8.0 7.5 7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 d1H
FIGURE 5.2 600 MHz1H NMR urine spectrum of control urine with expansions of selected spectral regions.
the interpretation of these complex modulations in spectral profiles, and can enable the classification of samples according to their metabolic status based on similarity of biochemical composition. Pattern recognition (PR) and related chemometric approaches can be used to discern significant patterns in complex data sets [27-30], and are particularly appropriate in situations where there are more variables than samples in the data set, such as is the case with spectral data. The general aim of PR is to classify objects (in this case 1H NMR spectra of biofluid or tissue samples) or to predict the origin of objects (the class of toxicity or disease in the current example), based on identification of inherent patterns in a set of indirect measurements. Moreover, PR can be used to reduce the dimensionality of complex data sets via two-dimensional 2D or 3D mapping procedures, thereby facilitating the visualization of inherent patterns in the data set. Both the theory and the application of the basic mathematical models used in PR have been well documented [27-30]. Typically, 1D biofluid spectra are digitized into a series of spectral descriptors that are scaled to provide input variables to map samples and construct mathematical models for specific types of physiological or pathological condition (Figure 5.3). Metabonomic technology has proved a powerful toxicological
M-interval spectra Data reduction M-dimensional space Classification
FIGURE 5.3 Schematic diagram of the sequence of spectral data reduction and analysis.
Was this article helpful?