Clustering of Red/White Wine and Allergen/Non-Allergen Data Sets by Using Descriptor Fingerprints

Abstract

An application of the descriptor fingerprints in relation to a clustering procedure of red and white wines and food allergen/non-allergens has been studied. The clustering method is based on a procedure using the Tanimoto similarity index. Different threshold values have been employed in order a best set of clusters to be obtained. It was shown that this procedure produces quite good discrimination between the red and white wines, but any correlation with the human sensation tastes (represented by a scale from 1 to 10) is more difficult to be obtained. The results from 4212 food allergens/non-allergens show a much better discrimination of those two groups starting from threshold value=0.5.