Skip to main content
ARS Home » Northeast Area » Wyndmoor, Pennsylvania » Eastern Regional Research Center » Characterization and Interventions for Foodborne Pathogens » Research » Publications at this Location » Publication #376557
Research Project: Shiga Toxin-Producing Escherichia coli in Biofilms and within Microbial Communities in Food

Location: Characterization and Interventions for Foodborne Pathogens

Title: Utilizing the microbiota and machine learning algorithms to assess risk of Salmonella contamination in poultry rinsate

Author
item BOLINGER, HANNAH - Clearwater Labs
item TRAN, DAVID - Clearwater Labs
item HARARY, KENNETH - Clearwater Labs
item Paoli, George
item GURON, GISELLE - Collaborator
item NAMAZI, HOSSEIN - Clearwater Labs
item KHAKSAR, RAMIN - Clearwater Labs

Submitted to: Journal of Food Protection
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/16/2021
Publication Date: 5/20/2021
Citation: Bolinger, H., Tran, D., Harary, K., Paoli, G., Guron, G., Namazi, H., Khaksar, R. 2021. Utilizing the microbiota and machine learning algorithms to assess risk of Salmonella contamination in poultry rinsate. Journal of Food Protection. doi.org/10.4315/JFP-20-367.. https://doi.org/10.4315/JFP-20-367.
DOI: https://doi.org/10.4315/JFP-20-367

Interpretive Summary: Traditional microbiological testing methods are slow, and even molecular-based techniques generally rely on culture-based enrichment to overcome low limits of detection. Recent advancements in sequencing technologies and computing power may make it possible to utilize machine learning to identify trends in microbiome data (an analysis of all the different microbes present) to predict the presence or absence of specific pathogens. In this study, 299 poultry wash samples from various locations along the processing chain were analyzed to determine if the microbiota could predict the risk of a sample containing Salmonella. An assessment of various machine learning tools ability to predict which samples were Salmonella-positive or -negative based on the microbiota from unenriched samples resulted in a predictive algorithm with high accuracy (86.7%), sensitivity (80%), specificity (90%). This data adds to the small but growing body of literature exploring novel ways to utilize microbiome data for predictive food safety

Technical Abstract: Traditional microbiological testing methods are slow, and even molecular-based techniques generally rely on culture-based enrichment to overcome low limits of detection. Recent advancements in sequencing technologies and computing power may make it possible to utilize machine learning to identify trends in microbiome data to predict the presence or absence of specific pathogens. In this study, 299 poultry rinsate samples from various locations along the processing chain were analyzed to determine if the microbiota could inform about the risk of a sample containing Salmonella. Each sample was culture confirmed as Salmonella-positive or -negative following modified USDA MLG protocols. The culture confirmation result was then used as a reference to compare with the output of microbiota analysis. Pre-chill samples tested positive (71/82) at a higher frequency than did post-chill samples (30/217) and were found to contain greater microbial diversity. Analysis of variance (ANOVA) identified a significant effect of chilling on the number of genera present, but analysis of similarities (ANOSIM) failed to provide evidence for microbial dissimilarity between pre and post-chilled samples. An assessment of various machine learning algorithms’ ability to predict which samples were Salmonella-positive or -negative based on the microbiota from unenriched samples concluded with the decision to move forward with a Random Forest algorithm whose performance was as follows: accuracy (86.7%), sensitivity (80%), specificity (90%). This data adds to the small but growing body of literature exploring novel ways to utilize microbiome data for predictive food safety.