Raman Optical Activity of Biological Molecules - Emerging Raman Applications and Techniques in Biomedical and Pharmaceutical Fields - page 168

7.9.1 Principal Component Analysis (PCA)

Pattern recognition methods have been widely used to better understand
many types of vibrational spectra of biological systems and biomolecules,
and principal component analysis (PCA) [72] was first implemented for ROA
in order to conduct an initial representation of the structural relationships
among polypeptide and protein states based on their ROA spectra [35]. In
this example, the set of 78 ROA spectra for proteins and polypeptides were
first normalized by scaling to the sum of the squared spectral intensities.
Figure 7.10 shows a plot of the two most significant basis functions, which to-
gether account for 50% of the total variance of the data, with the positions

of samples being colour-coded with respect to the seven different structural
types listed in the figure, which provide a useful initial classification of tertiary
structure. Since

∼

α -helix and β -sheet contents of proteins tend to be inversely

Fig. 7.10. Plot of the PCA coecients for the two most important basis functions
for a set of 78 polypeptide, protein and virus ROA spectra. Definitions of the struc-
tural types analysed are: all alpha, > ∼ 60% α-helix with little other secondary struc-
ture; mainly alpha, > ∼ 35% α-helix and a small amount of β-sheet ( ∼ 5-15%); alpha
beta, similar significant amounts of α-helix and β-sheet; mainly beta, > ∼ 35% β-
sheet and a small amount of α-helix ( ∼ 5-15%); all beta, > ∼ 45% β-sheet with little
other secondary structure; mainly disordered/irregular, little secondary structure;
all disordered/irregular, no secondary structure