The fast identification of microbial species in clinical samples is essential to provide an appropriated antiobiotherapy to the patient and to reduce the prescription of broad spectrum antimicrobials leading to antibioresistances. We have developed a new strategy for the fast identification of bacterial species in urine using a specific peptide signature designed by combination of proteomics data and machine learning approaches. Thereby, we have developed a 82 peptides signature which, we monitored by targeted proteomics, is able to distinguish between the 15 species the most frequently found in Urinary Tract Infections (UTIs). Our method allows the bacterial identification in less than 4 hours without culturing.