Decoding the Taste Puzzle: Toward a better understanding of taste profiling

Abstract

Personalizing a diet based on individual taste preferences can lead to healthier dietary habits. However, the lack of comprehensive data on meal taste profiles limits effective personalization. To address this gap, our study employed both a survey study and a tasting trial to gather a detailed taste profile of 18 expert-designed recipes. A total of 2046 participants (55.5% female, mean age of $47.4\pm 13.4$) in the survey, and in the tasting trial, 48 participants (83.3% female, mean age of $37.6\pm 14.5$) provided insights into their taste perception, their familiarity with meals and overall liking. Our data revealed a substantial variability in survey responses, suggesting relying solely on survey data may not yield sufficiently accurate data to predict the taste profile of meals. Among all tastes, sweetness emerged as the most precisely predictable taste, whereas bitter taste posed significant challenges. Comparative analysis using a linear mixed model showed that ingredient-based data is comparable to or slightly better predictor of the taste profile than the survey, except for sweetness. Furthermore, hierarchical analysis underscored the critical role of taste interactions in enhancing the model fit. Future research should aim to collect more comprehensive data, encompassing a greater variety of meals to cover broader taste and trigeminal profiles. Our study sets the groundwork for more sophisticated predictive modeling for dietary customization.