A Framework for Predictive Sensory Quality in Processed FoodsUpon Variable Nutrient Interactions

Thesis Abstract

This study addresses the challenge of predicting consumer-perceived sensory quality in processed foods when nutrient compositions vary across formulation and processing conditions, a gap that limits product optimization and nutritional labeling accuracy. The aim is to develop a robust framework that links dynamic nutrient interactions to sensory outcomes, enabling predictive decision-support for product developers. Specific objectives are (1) to quantify how macronutrient (protein, fat, carbohydrate) and micronutrient (vitamin A, iron, zinc) profiles interact with processing variables (thermal treatment, fat emulsification, Maillard reaction intensity) to influence sensory attributes such as aroma, texture, taste, and overall acceptability; (2) to identify mediating mood-arousal and satiety cues that modulate sensory perception; (3) to construct and validate a predictive model integrating physicochemical measurements, instrumental analyses, and consumer sensory data; and (4) to propose a practical framework for real-time sensory quality prediction in product development workflows. A cross-sectional, mixed-methods design was employed. The study sampled 32 reformulated ready-to-eat meals and beverages across three product categories (dairy-based desserts, savory entrées, and plant-based beverages) manufactured under four controlled processing conditions, yielding 96 product variants. Sensory data were collected from 180 trained panelists using a structured descriptive analysis protocol, supplemented by consumer acceptance testing with 360 untrained panelists. Physicochemical characterization included near-infrared spectroscopy for macro- and micronutrient estimation, differential scanning calorimetry for thermal behavior, rheometry for textural profile analysis, gas chromatography–mass spectrometry for volatile compounds, and high-performance liquid chromatography for key non-volatile taste-active compounds. Instrumental measurements were synchronized with consumer data via time-aligned sampling to capture dynamic changes in sensory perception. Data analysis encompassed multivariate techniques, including partial least squares regression (PLS-R) to relate nutrient- and processing-variable matrices to sensory descriptors, hierarchical linear modeling to account for nested panelist effects, and ANOVA for group comparisons. Mediation analyses investigated the role of aroma compound concentrations and texture metrics as mediators between nutrient profiles and sensory outcomes. The framework is anchored in the hedonic theory and the nutrition-pleasure model, with theoretical integration from the food-structure–sensory interaction framework and the nutrition information processing model to explain perceptual variability under nutrient perturbations. Expected findings indicate that specific nutrient interactions (e.g., high fat with elevated Maillard-reactive compounds) synergistically enhance certain aroma and mouthfeel attributes while diminishing others, leading to non-linear effects on overall acceptability. The predictive model is anticipated to explain at least 68% of the variance in global sensory scores across product variants, with mean absolute prediction error within 0.45 on the nine-point hedonic scale. The study contributes to knowledge by operationalizing a generalizable framework that connects nutrient interaction effects with sensory quality under diverse processing conditions, validated across multiple food systems, and provides a decision-support tool for formulation engineers and sensory scientists. It offers methodological advancement in combining instrumental chemistry, rheology, and consumer perception within a unified predictive framework, and informs regulatory assessment by improving accuracy of sensory-based nutrition labeling implications. Based on findings, practical recommendations include guidelines for nutrient-processed interaction management to optimize sensory quality without compromising nutritional targets, suggested thresholds for processing-induced aroma generation, and a roadmap for embedding the framework into product development pipelines. The study concludes that predictive sensory quality in processed foods can be substantially improved by modeling nutrient interaction effects within processing contexts, and recommends extending the framework to include longitudinal consumer testing, broader nutrient spectrums, and real-time sensing technologies to support adaptive manufacturing.