Classification of Latilactobacillus sakei subspecies based on MALDI-TOF MS protein profiles using machine learning...

Area of Science:

• Microbiology
• Bioinformatics
• Food Science

Background:

• Latilactobacillus sakei is vital in fermented foods, with subspecies having distinct roles (fermentation vs. spoilage).
• Accurate differentiation of L. sakei subspecies is critical for food quality and safety.
• Matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF MS) is standard for microbial identification but lacks subspecies resolution.

Purpose of the Study:

• To develop a novel method for differentiating L. sakei subspecies using MALDI-TOF MS and machine learning.
• To evaluate the performance of machine learning algorithms for high-resolution microbial classification.
• To provide a cost-effective tool for the food industry to monitor L. sakei subspecies.

Main Methods:

• Collected 227 L. sakei strains and obtained 908 MALDI-TOF MS spectra.
• Applied machine learning algorithms including Partial Least Squares-Discriminant Analysis (PLS-DA), Principal Component Analysis-K-Nearest Neighbor (PCA-KNN), Support Vector Machine (SVM), and Random Forest (RF).
• Compared machine learning model performance against the Biotyper database for subspecies identification.

Main Results:

• The Biotyper database achieved only 68.7% accuracy at the subspecies level.
• Machine learning models demonstrated high performance: PLS-DA (0.823), PCA-KNN (0.914), SVM (0.903).
• Random Forest (RF) achieved the highest accuracy (0.954) and an AUROC of 0.99, significantly outperforming other methods.

Conclusions:

• Machine learning combined with MALDI-TOF MS provides a powerful, high-resolution method for L. sakei subspecies classification.
• This approach significantly improves upon traditional identification methods for microbial subspecies.
• The developed model offers a promising, efficient solution for monitoring L. sakei in the food industry.