THE EFFECTS OF NUMBER OF REFERENCE INDIVIDUALS ON THE ACCURACY OF IMPUTATION FROM LOW AND MEDIUM DENSITIES TO HIGH DENSITY

摘要: Imputation is a common approach to infer the missing markers for individuals with low marker density (ie target population) from a reference population genotyped with higher-density Single-Nucleotide Polymorphism (SNP) panels. Several factors affect the imputation accuracy of untyped, including the number of reference individuals, marker density and population structure. This paper investigates the effects of these factors on the accuracy of imputation by using individuals of a single cattle breed or multiple cattle breeds in the reference population with 600k marker density, as well as assuming the target population was genotyped with low (15k) or medium (30k) marker density. To achieve a within breed imputation accuracy of> 90%, we required at least 500 individuals in the reference population when the target population was genotyped with 15k SNP panel. Whereas, if the reference population consisted of a mixture of purebred and multi-breed individuals, the SNP density must be at least 30k in the target population, and there must be more than 900 individuals in the reference population to achieve a similar level of accuracy.