Enhancing Bacterial Phenotype Classification Through the Integration of Autogating and Automated Machine Learning in...

Area of Science:

• Microbiology
• Computational Biology
• Biotechnology

Background:

• Flow cytometry is a powerful tool for analyzing microbial populations, but data processing is often subjective.
• Automated methods are needed to reduce bias and improve the accuracy of bacterial classification in flow cytometry.

Purpose of the Study:

• To integrate autogating with Automated Machine Learning (AutoML) for enhanced bacterial phenotype classification in flow cytometry.
• To assess the accuracy of gradient-boosting machine (GBM) models in classifying six common soil and groundwater bacterial strains across different metabolic phases.

Main Methods:

• Six bacterial strains (Bacillus subtilis, Burkholderia thailandensis, Corynebacterium glutamicum, Escherichia coli, Pseudomonas putida, Pseudomonas stutzeri) were analyzed using flow cytometry.
• The H2O-AutoML framework was employed to train Gradient-Boosting Machine (GBM) models for bacterial classification.
• Classification accuracy was evaluated across different metabolic phases: lag, early log, late log, and stationary phases.

Main Results:

• The GBM models achieved an overall classification accuracy of 82.34%.
• Highest accuracies were observed in the early log (89.37%), lag (88.43%), and late log (88.06%) phases, with slightly lower accuracy in the stationary phase (80.73%).
• Pseudomonas stutzeri showed high sensitivity and specificity, indicating distinct identification, while Escherichia coli presented classification challenges, especially in the stationary phase.

Conclusions:

• The integration of autogating and AutoML significantly reduces subjective bias in flow cytometry data analysis.
• This approach enhances the reproducibility and accuracy of microbial classification, offering a robust framework for microbial ecology studies.