Prediction of Listeria monocytogenes Clonal Complexes from Multilocus Variable Number Tandem Repeat Analysis Patterns...

Area of Science:

• Microbiology
• Bioinformatics
• Food Science

Background:

• Multilocus variable number tandem repeat analysis (MLVA) is a cost-effective molecular subtyping method for bacterial tracking.
• Unlike MLST, MLVA lacks a standardized database for Listeria monocytogenes, limiting its comparative analysis.
• Whole genome sequencing is resource-intensive, making MLVA a valuable alternative for many laboratories.

Purpose of the Study:

• To develop a predictive model for assigning MLVA patterns to Listeria monocytogenes clonal complexes (CCs).
• To assess the accuracy of machine learning in predicting CCs from MLVA data.
• To demonstrate the utility of machine learning in enhancing existing molecular subtyping data.

Main Methods:

• A predictive model was created using the XGBoost machine learning technique.
• The model was trained and validated using a 5-loci MLVA scheme.
• A simulation protocol for updating the model with new subtypes was developed.

Main Results:

• The XGBoost model accurately predicted CCs from MLVA patterns with approximately 85% (±4%) accuracy.
• The model demonstrated strong congruence with MLST-based CC assignments.
• A straightforward update protocol was simulated for future subtype emergence.

Conclusions:

• Machine learning techniques can effectively predict clonal complexes from MLVA data.
• This approach adds significant value to legacy molecular subtyping data in food processing environments.
• MLVA combined with machine learning offers a powerful tool for bacterial surveillance and food safety.