Pre-imputation analysis in racing line of Quarter Horses genotyped with equine chips of 54k and 65k
Equine genotyping has been an important tool for studying the mechanisms underlying important traits. The first commercialized array was the Equine SNP50 BeadChip (Illumina, Inc., USA) containing 54,602 SNPs. In 2013 the Equine SNP70 Beadchip (Illumina, Inc., USA) was released with 65,157 markers. These two SNP chips share approximately 45k markers and about 20k and 8k are exclusively of the 54k and 65k, respectively. The two-way genotype imputation is a good strategy to optimize the use of genomic information. However, the lack of samples genotyped with both the two panels can bring serious difficulties to ensure the quality of genotype imputation. The objective of this study was to perform preliminary analysis of imputation in a real dataset of a racing line of Quarter Horses. The dataset containing 233 and 116 animals was genotyped with the 65k and 54k, respectively. The analyses were performed in order to compute the imputation efficiency using different ranges of MAF (< 0.01; from 0.01 to 0.05; from 0.05 to 0.1), and considering the population substructure by using the multidimensional scaling (MDS). The exclusive markers of each panel were excluded and the two sub-samples (genotyped with 54k e 65k) were analyzed together. In the first scenario, the animals genotyped with the 54k SNP chip had their genotypes with MAF < 0.1 'masked'. In the second scenario, the samples genotyped with the 54k SNP chip were used as reference to impute the samples genotyped with the 65k SNP chip. The MDS analysis was performed by classical (metric) MDS using the N x N matrix of genome-wide identical-by-state pairwise distances. The results showed high accuracy of imputation between different ranges of MAF. Besides that, the correlation between imputed and real genotypes ranged from 0.97 to 0.99 and the proportion of genotypes that were correctly imputed ranged from 95% to 99%. The MDS plot revealed absence of population substructure between samples genotyped with the two different SNP chips. These results are important indicators of the reliable efficiency of imputation of real data between individuals genotyped with panels of different densities when they share SNPs.