The importance of Genetic Relationships and Phenotypic Record on Genomic Accuracy of Simulated Imputation Data Via Animal Models in Presence of Genotype × Environment Interactions

The objective of this study was to investigate the role of genetic relationships between training and validation set with considering different ratio of phenotypic records of training set on accuracy of genomic prediction via animal models containing genotype × environment interactions in simulated imputation data. For this purpose, four different scenarios using 15k density containing different levels of linkage disequilibrium (LD= low and high) and quantitative trait loci (90 and 300) was simulated. After simulating the population, the markers randomly masked with 50 and 95 percentage of missing rate for each scenario; afterwards, hidden markers were imputed. In the following, using coding in R software, the simulation was done to create the genetic relationships between the traits in three environments with different heritability’s in original and imputed data. The accuracy of imputation ranged from 0.737 to 0.941. LD was main factor on accuracy of imputation. Accuracy of genomic prediction generally increased when LD and heritability was increased. When non-phenotyped animals for the second environment were estimated using information of their relatives in the third environment, accuracy of genomic predictions was generally high. The accuracy of genomic prediction was the highest, when 75% of the second environment animals had phenotypic records and evaluated using information of their relatives in the first and third environments. The results showed that genotype imputation of very low density to 15k panel is not always helpful for genotypes prediction. Generally, simultaneous usage of relatives information and increasing phenotypic records of validation set increase the accuracy of genomic predictions using animal models in presence of genotype × environment interactions.