Imputation is a process of predicting unobserved data such as genotypes
. Imputation of high-density genotypes from low-density SNP panels and predicting missing genotypes is an essential step for the IGS Multi-breed Genetic Evaluation. All genotyped animals should have the same number of genotype markers without any missing data before they can enter into the genetic evaluation. The current IGS imputation pipeline is based on within-breed-association datasets. There are some animals in the IGS database which may have a parent from another breed. For example, a calf in the Simmental database may be sired by a Red Angus bull in the Red Angus database. That means the genotype information of his sire will not be used for imputation of the calf's genotypes in the current IGS single-breed imputation pipeline. This may impact the accuracy
of imputation and could have consequences on the quality control (QC) metrics used in the current IGS pipeline, leading to sire-offspring mismatches or the discarding of genotypes. The IGS genetic evaluation team recently developed a new multi-breed imputation, which also implements a new parentage tool for parentage QC as well as the new version of the software for imputation. The new multi-breed imputation pipeline is more accurate (98.2% vs 96.9% imputation accuracy for 12,676 tested animals) and results in fewer genotype QC failures (17% less discarded genotypes). There is more alignment between the parentage calls using the new approach and the stand-alone parentage test. The new IGS multi-breed imputation pipeline has little impact on GE-EPDs
. The correlations between GE-EPDs from beta test and actual production runs were high (0.98-0.99) for most traits in most breeds.