Predicting Staphylococcus aureus spa types using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry and machine learning

Abstract

Staphylococcus aureus is a major cause of intramammary infection (IMI) in dairy cattle, with persistent infections causing economic losses and facilitating transmission within herds. spa typing is a useful tool for epidemiological investigations but requires molecular testing. This study evaluated whether MALDI-TOF MS spectra combined with machine learning could predict the six most common S. aureus spa types found in Canadian dairy herds. A total of 373 previously typed isolates were analyzed by MALDI-TOF MS, and random forest models were trained using processed spectral data. Principal component analysis indicated that some spa types were distinguishable based on their spectra. Model performance was good for spa types t529 (AUC = 0.88), t605 (AUC = 0.94), and t13401 (AUC = 0.89), but limited for t359, t267, and t2445 (AUC = 0.53-0.68). These findings suggest that MALDI-TOF MS can accurately predict certain S. aureus spa types at no additional laboratory cost, offering a practical alternative to more resource-intensive molecular typing methods for epidemiological investigations.