This project explored how genetic variation at the population level underlies host specialization and ecological adaptation in fungal pathogens. We developed a novel genome-host association (GHA) framework that adapts genome-wide association mapping to natural field epidemics, using host cultivar identity as a proxy for pathogen selective environment. Applying this approach to a single wheat field epidemic, we analyzed whole-genome sequences from 832 Z. tritici isolates sampled across twelve wheat cultivars. This strategy revealed a polygenic architecture of host specialization, identifying between two and twenty pathogen genes associated with individual host cultivars, including known avirulence factors and numerous previously uncharacterized pathogenicity-related genes. Crucially, we supplemented theses statistical studies with functional validation experiments, generating targeted deletions for sixteen GHA-identified gene candidates, of which ten showed significant effects on the pathogen virulence, demonstrating the power of the GHA framework.