Genomic Selection Between Unrealized Wheat ( <i> <scp>Triticum aestivum</scp> L </i> .) Crosses Taking Epistatic and Environmental Interactions Into Account

ABSTRACT Genomic selection (GS) is most often applied using additive GS models, without including gene by gene (epistasis) and gene by environment (G × E) interactions. This publication aims to validate GS for breeding of bread wheat ( Triticum aestivum L.), using both additive GS models and models with epistasis that include two‐gene interactions under years with varying weather conditions. Falling number (FN) and germination index (GI) were used as quality criteria traits in a collection of 296 genotypes from 11 field trials of bread wheat, all with FN data and 3 also with GI data. The trials were divided into three groups based on the average FN, resulting in a genomic best linear unbiased prediction model with heritability ranging from 0.33 to 0.53 between environments. Epistasis seemed significant in all three groups, but total heritability was lower in the model where epistasis was included, indicating that the data and the methods tested in this study were not sufficient for more complex models with epistatic effects.