Spectroscopic Fingerprint of Pale Beers linked to Sensory Analysis and Consumers Preferences

During the last decade the awareness of consumers and society in general towards all aspects that concern food consumption has strongly increased. Ethics, sustainability, health, safety, quality, tradition are now everyday words, and communication and marketing are following more and more the trends that these terms represent. Research in food chemistry area has mainly focused on chemical analysis and characterization to contribute to fundamental issues such as food safety and quality, nutritional and health requirements. The present work is part of a larger project, which is aimed to take a step beyond the aforementioned approach. The fundamental idea is to use our analytical chemistry expertise to build new tools to aid consumers when choosing foodstuff (and have proper knowledge of it) and producers to meet consumer expectations, using food quality as a driver. To this aim the proposal is to use analytical spectroscopy to capture salient features of foodstuff (fingerprint) and build a reference database that can be efficiently searched through multivariate data analysis tools and, be likely in the near future, linked to applications for mobile smart devices implemented for consumers inquires. At the same time consumers’ choice may be oriented by showing how products of similar categories cluster according to different criteria. As a first benchmark to develop these ideas a survey on beer is presented. One hundred samples of light beer (i.e. no stout or dark beers have been considered) differing by brewery, alcohol content, yeast, brew type, etc., were collected. This work is focused on Vis-NIR, NMR and sensory data, with the aim of establishing a link between the “objective” information of the spectroscopic fingerprint and the “subjective” world of consumers’ assessments. The latter is represented by online beer ratings and for a reduced number of samples by sensory reports made by a panel of experts. Different combinations of decomposition methods, e.g. PCA, ICA, MCR and clustering (both linear and non-linear methods) were used to extract relevant information and as basis for data-fusion techniques to integrate the chemical information. PLS regression allowed establishing a link between spectral fingerprint/information and consumer preferences as expressed by ratings (www.ratebeer.com) aroma, appearance, taste, palate, overall. Furthermore, also the mostly used descriptive words used by consumer were codified, based on the approach described in [1]. This part of the work can contribute to build a “beer vocabulary”, which can be very useful to develop further real-life applications. References: [1] M. Bevilacqua, Visualization of comprehensive data mining problems in gastronomy, chap. 10, Development of chemometric approaches for ensuring food quality and safety, Doctoral Thesis, Roma La Sapienza (2014