There is a newer version of the record available.

Published September 18, 2021 | Version v1
Dataset Open

Rapid identification of MRSA using mass spectrometry and machine learning from over 20000 clinical isolates

  • 1. AI Innovation Center, China Medical University Hospital, Taichung City, Taiwan
  • 2. Department of Laboratory Medicine, China Medical University Hospital, Taichung City, Taiwan
  • 3. Proteomics Core Laboratory, Department of Medical Research, China Medical University Hospital, Taichung City, Taiwan
  • 4. Department of Neurosurgery, China Medical University Hospital, Taichung City, Taiwan
  • 5. Laboratory Medicine, Feng Yuan Hospital, Ministry of Health and Welfare, Taichung City, Taiwan
  • 6. Department of Laboratory Medicine, Taipei Medical University-Shuang Ho Hospital, Ministry of Health and Welfare, New Taipei City, Taiwan
  • 7. Department of Laboratory Medicine, National Taiwan University Hospital Yunlin Branch, Yunlin County, Taiwan
  • 8. Department of Laboratory Medicine, Hualien Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, Hualien County, Taiwan

Description

Rapidly identifying methicillin-resistant Staphylococcus aureus (MRSA) with high integration in the current workflow is critical in clinical practices. We proposed a matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-TOF MS) based machine learning model for rapidly MRSA prediction, the model was evaluated on a prospective test and four external clinical sites. On the dataset comprising 20359 clinical isolates, the area under the receiver operating curve of the classification model was 0.78–0.88. These results were further interpreted using SHapely Additive exPlanations. The important MRSA feature, m/z 6590–6599, was identified as a UPF0337 protein SACOL1680 and has a lower binding affinity or no docking results than to UPF0337 protein SA1452, which mainly detected in methicillin-susceptible S. aureus (MSSA). Our MALDI–TOF MS-based ML model for the rapid MRSA identification can be easily integrated into the current clinical workflows and can further support physicians prescribe proper antibiotic treatments. 

Files

Files (20.3 MB)

Name Size Download all
md5:bcd867037fb8b1a2d3838625d9de4144
10.8 MB Download
md5:7c3fbdb276a8039bae40c7bb76910762
9.4 MB Download