Machine-Learning Classification of Bacteria Using Two-Dimensional Tandem Mass Spectrometry.

Area of Science:

• Microbiology
• Analytical Chemistry
• Computational Biology

Background:

• Biothreat detection requires specialized identification methods for bacterial pathogens.
• Existing methods are often time-consuming and resource-intensive.

Purpose of the Study:

• To evaluate the utility of 2D MS/MS coupled with machine learning for rapid bacterial species identification.
• To demonstrate a high-accuracy method for biothreat agent classification.

Main Methods:

• Bacterial cell extracts were analyzed using a modified linear quadrupole ion-trap mass spectrometer with 2D MS/MS.
• Lipid profiles were generated and analyzed using random forest, k-nearest neighbor, and multilayer perceptron models.
• 2D MS/MS data was also processed using convolutional neural networks for image recognition.

Main Results:

• Bacterial species classification achieved accuracy greater than 99% across all tested machine learning models.
• Rapid analysis (within 5 seconds) of bacterial cell extracts was demonstrated.
• The lipid profiles provided sufficient information for high-confidence species-level identification.

Conclusions:

• 2D MS/MS coupled with machine learning offers a rapid and highly accurate method for bacterial pathogen identification.
• This approach significantly advances biothreat detection capabilities.
• The information-rich nature of 2D MS/MS data supports its application in microbial diagnostics.