| Abstract | Durum wheat (Triticum turgidum ssp. durum) suffers substantial yield losses from yellow rust (Puccinia striiformis) and leaf rust (Puccinia triticina). In this study, we employed genome‐wide association studies (GWAS) to identify loci associated with rust resistance and used genomic selection (GS) to evaluate the predictive accuracy of different statistical models and phenotyping metrics (AUDPC_GDD , Angle , GDD50 , and maxVar) in a Canadian durum wheat panel. The panel was evaluated in Mexico for yellow rust across three seasons near Toluca, and for leaf rust over two seasons at El Batán. Our GWAS identified 36 significant marker‐trait associations (MTAs), including known loci (Yr30, Yr57, Yr82, YrU1, Lr16, Lr17, Lr18 , and Lr65) and previously unreported regions. Yellow rust resistance was linked to loci on chromosomes 3A (602.7 Mbp) and 3B (243.4 Mbp), while leaf rust MTAs appeared on chromosomes 5A (552.8 Mbp) and 7A (570 Mbp). Candidate genes near novel MTAs encode defense‐related proteins such as serine/threonine kinases and NB‐ARC (nucleotide binding–Apaf‐1, R proteins, and CED‐4), F‐box, and RIN4 (RPM1‐interacting protein 4)‐domain proteins. Among four scoring metrics tested, AUDPC_GDD consistently outperformed others for yellow rust, whereas maxVar was most effective for leaf rust, reflecting differences in phenotypic distribution and trait variance. Bayesian GS models (BayesB) achieved the highest prediction accuracy, but including GWAS‐derived fixed effects did not improve predictions, likely due to complexities in modeling major‐effect loci. These results underscore the importance of rust‐specific phenotyping strategies and illustrate the difficulty of integrating GWAS into GS models to dissect complex resistance traits. |
|---|
