75
the ability of 2dCos RoA to reveal new structural information was shown by
Ashton et al. in their study on α-helix unfolding in the model homopolypeptide
poly-L-glutamic acid [95], which distinguished between fraying of the ends of helices and unfolding of the core helical structure.
4.3.2.2 Data Clustering Techniques
the analysis of protein structure remains a great challenge for the life sciences in the
post-genomic era. Consequently, a large number of RoA experimental researches
is dedicated to proteins and their building blocks, amino acids and peptides. Protein
RoA spectra are dominated by bands from the peptide backbone and can give direct
information about secondary and tertiary structure, whereas the bands from side
chains are usually weaker because of some degree of conformational freedom. As a
result, a lot of individual RoA signals and characteristic band patterns that are seen
in protein spectra have been assigned to elements of secondary and tertiary structure
such as α-helix and β-sheet, loops or turns. Thus, the ROA spectra of proteins with
a large number of structure-sensitive bands are suitable for the application of pattern recognition methods to obtain useful information about the structure. valuable
structural information has been successfully found by studying protein RoA spectra
using the method of principal component analysis (PCA) [96–100]. thanks to this
method the structural information from protein RoA spectra of unknown structure
can be extracted automatically by their location into clusters corresponding to different protein folds. more advanced multivariate analysis using non-linear mapping
(NLm) was found to give even better results [31, 101, 102]. A two-dimensional
NLm plot for a set of 80 RoA spectra (both ICP and SCP) of polypeptides, proteins
and viruses in aqueous solution measured in the range of 702–1773 cm
−1
shows
excellent clustering corresponding to the following seven structural categories: all
α, all β, mainly α, mainly β, αβ, mainly disordered/irregular and all disordered/
irregular [101]. Another two-dimensional NLm plot for a set of 85 RoA spectra
revealed the significant differences between the structural characteristics of natively
unfolded proteins and proteins unfolded by denaturation [102].
In conclusion, so far, applications of cluster analysis towards providing a better
understanding of RoA spectra were used successfully for proteins, and showed that
structural information can be easily and automatically extracted without any expert
knowledge of characteristic band assignments. however, cluster analysis can be
also used to study other class of biological molecules, as we have shown in our
recent work where we used hierarchical cluster analysis to determine the content of
chiral components in pichtae essential oil samples [103].
It is worth mentioning the application of another chemometric method, namely
partial least-squares (PLS). these algorithms were also found to yield fine results
for predictions of structural relationships among proteins from Raman as well as
RoA spectra [104]. It was shown that the optimized PLS algorithm gives, for whole
RoA spectra measured for 44 proteins, highly accurate secondary structure contents
with correlation coefficients of 0.96–0.98 and RSmd values from 2.5 to 2.9 %. this
4 Raman optical Activity of Biological Samples
the ability of 2dCos RoA to reveal new structural information was shown by
Ashton et al. in their study on α-helix unfolding in the model homopolypeptide
poly-L-glutamic acid [95], which distinguished between fraying of the ends of helices and unfolding of the core helical structure.
4.3.2.2 Data Clustering Techniques
the analysis of protein structure remains a great challenge for the life sciences in the
post-genomic era. Consequently, a large number of RoA experimental researches
is dedicated to proteins and their building blocks, amino acids and peptides. Protein
RoA spectra are dominated by bands from the peptide backbone and can give direct
information about secondary and tertiary structure, whereas the bands from side
chains are usually weaker because of some degree of conformational freedom. As a
result, a lot of individual RoA signals and characteristic band patterns that are seen
in protein spectra have been assigned to elements of secondary and tertiary structure
such as α-helix and β-sheet, loops or turns. Thus, the ROA spectra of proteins with
a large number of structure-sensitive bands are suitable for the application of pattern recognition methods to obtain useful information about the structure. valuable
structural information has been successfully found by studying protein RoA spectra
using the method of principal component analysis (PCA) [96–100]. thanks to this
method the structural information from protein RoA spectra of unknown structure
can be extracted automatically by their location into clusters corresponding to different protein folds. more advanced multivariate analysis using non-linear mapping
(NLm) was found to give even better results [31, 101, 102]. A two-dimensional
NLm plot for a set of 80 RoA spectra (both ICP and SCP) of polypeptides, proteins
and viruses in aqueous solution measured in the range of 702–1773 cm
−1
shows
excellent clustering corresponding to the following seven structural categories: all
α, all β, mainly α, mainly β, αβ, mainly disordered/irregular and all disordered/
irregular [101]. Another two-dimensional NLm plot for a set of 85 RoA spectra
revealed the significant differences between the structural characteristics of natively
unfolded proteins and proteins unfolded by denaturation [102].
In conclusion, so far, applications of cluster analysis towards providing a better
understanding of RoA spectra were used successfully for proteins, and showed that
structural information can be easily and automatically extracted without any expert
knowledge of characteristic band assignments. however, cluster analysis can be
also used to study other class of biological molecules, as we have shown in our
recent work where we used hierarchical cluster analysis to determine the content of
chiral components in pichtae essential oil samples [103].
It is worth mentioning the application of another chemometric method, namely
partial least-squares (PLS). these algorithms were also found to yield fine results
for predictions of structural relationships among proteins from Raman as well as
RoA spectra [104]. It was shown that the optimized PLS algorithm gives, for whole
RoA spectra measured for 44 proteins, highly accurate secondary structure contents
with correlation coefficients of 0.96–0.98 and RSmd values from 2.5 to 2.9 %. this
4 Raman optical Activity of Biological Samples