Machine learning-supported sensor array for multiplexed foodborne pathogenic bacteria detection and identification

Background: Foodborne pathogens present a significant challenge to food safety. Traditional culture-based methods are often time-consuming and labor-intensive, while newer technologies have limitations, such as requiring specialized expertise or costly equipment. This has driven the development of sensor arrays, like electronic noses (e-noses) and optical sensor arrays, which use multiple cross-reactive sensor elements to generate unique fingerprints for various analytes. Scope and approach: This review highlights recent advances in the design of sensor arrays and the materials commonly used as their building blocks. We outline four key principles for constructing sensor arrays: detecting volatile organic compounds (VOCs), antibody-based sensors, bacterial surface physiology and microenvironments, and metabolic activity. We also discuss the use of machine learning (ML) in sensor array interpretation and output. Additionally, we explore the challenges in multiplexed pathogen detection and emerging trends in the field. Key findings and conclusions: Bacterial cell envelope microenvironments and metabolic activities have received the most attention in the development of sensor arrays. ML models play a critical role not only in pattern recognition but also in tasks like data preprocessing, such as correcting signal drift in e-noses and handling outliers. Challenges like small datasets are addressed through potential solutions such as few-shot learning and leave-one-out cross-validation. Sensor arrays show great promise for in-field pathogen identification, offering valuable benefits to food producers and processors alike.