Multivariate relationships between instrumental attributes, microstructure, sensory profiles and consumer preference in roasted peanuts (Arachis spp)

The objective of this research was to determine how raw material and process technology selection can affect the organoleptic characteristics of roasted peanuts, and further identify which of those characteristics drive liking with European consumers. Twelve different raw peanuts of various market types, origins and grades were treated by eleven different process (maceration in water, aqueous glucose and at different pH followed by frying or baking), resulting in 134 unique samples, which were profiled by a sensory panel (SPECTRUM, DSA) and analysed for colour (CIELAB), fatty acid composition (FAMEs-GC-MS), headspace volatile composition (DHS-GC-MS, SPME-GC-MS, and GC-MS-O), sugar profile (ion chromatography), and textural characteristics (large deformation compression). Principal Component Analysis, Canonical Variate Analysis and General Linear Model regressions were used to identify differences in sensory attributes, fatty acid and headspace volatile profiles, and to relate them to raw materials and process conditions. Process selection had a large impact on the final sensory characteristics. Specifically, baking reduced ‘roasted peanut’ and ‘dark roast’ and increased ‘raw bean’ aromas compared to frying. Maceration significantly increased ‘roasted peanut’ and ‘dark roast’, and reduced ‘sweet’, ‘raw bean’ aromas, and sweetness. ‘Crispy’, ‘crunchy’ and ‘hardness’ attributes were significantly rated higher in the presence of glucose in the medium, while the effect of pH was minor. The microstructure was further probed with confocal microscopy and X-ray tomography. The degree of alveolation was similar in differently processed macerated peanuts, even though sensory attributes were significantly different. Quantitative data on alveolation showed that microstructure disruption through steam generation cannot explain all the texture differences among processed peanuts. Correlations between sensory and instrumental attributes were also explored using Partial Least Square