K. Chruszcz-Lipska and E. W. Blanch
74
4.3.2 Multivariate Analysis of ROA Spectra
4.3.2.1 Two-dimensional Correlation Analysis
the RoA spectra of even simple biomolecules are complex and rich in detail. Although quantum chemical calculations are now opening many new insights into the
structural origins of RoA features, these are not routine and are generally limited to
small-to-medium sized molecules. In order to utilize the sensitivity of RoA to biomolecular structure to investigate conformational transitions, the informatics method of generalized two dimensional correlation analysis (2dCos) has proven to be a
particularly promising tool. 2dCos provides a cross-correlation analysis of the measured spectra as a function of any two independent wavenumber positions, which
spreads the spectra over an additional dimension. the generalized 2dCos method
developed by Noda [88, 89] has been manifested in a wide, and still growing, range
of representations, but the most common are 2d synchronous and asynchronous
contour maps which show visual similarities to 2d NmR correlation maps.
upon collection of a series of spectra from a sample subjected to varying degrees
of a particular perturbation, 2d correlation generates a synchronous plot, in which
relative similarities in behaviour exist between bands, and an asynchronous plot,
where overall differences between the behaviour of bands are shown. using simple
sign associations, 2d spectral correlation analysis can provide much improved visualization of spectral features, consequently simplifying the identification of dynamic spectral changes, as well as probing the sequential information contained
within the experimental data [90, 91]. 2d correlation also allows perturbation-induced data sets from two different spectroscopic techniques, such as corresponding
Raman and RoA spectra measured simultaneously on the same sample, to be compared through so-called heterospectral or heterocorrelation spectroscopy. this can
provide a source of new band assignments for RoA spectroscopists as the Raman
literature is still more extensive.
Ashton et al. presented the first report of 2dCos RoA, in a study of the transition from α-helical to β-sheet structure in the homopolypeptide, poly-L-lysine [92].
A well known aspect of 2dCos, being both a key advantage and a potential source
of error, is the enhanced sensitivity to small spectral changes, which necessitates
the exercise of care in data pretreatment prior to performance of the 2d correlation
analysis. Ashton et al. established an optimal protocol for pretreating RoA spectra
in order to generate reliable synchronous and asynchronous correlation maps [92,
93]. the authors also found that the pseudoscalar nature of RoA spectra leads to
an additional, but solvable, complication in the analysis of the contour plots. the
general rules for interpreting the signs of 2d contour plots were based on monosignate spectra, e.g. Raman spectra, but the bisignate nature of RoA spectra needs to
be taken into account in order to correctly identify the direction of intensity change.
Recently, Pazderka and Kopecky have combined 2dCos with principal components
analysis (PCA) in order to reduce noise and baseline distortions prior to generation
of the 2d contour maps [94].
74
4.3.2 Multivariate Analysis of ROA Spectra
4.3.2.1 Two-dimensional Correlation Analysis
the RoA spectra of even simple biomolecules are complex and rich in detail. Although quantum chemical calculations are now opening many new insights into the
structural origins of RoA features, these are not routine and are generally limited to
small-to-medium sized molecules. In order to utilize the sensitivity of RoA to biomolecular structure to investigate conformational transitions, the informatics method of generalized two dimensional correlation analysis (2dCos) has proven to be a
particularly promising tool. 2dCos provides a cross-correlation analysis of the measured spectra as a function of any two independent wavenumber positions, which
spreads the spectra over an additional dimension. the generalized 2dCos method
developed by Noda [88, 89] has been manifested in a wide, and still growing, range
of representations, but the most common are 2d synchronous and asynchronous
contour maps which show visual similarities to 2d NmR correlation maps.
upon collection of a series of spectra from a sample subjected to varying degrees
of a particular perturbation, 2d correlation generates a synchronous plot, in which
relative similarities in behaviour exist between bands, and an asynchronous plot,
where overall differences between the behaviour of bands are shown. using simple
sign associations, 2d spectral correlation analysis can provide much improved visualization of spectral features, consequently simplifying the identification of dynamic spectral changes, as well as probing the sequential information contained
within the experimental data [90, 91]. 2d correlation also allows perturbation-induced data sets from two different spectroscopic techniques, such as corresponding
Raman and RoA spectra measured simultaneously on the same sample, to be compared through so-called heterospectral or heterocorrelation spectroscopy. this can
provide a source of new band assignments for RoA spectroscopists as the Raman
literature is still more extensive.
Ashton et al. presented the first report of 2dCos RoA, in a study of the transition from α-helical to β-sheet structure in the homopolypeptide, poly-L-lysine [92].
A well known aspect of 2dCos, being both a key advantage and a potential source
of error, is the enhanced sensitivity to small spectral changes, which necessitates
the exercise of care in data pretreatment prior to performance of the 2d correlation
analysis. Ashton et al. established an optimal protocol for pretreating RoA spectra
in order to generate reliable synchronous and asynchronous correlation maps [92,
93]. the authors also found that the pseudoscalar nature of RoA spectra leads to
an additional, but solvable, complication in the analysis of the contour plots. the
general rules for interpreting the signs of 2d contour plots were based on monosignate spectra, e.g. Raman spectra, but the bisignate nature of RoA spectra needs to
be taken into account in order to correctly identify the direction of intensity change.
Recently, Pazderka and Kopecky have combined 2dCos with principal components
analysis (PCA) in order to reduce noise and baseline distortions prior to generation
of the 2d contour maps [94].