News Center

Computer model predicts Listeria risk-reduction outcome

March 10, 2020

No two packinghouses or fresh-cut processing facilities are alike when it comes to the potential contamination risk from Listeria monocytogenes and the corrective measures implemented, should the pathogen be detected.

Key Take-Aways

* Computer models will help participating packinghouses and fresh-cut facilities customize Listeria prevention and control programs.

* The industry in general will be able to use a number of the learnings to justify their Listeria control programs.

* Models would not be possible without industry cooperators' "open-door" attitude.

That is why Renata Ivanek, Ph.D., and an associate professor with Cornell University, is leading the effort with her project, "Modeling tools for design of science-based Listeria environmental monitoring programs and corrective action strategies," to develop a science-based computer model to help four individual facilities customize Listeria risk-reduction and corrective measures.

Once completed, the models will allow the plants' food safety managers to ask "what if" questions about potential risk-reduction or corrective measures without having to actually conduct trials in their facilities.

Although the simulation will be most useful to the four participating facilities, she said, the industry in general will be able use a number of the learnings to justify their Listeria control programs.

The models Ivanek's team are developing are based on real-world observations from four facilities, while most other models are based on hypothetical facilities and occurrences. 

This is not the first model of its type Ivanek has been involved with, either. She and her team previously developed a computer program, known as Environmental Monitoring with an Agent-Based Model of Listeria (EnABLe), for cold smoked-salmon processing plants.

As with the earlier efforts, she said developing these models wouldn't be possible without great industry cooperators who allowed her team into their facilities. Two packinghouses and two fresh-cut operations participated in this effort.

"If you don't have facilities that allow you to come in with open doors, observe their production, collect samples and talk to us, there would be no model," Ivanek said. "Their food safety managers were also interested in having such tools."

"And it's not just being able to have test results, but observing workers moving, how much they move from their stations, how they clean and sanitize, how the water flows in the facilities, and how the equipment is moved from one location to the other. There are a lot of observations going on so we can represent that facility accurately."

Joining her in the project is co-principal investigator, Martin Wiedmann, Ph.D., and a professor of food science at Cornell University. Ivanek said Wiedmann has been invaluable because of his long-standing expertise in environmental modeling and food safety, specifically with Listeria.

"He provided the applied science to the project," she said. "His microbiological expertise and many years of environmental modeling in food processing facilities were helpful as well."

Two food science graduate students, Genevieve Sullivan and Cecil Wilfried Barnett-Neefs, also are developing the models and are an integral part of the project, Ivanek said. Developing computer models like these requires a fine balance between incorporating enough real-world data for it to be useful but not so much that the job is never completed, Ivanek said.

"In the model simulations, we monitored different locations over time and have the luxury of sampling many, many locations over time, which we know would never be done in real life," she said.

In essence, the researchers are creating a digital twin reality that tries to, as closely as possible, parallel happenings in the produce facilities.

Once they finish a model prototype, they run it for predictive results and then validate the outcome, which Ivanek said is the most important step.

"If something is not represented correctly, you go back to the equations and work on them until we can have a system that's checked for interim and final predictive stress tests," she said. "If you change a parameter in a certain way, even if it's unrealistic, you expect the type of change that will be in the model results."

To ensure the algorithms make sense, the researchers asked experts to read and review the coding. As it's written currently, the model is complex and carries a steep learning curve. But Ivanek said she'd like to eventually incorporate it into an easy-to-use web-based interface with simple "click and play" functions.

"For a decision tool to be found useful by many, it has to have a really simple interface," Ivanek said.

Twitter | Facebook