Predicting microbial growth: the consequences of quantity of data

Area of Science:

• Microbiology
• Food Science
• Mathematical Modeling

Background:

• Salmonellae contamination poses a significant risk to food safety.
• Accurate modeling of bacterial growth is crucial for predicting spoilage and ensuring food safety.
• Previous models have focused on various environmental factors affecting bacterial growth.

Purpose of the Study:

• To investigate the impact of data selection on mathematical models of salmonellae growth.
• To assess the consequences of systematic data removal on model accuracy.
• To demonstrate the risk of developing erroneous models due to insufficient data.

Main Methods:

• Developed a mathematical model for salmonellae (mixed inoculum of Salmonella thompson, S. stanley, and S. infantis) growth.
• Varied pH level, NaCl concentration, and storage temperature in a laboratory medium.
• Systematically removed data points using three different strategies.
• Analyzed model performance using three-dimensional plots of fitted response surfaces.

Main Results:

• Differences in model predictions were observed between the full dataset and reduced datasets.
• Incomplete data led to significant alterations in the response surface plots.
• The study demonstrated that data reduction strategies can result in erroneous model outputs.

Conclusions:

• Mathematical models for bacterial growth are highly sensitive to the data used for their development.
• Incomplete or selectively removed data can lead to misleading predictions and erroneous models.
• Researchers must exercise caution and ensure data representativeness when developing and validating predictive models for food safety.