Discriminating ST3 and non‐ST3 Staphylococcus lugdunensis using MALDI‐TOF and machine learning analysis

Purpose: Staphylococcus lugdunensis has gradually become an important pathogen because of its broad range of infectious symptoms, especially the high mortality associated with endocarditis. Previous epidemiological surveillance has shown that most oxacillin-resistant isolates belong to the ST3 group, the predominant population in communities. Therefore, there is a need to rapidly and efficiently evaluate antimicrobial resistance in S. lugdunensis. Methods: To rapidly and efficiently discriminate between ST3 and non-ST3 populations, a matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) platform with a machine learning approach was used to analyze 107 clinical isolates collected between 2010 and 2014. Results: Our data showed that the signals located at both 3676 m/z and 7352 m/z in ST3 isolates varied from those of non-ST3 isolates (3683 m/z and 7366 m/z). Further, 81 isolates collected from 2016 to 2019 were used to evaluate this finding; 59 isolates were classified as ST3, and multilocus sequence typing (MLST) validation confirmed that 50 isolates belonged to ST3. Using MLST, the remaining 22 isolates classified as non-ST3 were found to be non-ST3 types. Overall, our approach had a sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of 100%, 71%, 85%, 100%, and 89%, respectively. Conclusion: Our data demonstrate that MALDI-TOF provides a reliable way to discriminate between ST3 and non-ST3 S. lugdunensis , which is valuable for clinical identification applications.