Figure 3.

k-means cluster analysis of T. delbrueckii strains’ characterization. (A) Features (phenotypic tests and metabolites quantified by HPLC) mostly contributing for the division of strains into two clusters, in terms of information gain and information gain ration; (B) Number of strains categorized in the two clusters defined with k-means algorithm, colored according with their technological application or origin: ●—winemaking; ●—arboreal/soil; ●—food; ●—bread; ●—water; ●—clinical; ●—other beverages; ●—unknown origin.