Applications of Raman and Surface-Enhanced Raman Scattering to the Analysis of Eukaryotic Samples - Emerging Raman Applications and Techniques in Biomedical and Pharmaceutical Fields - page 78

evaluation of overall variance between different pollens and to obtain an un-
biased perspective on possible chemical differences, we aim at a classification
method that does not require pre-selection of chemical parameters or weighing
of input information. As a consequence, we chose to use hierarchical cluster
analysis (HCA) as an unsupervised method.
Prior to analysis, the Raman shift axes of the spectra were calibrated
using the Raman spectrum of 4-acetamidophenol. Pretreatment of the raw
spectra, such as vector normalization and calculation of derivatives were done
using Matlab (The Mathworks, Inc.) or OPUS (Bruker) software. OPUS NT
software (Bruker, Ettlingen, Germany) was used to perform the HCA. The
first derivatives of the spectra were used over the range from 380 cm − 1 to
1700 cm − 1 . To calculate the distance matrix, Euclidean distances were used
and for clustering, Ward's algorithm was applied [59].
We applied HCA to a spectral data set containing 91 spectra from differ-

ent samples of 15 tree species using almost the full spectrum 380-1700 cm − 1
(Fig. 4.3). The descriptions of the samples are listed in Table 4.2. The result-
ing dendrogram reveals formation of three large clusters (see Fig. 4.3). Except
for the case of the different oak species and < 4% other misclassifications, the
spectra of the different samples of each respective species form individual
groups within these larger clusters (Fig. 4.3). This result is in accordance
with the assumption that the variation between different specimens of a par-
ticular species is smaller than the variation between species and permits us to
conclude that most of the pollen samples of the different species studied here
can be distinguished based on their Raman spectra. The average spectra of
each of the species were displayed in Fig. 4.1.
As evidenced by the dendrogram in Fig. 4.3, separation down to species
level is superimposed by a high similarity within genera. Examples for a high
chemical similarity within the same genus are the spectra obtained from two
species of Prunus , Bird Cherry and Wild Cherry. They form a distinct group,
as do also two Acer species, Norway Maple and Sycamore Maple (see Fig. 4.3).
However, the pollen spectra from a third species of Acer , the Ash-leaf Maple
(or Maple Ash), resemble those of oak and sycamore tree pollen rather than
the other Acer species. The tree is native to North America, different from
the other two Acer species, which are natives to Europe and Asia.
In good agreement with the order of the systematic taxa, the next level
of similarity revealed by HCA is that of plant families, as illustrated by the
examples of the Birch and of the Soapberry families (Betulaceae and Sapin-
daceae). Silver birch (genus Betula ) and Hophornbeam (genus Ostrya )are
members of the Birch family (Betulaceae), and their spectra form one of the
three main data clusters. Sycamore Maple, Norway Maple and Horse-chestnut
are all members of the Soapberry family (Sapindaceae), and are also contained
in one group.
When classifying biological samples by analytical methods, usually the
question arises about individual variation vs. variation between groups of in-
dividuals. To take into account such variation between individuals, each of