A framework for simulating genotype by environment interaction using multiplicative models

Abstract This paper develops a general framework for simulating genotype by environment interaction (GEI) using multiplicative models. Many stages of plant breeding are complicated by GEI, from the selection of potential parents to the development of improved genotypes for release to growers. Despite its importance, however, current plant breeding simulations do not adequately capture the complexity of GEI because they either use unrealistic models to simulate it or they ignore it completely. The framework developed in this paper simulates the two main components of GEI, that is non-crossover and crossover interaction, using the class of multiplicative models. The framework is demonstrated using two working examples supported by R code. The first example embeds the framework into a linear mixed model to generate MET datasets with low, moderate or high GEI, which are then used to compare various statistical models widely used in plant breeding. The results show that the prediction accuracy of all models increases as the level of GEI decreases or the number of sampled environments increases. The second example integrates the framework into a breeding programme simulation to compare genomic and phenotypic selection strategies over time. The results show that genomic selection outperforms phenotypic selection by 1.4−1.8 times, depending on the level of GEI. These examples demonstrate how the new framework can be used to generate realistic MET datasets and model plant breeding programmes that better reflect the complexity of real-world settings, making it a valuable tool for breeders to optimize their breeding programmes. The framework also has broader applications beyond plant breeding, including animal, aquaculture and tree breeding as well as other analysis comparison settings.