The use of predictive models in the context of food spoilage: the case of white pudding

Food spoilage is a major issue for the food industry and consumers as products become unacceptable for human consumption leading to significant food waste and economic losses. This study combines the use of predictive microbiology and metagenetics in order to predict bacterial evolution in Belgian white pudding. The ecology of the product was studied at several times during the storage at constant temperatures, and under three different packaging (food wrap, modified atmosphere 30% CO2 – 70% N2 and vacuum packaging), by association of classical microbiological plate counting and 16S rRNA metagenetic analysis on each sample. The bacterial evolution could thus be deduced for the two major spoilage populations in the product, Brochothrix thermosphacta and Pseudomonas spp. The growth parameters were estimated using the nlsMicrobio package from R and then used to simulate the microbial behavior in dynamic conditions with three different softwares: ComBase, Sym’Previus and baranyi growth function in R. These results are compared with validation curves, obtained from these durability studies. A relatively good agreement was obtained between the validation data set and the simulations, showing that the approach combining the metagenetics and the simulations based on an accurate database is promising. Performance factors (bias and accuracy factors) indicated no significant structural deviation of the maximal growth rates simulations between observed and predicted values with R and Sym’Previus. An overestimation was mainly observed with R, while an underestimation was generally observed with Sym’Previus and ComBase. None of predictive simulations give an identical microbial curve that the validation data set, but all models show relatively good statistical fittings. This work gives a proof of concept on the feasibility to combine predictive models and metagenetics in order to predict bacterial evolution using different predictive tools. In the future, predictive models needed to be more accurate by taking into account as many growth parameters as possible.