Spectral matching approaches

To bypass the problems of conventional database search methods, a number of strategies for unrestricted PTM identification have been developed, such as multistep database search, sequence tag by de novo sequencing, and spectral matching approaches (reviewed in Table 2). 

In addition to the limitations of the direct spectral matching approach discussed in Section 15.3.2, there is another, more subtle difficulty that arises when full spectral scans are used for automatic computerized identification of unknown materials. This limitation applies to both of the spectral methods used. To examine the cause of this, let us imagine that we have collected two spectra of the same material, and let us imagine further that we have properly corrected for the effect of particle size shifts on the data. In this case, the two spectra should be essentially identical. However, they will not be actually identical because of the effect of instrumental noise. In this case the effect of noise on the data is easily calculated, and the two spectra will be separated by a multidimensional Euclidean distance equal to an, where a is the noise level of the instrument. 

DEVELOPMENT OF SPECTRAL PATTERN-MATCHING APPROACHES TO MATRIX-ASSISTED LASER DESORPTION/ IONIZATION TIME-OF-FLIGHT MASS SPECTROMETRY FOR BACTERIAL IDENTIFICATION... 

Following are descriptions of the most common mathematical approaches used through spectral matching software. 

One mechanism that is proposed to ameliorate this problem is to first identify the product being tested before quantifying the analyte of interest. Using a well-established algorithm for spectral matching, such as principal component analysis followed by an appropriate qualitative pattern recognition routine like the Mahalanobis distance calculation, would be a viable approach. 

Peak identification is based on the comparison of normalized spectra representative for the peak with spectra of one or several standard compounds run in the same separation system and stored in a spectral library [107,116]. This approach is less powerful than for mass or infrared spectral searches due to the rather broad and featureless bands that typify absorption spectra. Absorption spectra of similar compounds and compounds with a chromophore well separated from the variation in molecular structure are often virtually identical. Also, spectral changes dependent on the experimental conditions (pH, mobile phase composition, temperature, etc.) occur frequently. For this reason user prepared local libraries tend to predominate over general libraries, in contrast to common practices in infrared and mass spectral searches. A favorable spectral match for an absorption spectrum by itself is not acceptable for absolute identification. 

The second approach is one of correlation. A spectral match is calculated by comparing the slope generated from comparative data points on the sample and reference spectra. The cosine between the slopes is calculated and a spectral match index or correlation determined (Fig. 9.23). 

One new approach is to match the magnetic susceptibility of the materials used in probe construction to a non-zero value. Olson et al. have obtained high-resolution NMR spectra using this susceptibility matching approach. A microcoil with an active volume of 5 nL was immersed into a fluid with a magnetic susceptibility similar to that of the copper coil, and thus compensated for the susceptibility mismatch at the air/copper interface. The spectral line width achieved was under 1 Hz, and the LODs were in the 10 to lOOpmol range. 

Technically valid calibrations transfer is not the trivial process that some would propose. In fact, due to advancing calibration mathematics such as PCR, PLS, and spectral matching/search algorithms, it becomes even more critical that transfer technologies be scientifically scrutinized. To date, the most successful approach for transferring calibrations for use with all multivariate mathematical modeling techniques is found in a three-pronged approach ... 

This approach to spectral matching was initially developed for use in the UV region of the spectrum [9] and in the mid-IR [10,11] where its use as a matching criterion was applied to the interferograms produced by Fourier Transform (FT) IR spectrometers [12,13] as well as to spectra. Use in the NIR region quickly followed [14,15],... 

Another advantage of the Mahalanobis distance approach is the fact that at this time it is much better understood theoretically. This gives rise to two characteristics first, it allows theoretical bounds to be specified as the limits for classifying samples and for detecting samples not part of the training set. For both spectral match methods in current use, limiting bounds must be determined empirically. 

---