Abstract
BACKGROUND: The traditional method for assessing the capacity of a microorganism to produce biofilm is generally a static in vitro model in a multi-well plate using the crystal violet (CV) binding assay, which takes 96 h. Furthermore, while the method is simple to perform, its reproducibility is poor. OBJECTIVE: We evaluated whether matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) could make it possible to differentiate between high-and low-biofilm-producing microorganisms on 24-h cultures of Staphylococcus aureus and Candida albicans. METHODS: We included 157 strains of S. aureus and 91 strains of C. albicans obtained from the blood cultures of patients with bacteremia/candidemia. We tested biofilm production using the CV binding assay as the gold standard to classify strains as low or high biofilm producers. We then applied MALDI-TOF MS to create a machine learning-based predictive model using 40 strains of S. aureus and C. albicans, each with extreme absorbance values, and validated this approach with the remaining 117 and 51 strains using the random forest algorithm and the support vector machine algorithm, respectively. RESULTS: Overall, 81.2% of the S. aureus strains (95/117) and 74.5% of the C. albicans strains (38/51) used for validation were correctly categorized, respectively, as low and high-biofilm-producing. CONCLUSION: Classification based on MALDI-TOF MS protein spectra enables us to predict acceptable information about the capacity of 24-h cultures of S. aureus and C. albicans to form biofilm.