Skip to main content
ARS Home » Northeast Area » Wyndmoor, Pennsylvania » Eastern Regional Research Center » Residue Chemistry and Predictive Microbiology Research » Research » Publications at this Location » Publication #384866

Research Project: Data Acquisition, Development of Predictive Models for Food Safety and their Associated Use in International Pathogen Modeling and Microbial Databases

Location: Residue Chemistry and Predictive Microbiology Research

Title: Use of ComBase Data to Develop an Artificial Neural Network Model for Nonthermal Inactivation of Campylobacter jejuni in Milk and Beef and Evaluation of Model Performance and Data Completeness Using the Acceptable.....

Author
item Boleratz, Beth
item Oscar, Thomas

Submitted to: Journal of Food Safety
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/17/2022
Publication Date: 5/5/2022
Citation: Boleratz, B.L., Oscar, T.P. 2022. Use of ComBase Data to Develop an Artificial Neural Network Model for Nonthermal Inactivation of Campylobacter jejuni in Milk and Beef and Evaluation of Model Performance and Data Completeness Using the Acceptable...... Journal of Food Safety. e12983. https://doi.org/10.1111/jfs.12983.
DOI: https://doi.org/10.1111/jfs.12983

Interpretive Summary: Campylobacter bacteria are the leading cause of bacterial foodborne illness in the United States. Models that predict death of Campylobacter in food as a function of time and temperature are valuable tools for helping assess and manage this food safety risk. Existing data were used in the present study to develop a model for predicting Campylobacter death in milk and beef as a function of times and temperatures observed during harvest, processing, distribution, and storage of these products. The study successfully demonstrated that the planned modeling and validation methods can be used in future studies to develop and validate predictive models for Campylobacter in food for use in food safety applications.

Technical Abstract: Models that predict behavior of pathogens in food are valuable tools for food safety. However, there is a shortage of predictive models for Campylobacter jejuni. However, because of a worldwide pandemic, it was not possible to collect new data. Therefore, existing data were used to develop an artificial neural network (ANN) model for predicting death of C. jejuni in milk and beef as a function of strain, temperature, and time. The ANN model was developed using commercial software applications (Excel, NeuralTools). Four ANN were trained and tested, and the best-performing was a General Regression Neural Network (GRNN). Performance of the GRNN model was evaluated using criteria of the Acceptable Prediction Zones (APZ) method in the Validation Software Tool (ValT) for predictive microbiology. In the APZ method, a model is considered to provide acceptable predictions when the proportion of residuals in the APZ (pAPZ) = 0.7. Death of C. jejuni in milk and beef at 20, 30, and 40C > -20C > 1 and 10C. Also, death of C. jejuni in milk > beef at all temperatures except 1C where it was similar. Relative variable impacts were 42.5% for time, 31.5% for temperature, 20.1% for food, and 5.9% for strain. The pAPZ for individual death curves (n = 24) ranged from 0.77 to 1 for milk and from 0.98 to 1 for beef and there were no local prediction problems. Although performance of the GRNN model was acceptable, there were insufficient data to satisfy all criteria for test data. Consequently, the GRNN model failed validation for dependent data and it failed validation for interpolation, and for extrapolation to the other food. The pAPZ for extrapolation to the other food were < 0.7 for all comparisons except at 1C where death of C. jejuni was similar in milk and beef.