Direct Antimicrobial Resistance Prediction from MALDI-TOF mass spectra profile in clinical isolates through Machine Learning

Early administration of effective antimicrobial treatments improves the outcome of infections. Culture-based antimicrobial resistance testing allows for tailored treatments, but takes up to 96h. We present a revolutionary approach to predict resistance with unmatched speed within 24h, using calibrated logistic regression and LightGBM-classifiers trained on species-specific MALDI-TOF mass spectrometry measurements. For this analysis, we created an unprecedented large, publicly-available dataset combining mass spectra and resistance information. Our models provide highly valuable treatment guidance 12–72h earlier than classical approaches. Rejection of uncertain predictions enables quality control and clinically-applicable sensitivities and specificities for the priority pathogens , , and .