Candidemia is the fourth most common bloodstream infection, with Candida albicans being the most common causative species. Success in reducing the associated morbidity and mortality has been limited by the inadequacy and time delay of currently available diagnostic modalities. Focusing on host response to infection, we used a murine model to develop a blood gene expression signature that accurately classified mice with candidemia and distinguished candidemia from Staphylococcus aureus bacteremia. Validation of the signature was achieved in an independent cohort of mice. Genes represented in the signature have known associations with host defense against Candida and other microorganisms. Our results demonstrate a temporal pattern of host molecular responses that distinguish candidemia from S. aureus-induced bacteremia and establish a novel paradigm for infectious disease diagnosis.