In silico modelling to predict the odor profile of food from its molecular composition using experts' knowledge, fuzzy logic and optimization: Application on wines

Aroma analysis follows a well-established procedure which provides a list of odorants that contribute to a given food aroma. However, such a procedure does not allow establishing the actual sensory profile of the food because the perceptual influence of mixed odorants is poorly considered. To improve the aroma analysis efficiency, we explored an innovative strategy which combines classical aroma analysis results with expert knowledge on aroma formulation through a modelling approach relying on fuzzy logic and optimization. The model queries analytical and sensory databases in order to predict the odor profile of wines, namely the intensity of a series of odor descriptors. By comparing the output of the model and the actual sensory data we estimated that the model can predict sensory scores in a promising way.