, HD\_A, HD\_I and SP contained 429 624, 365 225 and 658 770 markers distributed across the whole genome. The overall mean proportion of markers pruned out per chromosome for HD\_A, HD\_I and SP was 15.17%, 43.18%, 38.63% respectively. The HD\_I panel presented the highest mean number of runs-of-homozygosity segments per animal (45.48%, an increment of 5.11% compared with SP) and longer segments, on average (3057.95 kb per segment), than did both HD\_A and SP. HD\_I also showed the highest mean number of SNPs per run-of-homozygosity segment. Consequently, the majority of animals presented the highest genomic inbreeding levels when genotyped using HD\_I. The visual examination of marker distribution along the genome illustrated uncovered regions among the different panels. Haplotype-block comparison among panels and the average haplotype size constructed on the basis of HD\_A were smaller than those from HD\_I. The average number of SNPs per haplotype was different between HD\_A and HD\_I. Both HD\_A and HD\_I panels achieved high imputation accuracies when used as the lower-density panels for imputing to SP. However, imputation accuracy from HD\_A to SP was greater than was imputation from HD\_I to SP. Imputation from one HD panel to the other is also feasible. Low- and medium-density panels, composed of markers that are subsets of both HD\_A and HD\_I panels, should be developed to achieve better imputation accuracies to both HD levels. Therefore, the genomic analyses performed in the present study showed significant differences among the SNP panels used. (AU)