Authors: Charlotte Brault 1, Emily Conley 1, Jason Fiedler 2, and James Anderson 1
1. Department of Agronomy and Plant Genetics, University of Minnesota, Saint Paul, MN, United States
2. Biosciences Research Laboratory, USDA-ARS Genotyping Laboratory, Fargo, ND, United States
Corresponding author: Charlotte Brault, cbrault@umn.edu
Presenting Author: Charlotte Brault
Poster Video:
Abstract
Variability in FHB susceptibility is influenced by genotype, environment, and genotype-environment interaction (GxE). This study examines GxE in a multi-environment trial using data from the Uniform Regional Scab Nursery (URSN), a collaborative effort initiated in 1995 to assess resistant genotypes from spring wheat breeding programs across the Northern U.S. Of the 780 genotypes tested, 222 were genotyped using a 3K array, with a focus on the visual scabby kernel (VSK) trait.
Traditionally, GxE has been studied as a reaction norm over an environment index. Here, the environment index was predicted from a linear combination of a set of environmental covariables (ECs) specific to each environment. Three prediction scenarios were compared, all aimed at predicting untested genotypes with varying degrees of relatedness between training and target environments.
GxE prediction was performed using joint-genomic regression analysis (JGRA) with two modeling approaches: (1) JGRA Reaction Norm (JGRA RN), which estimates genotype responses to the environment through intercept and slope regression against the environment index, and (2) JGRA Marker, where genomic selection (GS) was applied in each environment, and marker effects were regressed against the environment index. These methods were compared to a baseline genomic selection model without environmental covariates (GS), with predictive abilities calculated within and across environments.
Results showed that within-environment predictions were more accurate for JGRA Marker than for GS or JGRA RN, though the differences were small (mean predictive abilities of 0.509 for JGRA Marker and 0.487 for GS). Predictive ability decreased when the target environment was less related to the training environments. For across-environment predictions, the advantage of the JGRA Marker was more pronounced, with a difference in predictive ability up to 0.203 compared to GS. Incorporating environmental covariables improved prediction accuracy and aided in selecting resistant genotypes across diverse environments.