Fusarium head blight (FHB) disease resistance and winter hardiness are the two priority traits in the Canadian winter durum wheat breeding program. These traits, influenced significantly by polygenic factors and environmental conditions, are crucial given the ongoing climate change challenges in Western Canada, such as rising temperatures and escalating drought risks.

There is only a single winter durum cultivar (OAC Amber) registered in Eastern Canada in 2010, which remains uncommercialized. At present, no winter durum wheat candidates are available for testing in Western Canada. Recently, joint Genome-Wide Association Study-Genomic Selection (GWAS-GS) and haploblock based GS approach have shown great potential for deciphering the genetic basis of complex traits, facilitating more accurate prediction of breeding values and enhancing genetic gains for difficult-to-phenotype and complex traits, such as FHB resistance and winter hardiness.

In our study, we assembled a diverse panel of 292 winter durum accessions from Canada, Europe, and the USA, including in-house winter hexaploid wheat x durum cross derivatives. The panel was phenotyped for FHB resistance and winter hardiness at multi-location trials in Winnipeg, Carman, and Ottawa during the 2021-23 seasons, with ongoing trials in 2023-24. Genotyping was performed using Genotyping-by-Sequencing (GBS), and SNPs were called against the durum cv Svevo.v1 and IWGSC CS RefSeq v2.1 reference genomes using an in-house GBS pipeline to enhance the robustness of the downstream genetic analyses.

Genetic structure analyses (phylogenetic, principal component and ancestry analyses) clustered the accessions into six sub-populations. FarmCPU GWAS detected loci associated with winter hardiness on chromosome 2A, 4B, 5A, 5B and 7B and FHB resistance on chromosome 3B and 7B.

Further, we aim to employ joint GWAS-GS approach using several parametric models such as GBLUP, EGBLUP, RRBLUP and Bayesian models (BRR, BLASSO, EBLASSO, Bayes B, Bayes C), non-parametric models such as Random Forest and RKHS, and haploblock based genomic prediction. We will utilize GWAS-tagged markers derived from multiple GWAS models such as, MLM+K+Q, FarmCPU, MLMM and BLINK in the GS models as fixed effects.

Significant markers derived from each GWAS models will be consolidated into a streamlined, non-redundant marker set for further use in genomic prediction. The haplotype blocks will be constructed using different fine-tuned LD thresholds, number of adjacent markers and algorithms available in HaploView and HaploBlocker (R package) and will be utilized for haploblock based genome prediction using suite of GS models.

Our ongoing study focuses on optimizing GS for Canadian winter durum development program.