From deleterious alleles to digital infrastructures : bridging quantitative genetics and data science for wheat breeding

Abstract

Data-driven approaches are increasingly central to plant breeding, with quantitative genetics providing the conceptual backbone. This thesis identifies contemporary data sharing practices and introduces the concept of a data cohort—a functional unit comprising breeding data from a defined study or experiment. Using cohorts from experimental hybrid populations, it first demonstrates that genomic prediction models achieve higher accuracy when accounting for evolutionary deleterious substitutions. The work then explores federated data sharing infrastructures to integrate these cohorts into Big Data. Building on this, the thesis shows that linking genomic, phenotypic, and environmental information reveals genotype–phenotype and environment–phenotype signals, informing actionable breeding decisions. Finally, by modeling genotype × environment interactions, it proposes the use of enviromically adapted varieties to help close the global yield gap.