This paper proposes a feature selection framework for statistical machine learning–based bacterial genome-wide association studies aimed at uncovering resistance-causing traits. Using a well-defined Staphylococcus aureus pangenome as ground truth for labeling true causal variants, we demonstrate improved control for population structure and enhanced interpretability through the explicit incorporation of genomic context derived from graph-structured data. Our framework successfully uncovers resistance-causing traits in 9 of 14 resistance phenotype contexts, highlighting differences in capabilities compared to traditional methods as well as the possibility for the presence of more complex resistance profiles that may further conflict with model assumptions and mask causal signal.