Although concentration changes in the target compounds were
seemingly random in nature, close examination by chemical families
revealed that the concentration reach the higher, except for
alcohols, after the second or third batch fermentation. This trend
was examined by cluster analysis according to Ward's method,
using the contents in volatiles and polyols as classifying variables.
The term cluster analysis encompasses a number of different algorithms
and methods for grouping samples of similar kind into
respective categories. The smaller the distance between clusters is,
the greater is their similarity. In our study, cluster analysiswas used
to study the similarity among the wines obtained in the different
batch fermentation. As can be seen from Fig. 3, the wines obtained
by partial fermentation differ markedly from the traditional Pedro
Ximenez wine (unfermented wine, T). Also, the cluster analysis
distinguished two groups: one encompassing the wines obtained
with free yeast cells (F) and those provided by the first run with
yeast biocapsules (B1); and the other comprising the wines obtained
in the successive reuses. This may have been a result of
cellular growth for F and B1. Then, when cellular growth is less
important a gradual adaptation of the yeasts to the fermentation
conditions in the successive fermentation runs may be reached
(B2eB5).