Applications of Raman Spectroscopy to Virology and Microbial Analysis - Emerging Raman Applications and Techniques in Biomedical and Pharmaceutical Fields - page 441

of bacteria and other microorganisms, e.g., archaea, exhibiting dimensions in
the micrometer range are favorable targets for micro-Raman spectroscopy.
The obtained Raman signatures reflect the overall composition of the cell, in
case that single prokaryotes exhibit a spatial homogeneity, which is generally
true [45].
However, it has to be kept in mind that only very small sample volumes

0 . 5-2 . 5

m 3 are analyzed [45], so a lower signal-to-noise ratio and additional
signals due to the substrate might appear in contrast to the aforementioned
bulk measurements.
Like already mentioned for the bulk studies presented above different cul-
turing environments lead to changes in the biochemical composition of a mi-
crobial cell and could therefore affect the ability to discriminate and identify
the investigated species [43]. It is also a striking feature of Raman spectra of
single cells that the spectra are strongly affected by the cells' individual phys-
iological states [44, 59]. This is certainly also true for bulk Raman spectra,
but to a lesser extent since the spectra are an average over several bacterial
cells [41, 43, 57].
This results in a dilemma: which information in Raman spectra are rep-
resentative for the whole ensemble of the bacterial community and which
emerge from the specific individualities of the single cell? This information
has to be extracted out of the vast amount of multidimensional data obtained
by measuring a number of single cells making a chemometrical analysis of the
data necessary. In doing so, only a handful of multivariate methods, unsuper-
vised and supervised, are applied in every work dealing with chemotaxonomic
identification of single cells by micro-Raman spectroscopy. Frequently, a data
reduction step is performed using the principal component analysis (PCA),
which is a well-known method for reducing the dimensionality in a data set.
Also well established in this context is the unsupervised hierarchical cluster
analysis (HCA), mainly useful as preliminary analysis step. Finally as su-
pervised classification method the so-called support vector machine (SVM)
provided hugely satisfying results in combination with micro-Raman analysis.
However, like all classification methods an SVM model is only as good as the
database backing it up is comprehensive. A valid signature library has to be
customized to the objective: If, for example, Bacillus spores isolated out of
soil need to be identified the library must comprise spectra of several Bacillus
strains of different species, of genetic near neighbors, and also of typical soil
material.
The basis of phenotypic discrimination of closely related species via
Raman spectroscopy lies in its sensitivity to the intracellular molecular com-
ponents including extrachromosomally encoded phenotypes, such as the Bacil-
lus anthracis or B. thuringiensis toxins or polyglutamic acid capsules. Other
prominent examples are cell storage materials like the polyhydroxy butyric
acid (PHB), carotenoid-based pigments like sarcinaxanthin, hemoproteins like
cytochrome or calcium dipicolinate (CaDPA). Raman spectra of single bac-
teria, in which the latter four intracellular substances occur, are shown in

μ