Inference, exploration, and prediction in hard red winter wheat: multi-environment genomic-to-phenotype framework

Abstract

Wheat breeding must deliver genetic gain while coping with strong environmental variability, polygenic trait architectures, and rapidly expanding data streams. The framework for breeding analytics organizes methods around inference, exploration, and prediction, emphasizing that model usefulness depends on alignment between the biological question, the available data, and realistic validation. Using hard red winter wheat as the primary system, this dissertation first quantifies genotype-by-environment interaction and genetic progress from long-term multi-environment trials in Kansas, showing how models and stability analyses can separate genetic signal from environmental and design noise to support cultivar selection and deployment. It then investigates the genetic architecture of key developmental, floral, and disease traits in a D-genome–enriched population derived from Aegilops tauschii, integrating controlled greenhouse and field phenotyping. Finally, it evaluates multi-omics prediction across locations, years, and new genotypes by combining genomics with environmental and phenotyping features under breeding-relevant validation schemes, demonstrating that the value of added data layers and model complexity is scenario- and trait-dependent. Collectively, these results support an integrated, goal-driven approach to wheat improvement that couples robust multi-environment inference with hypothesis-guided discovery and deployment-oriented prediction, providing actionable guidance for selection, parent choice, and cultivar placement.