Plant-pathogenic microbes, including the wheat fungal pathogen Zymoseptoria tritici, adapt to their host environment. In plants, genome-wide association studies (GWAS) have been extensively used to uncover the complexity of local adaptation and disease resistance. However, the application of GWAS to decipher the mechanisms underlying fungal pathogenicity and host adaptation trails far behind. Here, we established a genome-host association (GHA) approach to infer statistical associations between pathogen allele frequencies and host of origin for 832 fungal strains isolated from twelve different host cultivars during a natural field epidemic. We identified from two to twenty genes associated with specialization to the different wheat cultivars, including one known effector gene that provided a proof-of-concept for our GHA approach, as well as two new virulence-related genes that we validated with targeted gene knockouts. Our study highlights the polygenic genetic architecture of host adaptation and provides a novel application of GWAS in plant pathogens that transcends the limitations imposed by traditional phenotyping methods.