Methodologies for estimation of genotype with environment interaction

Abstract

Several methods to estimate the genotype with environment interaction, with one trait recorded on each animal, are reviewed. Genetic variances and covariances estimated using bivariate REML methodology result in unbiased estimates of the genetic correlation, which is a measure of the genotype with environment interaction. When the genetic covariance is estimated by weighted crossproducts of univariate sire breeding value estimates, the genetic correlation estimate will be biased, unless appropriate account is taken of the progeny distribution between fixed effects. Developments in computer technology have allowed more complex models to be fitted to larger data sets than was previously possible. Derivative Free REML (DFREML) algorithms have enabled individual animal models with different fixed effects with several thousand records for each trait to be accommodated in multivariate analyses for the estimation of genetic variances and covariances. Examples of the estimation of genetic variances and covariance using REML and DFREML algorithms, in a genotype with environment framework, were presented, but in each case there was no evidence of a genotype with environment interaction.