A Machine Learning-Enabled SERS Sensor: Multiplex Detection of Lipopolysaccharides from Foodborne Pathogenic Bacteria.

Abstract

Foodborne pathogenic bacteria pose a global threat to public health, causing dangerous and expensive outbreaks. While therapeutics are being developed to target antibiotic-resistant and foodborne pathogenic bacteria, approaches to minimize their spread are also crucial. As estimated by the World Health Organization, each year around 600 million cases of foodborne illnesses are reported, with over 420,000 deaths annually. Lipopolysaccharides (LPS) are glycolipids found on the outer membrane of Gram-negative bacteria, serving as biomarkers for pathogenic strains. With a clear need for a sensing platform to quickly and cost-effectively detect and identify microbial pathogens, this work focuses on surface-enhanced Raman scattering (SERS)-based bacterial sensors for the detection of LPS from various pathogenic bacteria. Specifically, linear polymer affinity agents with a broad affinity for LPS immobilized on metal film over nanosphere plasmonic substrates and the integration of machine learning analysis are used to characterize and discriminate between the Raman "fingerprints" of LPS from Salmonella typhimurium, Escherichia coli O111:B4, and Escherichia coli O26:B6. Our findings demonstrate the feasibility of a linear polymer-based SERS sensing platform coupled to machine learning analysis with principal component analysis and support vector machines for the sensitive detection and differentiation of LPS from different bacterial species, even in a complex food matrix like apple juice, highlighting its potential for real-world applications in food safety monitoring.