Genomic selection using information on multiple phenotypic traits and multiple growing environments

Plant breeders identify superior genotypes by collecting phenotypic data on multiple traits across field trials in multiple environments. The factor analytic linear mixed model is an effective method for analysing multi-environment field trial data, but has yet to be extended to genomic selection (GS) on multiple traits and multiple environments. The advantage of including both sources of information is that breeders can utilise genotype by environment by trait interaction to obtain more accurate predictions of the genetic effects across correlated traits and environments.
The objective of this research was to develop a new GS approach that incorporates multiple traits and multiple environments within a partially separable factor analytic framework. The so-called separable factor analytic (SFA) model is based on a three-way separable structure with a factor analytic model for traits, a factor analytic model for environments and a genomic relationship matrix for genotypes. This structure was then modified to enable a different genotype by environment interaction (GEI) pattern for each trait, and a different genotype by trait interaction (GTI) pattern for each environment. In this manner, the SFA model provides a natural framework for applying plant breeding selection tools to obtain measures of overall performance and stability for each trait.
The SFA model is demonstrated using a late-stage rice breeding dataset comprising ~266 genotypes and 12 environments in the south-eastern rice growing region of Australia, with phenotypic data collected on three key traits; grain yield, days to flowering and plant height.
The results show that the GEI pattern for grain yield is remarkably different from the other two traits, and that the GTI pattern differs across environments. Lastly, we demonstrate how this information can be efficiently utilised within a selection index to select for overall performance and stability.
This work represents an important continuation in the advancement of plant breeding analyses, particularly with the advent of high throughput phenotypic data involving a very large number of traits and environments.