Discovering the unknown: detection of emerging pathogens using a label-free light-scattering system

Area of Science:

• Microbiology
• Optical Physics
• Machine Learning

Background:

• Bacterial pathogen recognition is crucial for public health.
• Optical sensors, like BARDOT using forward-scatter phenotyping (FSP), offer rapid detection.
• Current methods struggle with incomplete training datasets, limiting accuracy.

Purpose of the Study:

• To apply the BARDOT system for Salmonella classification using a nonexhaustive framework.
• To develop an automated detection method for unknown bacterial classes.
• To improve the accuracy and specificity of optical-based pathogen recognition.

Main Methods:

• Utilized BARDOT system based on forward-scatter phenotyping (FSP).
• Employed a Bayesian approach to machine learning with a nonexhaustive training dataset.
• Focused on classifying bacteria within the Salmonella class.

Main Results:

• Demonstrated successful classification of Salmonella bacteria.
• Showcased the capability to detect unknown bacterial classes.
• Achieved improved accuracy and specificity compared to traditional methods.

Conclusions:

• The Bayesian approach enhances BARDOT's ability to classify bacteria, even with incomplete data.
• This method allows for the automated detection of previously unencountered bacterial species.
• Advances in optical sensing and machine learning improve biodetection capabilities.