Machine learning for microbial identification and antimicrobial susceptibility testing on MALDI-TOF mass spectra: a...

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Matrix-assisted laser desorption/ionization and time-of-flight mass spectrometry (MALDI-TOF MS) is a key technology in microbiology for rapid species identification.
• Machine learning (ML) techniques are increasingly applied to MALDI-TOF MS data to improve accuracy and efficiency.

Purpose of the Study:

• To systematically review and evaluate studies that utilize ML for the analysis of MALDI-TOF mass spectra.
• To assess the application of ML in microbial species identification and antimicrobial susceptibility determination using MALDI-TOF MS data.

Main Methods:

• A systematic literature search was conducted across PubMed/Medline, Scopus, and Web of Science.
• Studies employing ML for MALDI-TOF MS analysis in microbiology were included, excluding case studies and reviews.
• A quality assessment of the ML models was performed according to PRISMA guidelines.

Main Results:

• Out of 36 included studies, 27 focused on species identification and nine on antimicrobial susceptibility testing.
• Commonly used ML algorithms included Support Vector Machines, Genetic Algorithms, Artificial Neural Networks, and Quick Classifiers.
• While most studies reported interpretability and clinical applications, only a small fraction (11.11%) validated their ML algorithms on external datasets.

Conclusions:

• The use of ML with MALDI-TOF MS is growing for optimizing microbial analysis.
• Current ML-supported approaches have limitations that require addressing for widespread clinical adoption.
• Further validation and refinement are necessary to integrate these advanced analytical methods into routine clinical practice.