|LIU, XIAOBAO - University Of Massachusetts|
|YANG, MANYUN - University Of Massachusetts|
|Luo, Yaguang - Sunny|
|WANG, SHILONG - University Of Massachusetts|
|TENG, ZI - US Department Of Agriculture (USDA)|
|DILLOW, HAYDEN - University Of Massachusetts|
|GU, TINGTING - University Of Massachusetts|
|REED, KEVIN - University Of Massachusetts|
|SHARM, ARNAV - University Of Massachusetts|
|JIA, ZHEN - University Of Massachusetts|
|YU, HENGYONG - University Of Massachusetts|
|ZHANG, BOCE - University Of Massachusetts|
Submitted to: Nature Food
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/18/2021
Publication Date: 2/18/2021
Citation: Liu, X., Yang, M., Luo, Y., Wang, S., Zhou, B., Teng, Z., Dillow, H., Gu, T., Reed, K., Sharm, A., Jia, Z., Yu, H., Zhang, B. 2021. Machine learning-enabled non-destructive paper chromogenic array detection of multiplexed viable pathogens on food. Nature Food. 2:110-117. https://www.x-mol.com/paperRedirect/1362513694001762304.
Interpretive Summary: The presence of harmful bacteria on food products is a major cause for food-borne illness outbreaks and the associated loss in human lives and the negative impact on the financial wellbeing of the food industry. Early detection and subsequent removal of contaminated foods from the supply chain is a critical measure to ensure food safety. In this study, we report a novel pathogen detection technology using a machine learning empowered paper chromogenic array capable of detecting and differentiating different pathogens with high accuracy. This research benefits pathogen detection method developers and the food industry.
Technical Abstract: Reliable pathogen detection in food is critical to the protection of public health by preventing food-borne illness outbreaks. Here we report a novel technology to detect viable human pathogens in complex food matrices using a paper chromogenic array (PCA) empowered by machine learning. The PCA was developed by impregnating a paper substrate with 5-23 chromogenic dyes, in which exposure to the characteristic volatile organic compounds (VOCs) of the target microorganisms elicits color changes. These color changes were digitized, and used to train a multi-layer Neural Network (NN), giving it strain-specific pathogen detection and quantification abilities with 91-95% accuracy. The trained PCA-NN system was capable of detecting and differentiating between viable Escherichia coli and Escherichia coli O157:H7 in fresh cut lettuce, a realistic and chemically complex environment. Overall, this machine learning empowered PCA technology represents a great potential tool for continuous, accurate, and non-destructive detection of microbial contamination in food.