Comparison of imputation methods in Czech Holstein population

KRANJČEVIČOVÁ, Anita, KAŠNÁ, Eva, VOSTRÝ, Luboš a BRZÁKOVÁ, Michaela., 2020 Comparison of imputation methods in Czech Holstein population. In Book of Abstracts of the 71st Annual Meeting of the European Federation of Animal Science. Wageningen: Wageningen Academic, s. 380. ISSN
Internal link20178.pdf

The importance of genomic data in the breeding of dairy cattle is growing. Genomic information is used not only for genomic selection but also for revealing the genetic architecture. Data from DNA genotyping chips of various densities are available. Genotype imputation is the process of using reference population genotyped at a higher density to predict genotypes in population genotyped at a lower density. Our aim was to compare three different methods of genetic imputation with FImpute software and to choose the best one for Holstein population in the Czech Republic. We compared (1) imputation based on pedigree data, (2) population imputation and (3) combination of both approaches. A simulation study was performed on 3,994 animals genotyped on Illumina 50kBeadChip v.2, of which 994 animals were artificially masked on Illumina LD Beadchip v.2. Masked animals were imputed by all three methods. The success rate was based on two parameters: (1) imputation accuracy and (2) the percentage of correctly imputed SNPs. Since the animals to be masked were randomly selected, the whole calculation process was repeated 100 times, and the resulting parameters were averaged. We achieved the highest average accuracy (0.831±0.0004) and the highest percentage of correctly imputed SNPs (0.963±0.0008) with population imputation. For the combined method of imputation, the average accuracy was 0.827±0.0004 and the percentage of correctly imputed SNPs 0.957±0.0009. Since the massive genotyping of animals in the Czech Republic began a few years ago, our database does not contain enough genotyped ancestors. For this reason, pedigree imputation achieved low average accuracy of 0.132±0.0039 and the percentage of correctly imputed SNPs was 0.274± 0.0109. The results showed that pedigree imputation is currently not suitable for our population, and we need to focus on the population approach.