Genomic selection for multi-trait crop improvement in wheat (Triticum aestivum L.): a practical modeling approach

Wheat (Triticum aestivum L.) is a staple crop essential for global food security. Enhancing its yield and securing resistance to biotic/abiotic stress to help address the issues of population growth and climate change, however, remain urgent challenges. This study aimed to enable data-driven digital breeding in wheat by characterizing the genetic basis of major agronomic traits and evaluating the effectiveness of genomic selection. We analyzed 566 globally diverse bread wheat accessions for 15 traits spanning morphology, development, disease resistance, and grain quality. A genome-wide association study identified 32 significant marker–trait associations, supported by candidate gene enrichment around associated loci, providing biologically plausible targets for improvement. Moreover, we benchmarked a broad suite of genomic prediction models. Prediction accuracy varied among traits and models, reflecting differences in heritability and genetic architecture. Linear mixed and Bayesian models were generally superior, with EGBLUP offering the best compromise between accuracy and computational cost, and RKHS and Random Forest providing robust alternatives. In contrast, the machine learning model SVM exhibited poor prediction accuracy. Morphological and developmental traits such as lodging, culm length, days to heading, and days to maturity achieved high prediction accuracies (up to 0.65), enabling reliable early-generation selection, whereas disease resistance and quality traits were less predictable and will require multi-trait and more flexible modeling. Crucially, genomic estimated breeding values from our models produced clear phenotypic gains in an independent, genetically distant population, demonstrating that genomic selection can be both accurate and portable across germplasm pools. These results provide concrete evidence that genomic-informed selection is a transformative tool for wheat improvement and outline a practical roadmap for implementing digital breeding pipelines to accelerate genetic gain and deliver high-yielding, climate-resilient wheat varieties.