Stripe rust (YR) is an economically important disease threatening wheat production globally. Resistance to YR is affected by environment and genotype by environment interaction (GEI) effects. We aimed to derive selective breeding strategies to achieve rapid improvement for YR resistance that is stable across environments. We investigated disease responses of YR in the Vavilov wheat diversity panel (n = 295) in 11 field experiments, including three in Australia and eight in Ethiopia during multiple years.

A Factor analytic (FA) model was used to fit to GEI of YR resistance across environments. Using the FA model, two breeding zones, Australia and Ethiopia, could be clearly defined based on the genetic correlations. Line-level selection was conducted with the overall performance (OP) and stability of YR resistance in environments, where OP was the genotype score for the first latent factor (F1) expressed in trait units across environments and stability was the root-mean-square deviation (RMSD) from the regression line associated with latent variable F1. To calculate GEBV for these two traits, 34,899 genome wide markers and a GBLUP model were used. These GEBV were compared with GEBV from a multi-trait GBLUP model, where each environment was fitted as a separate trait, and performance was the average GEBV across traits and stability was the standard deviation of GEBV across environments (traits).

Results showed that the GEBV from the FA approach and the multi-trait approach were in good agreement, having high correlations (Spearman r = 0.49 – 0.68). By adopting haplotype-based local GEBV, the haploblocks with the highest variances of haplotype effects were identified to be associated with OP and RMSD of YR resistance, respectively, in combined Australia and Ethiopia environments, and the haplotypes with favorable effects on both OP and RMSD were identified. Compared with truncation selection, longer-term and higher-ability genetic gain of both OP and RMSD could be achieved by optimal haplotype selection (OHS) to choose parental crosses based on a selection index (SI) of OP and RMSD.

Through simulation we also demonstrate that selections in Australia/ Ethiopia can improve OP and RMSD of YR resistance in the constant breeding zone (i.e. Australia/ Ethiopia), but not in changing breeding zones (i.e. Ethiopia/ Australia), although OP to some extend can be improved.

Parental selections with OHS and SI in combined environments were predicted to advance long-term genetic improvement for both average performance and stability of YR resistance across global breeding environments, including Australia and Ethiopia.