241
8 Two-Dimensional Mid-Infrared Correlation Spectroscopy in Protein Research
tained for other types of systems [142–145] or for different spectroscopies [102],
multivariate curve resolution (mCR), mainly of pure concentration profiles, could
also contribute valuable information useful in the interpretation of infrared 2dCoS
results. It can be expected that the 2dCoS-mCR combination will find greater applications in the analysis of infrared 2dCoS studies of protein, as in [70]. Numerous
literature data from tauler’s group and our results have shown that mCR itself has
great potential in protein investigations [146, 147].
despite their complementarity, 2dCoS and PCA are totally independent computational methods. In PCA, as in many other chemometric methods, each spectrum
in the entire range is considered to be one integral object represented by scores and
loading vectors. In contrast, 2dCoS is not based on the integral approach, and the
spectral variations at each frequency contribute individually to the synchronous
and asynchronous values. this combination of 2dCoS with PCA or another chemometric method is highly recommended because it supports the reliability of obtained results and primarily aids in the interpretation of the sequence of spectral intensity changes, which are determined from the so-called Noda’s rules. moreover,
a plot of the loading values as a function of the original variables (frequencies)
helps to deconvolute the amide bands and assign the bands to successive perturbation steps. Comparative analysis of the characteristics of PCA and 2dCoS has been
performed for the near-IR spectra of hSA measured as a function of concentration
at room temperature [92]. the PCA model, which is based on the two component
provided matrix of two loadings plotted as a function of frequency, has been compared with the autopower spectrum and slice spectra extracted from the asynchronous spectrum at the frequency corresponding to the maximum asynchronicity.
A strong similarity has been observed between the autopower spectrum and the
first PCA loading vector. Asynchronous changes from the slice spectrum were well
matched with the second PCA loading vector, which is orthonormal to the first
vector. this analysis is supported by the distribution of the score values, which
shows the relationships between the samples and provides additional information
that could be very helpful in the interpretation of the sequence of spectral intensity
changes from the 2dCoS results. In most cases, if the two methods were combined, only the scores plot, which shows a possible cluster of samples in response
to the perturbation, has been analyzed. 2dCoS of spectra separated into clusters
has provided specific information that is strictly correlated with a well-defined
perturbation range [30, 57, 74, 76, 77, 83, 87, 98]. For 2dCoS of data covering the
entire perturbation, some information could be unnoticed, mainly information from
intermediate states with low population that occur in a narrow perturbation range.
PCA has also been employed at the data pretreatment level to remove noise from
spectra before 2dCoS [148].
An overview of the different modern statistical and numerical methods used in
the analysis of spectral data and, in particular, for the quantitative characterization of protein structural evolution is given in [149]. this survey provides valuable
guidelines for biospectroscopists who would like to use these powerful methods in
their studies. Such information can also be found in [150], which provides comprehensive information on procedures oriented toward improving the interpretation
8 Two-Dimensional Mid-Infrared Correlation Spectroscopy in Protein Research
tained for other types of systems [142–145] or for different spectroscopies [102],
multivariate curve resolution (mCR), mainly of pure concentration profiles, could
also contribute valuable information useful in the interpretation of infrared 2dCoS
results. It can be expected that the 2dCoS-mCR combination will find greater applications in the analysis of infrared 2dCoS studies of protein, as in [70]. Numerous
literature data from tauler’s group and our results have shown that mCR itself has
great potential in protein investigations [146, 147].
despite their complementarity, 2dCoS and PCA are totally independent computational methods. In PCA, as in many other chemometric methods, each spectrum
in the entire range is considered to be one integral object represented by scores and
loading vectors. In contrast, 2dCoS is not based on the integral approach, and the
spectral variations at each frequency contribute individually to the synchronous
and asynchronous values. this combination of 2dCoS with PCA or another chemometric method is highly recommended because it supports the reliability of obtained results and primarily aids in the interpretation of the sequence of spectral intensity changes, which are determined from the so-called Noda’s rules. moreover,
a plot of the loading values as a function of the original variables (frequencies)
helps to deconvolute the amide bands and assign the bands to successive perturbation steps. Comparative analysis of the characteristics of PCA and 2dCoS has been
performed for the near-IR spectra of hSA measured as a function of concentration
at room temperature [92]. the PCA model, which is based on the two component
provided matrix of two loadings plotted as a function of frequency, has been compared with the autopower spectrum and slice spectra extracted from the asynchronous spectrum at the frequency corresponding to the maximum asynchronicity.
A strong similarity has been observed between the autopower spectrum and the
first PCA loading vector. Asynchronous changes from the slice spectrum were well
matched with the second PCA loading vector, which is orthonormal to the first
vector. this analysis is supported by the distribution of the score values, which
shows the relationships between the samples and provides additional information
that could be very helpful in the interpretation of the sequence of spectral intensity
changes from the 2dCoS results. In most cases, if the two methods were combined, only the scores plot, which shows a possible cluster of samples in response
to the perturbation, has been analyzed. 2dCoS of spectra separated into clusters
has provided specific information that is strictly correlated with a well-defined
perturbation range [30, 57, 74, 76, 77, 83, 87, 98]. For 2dCoS of data covering the
entire perturbation, some information could be unnoticed, mainly information from
intermediate states with low population that occur in a narrow perturbation range.
PCA has also been employed at the data pretreatment level to remove noise from
spectra before 2dCoS [148].
An overview of the different modern statistical and numerical methods used in
the analysis of spectral data and, in particular, for the quantitative characterization of protein structural evolution is given in [149]. this survey provides valuable
guidelines for biospectroscopists who would like to use these powerful methods in
their studies. Such information can also be found in [150], which provides comprehensive information on procedures oriented toward improving the interpretation