Spectral similarity search

Spectral similarity search is a routine method for identification of compounds, and is similar to fc-NN classification. For molecular spectra (IR, MS, NMR), more complicated, problem-specific similarity measures are used than criteria based on the Euclidean distance (Davies 2003 Robien 2003 Thiele and Salzer 2003). If the unknown is contained in the used data base (spectral library), identification is often possible for compounds not present in the data base, k-NN classification may give hints to which compound classes the unknown belongs. 

A great variety of different methods for multivariate classification (pattern recognition) is available (Table 5.6). The conceptually most simply one is fc-NN classification (Section 5.3.3), which is solely based on the fundamental hypothesis of multivariate data analysis, that the distance between objects is related to the similarity of the objects. fc-NN does not assume any model of the object groups, is nonlinear, applicable to multicategory classification, and mathematically very simple furthermore, the method is very similar to spectral similarity search. On the other hand, an example for a rather sophisticated classification method is the SVM (Section 5.6). 

There has been little interest in the MCSS method for the interpretation of spectra of organic compounds until recently. An early work on the application of MCSS approach for unknown mass spectra was reported by Cone et al. [56] in 1977. Chen and Robien developed a novel approach for the automatic deduction of common structural features from a set of structures obtained using 13C-NMR spectral similarity search [72] in 1994. They demonstrated that the detected MCSSs often show the main structural features of the unknown compounds under investigation. This method has recently been adopted by Varmuza et al. [73] for the automatic extraction of common structural features from the hitlist structures obtained using infrared (IR) spectral similarity searches. 

MS spectral similarity searches play a significant role in structure elucidation. Assuming that the spectrum of the query structure is not contained in the library, the only way to obtain information about the structure is to investigate the hitlist for common structural features. The SISCOM algorithm in MassLib has been optimized for that task (Figure 3). 

MassLib is not only applicable to spectrum similarity searches. The system contains an MS-specific coding technique for chemical structures based on 160 structure descriptors. This enables the spectroscopist to apply chemometric techniques to analyse the results of spectral similarity searches in the structure space. 

It is not trivial to define or select substructures that should be considered for this purpose. For the STIRS system, considerable effort has gone into the search for substructures that can be successfully classified by the implemented spectral similarity search. The Mass-Lib system uses a predefined set of 180 binary molecular descriptors to characterize the similarity of structures. In most investigations a more or less arbitrary set of substructures, functional groups or more general structural properties (compound classes) has been considered. Self-adapting methods that automatically analyse the molecular structures in the hitlist (for instance by searching for frequent and large substructures) have not been used up to now in MS. 

Multivariate data analysis usually starts with generating a set of spectra and the corresponding chemical structures as a result of a spectrum similarity search in a spectrum database. The peak data are transformed into a set of spectral features and the chemical structures are encoded into molecular descriptors [80]. A spectral feature is a property that can be automatically computed from a mass spectrum. Typical spectral features are the peak intensity at a particular mass/charge value, or logarithmic intensity ratios. The goal of transformation of peak data into spectral features is to obtain descriptors of spectral properties that are more suitable than the original peak list data. 

The importance of an appropriate transformation of mass spectra has also been shown for relationships between the similarity of spectra and the corresponding chemical structures. If a spectra similarity search in a spectral library is performed with spectral features (instead of the original peak intensities), the first hits (the reference spectra that are most similar to the spectrum of a query compound) have chemical structures that are highly similar to the query structure (Demuth et al. 2004). Thus, spectral library search for query compounds—not present in the database—can produce useful structure information if compounds with similar structures are present. 

The peak identification for the chromatogram shown in Figure 6.1.25 was done using MS spectral library searches. This identification is not always possible, since most compounds with a higher MW are not found in the commercial mass spectral libraries (such as NIST 2002, Wiley 7. etc.). The similarity between the spectra in each series of compounds can be used for peak identification, even when the compound is not found in the mass spectral library. This is exemplified in Figure 6.1.26, which shows the spectra of the B series of compounds shown in Table 6.1.10. 

An example of a pyrogram of a polystyrene sample (M = 280,000) is shown in Figure 6.2.2. The pyrolysis was done at 600° C in He with separation of a Carbowax column and MS detection, similarly to other polymers discussed in this book (see also Table 3.4.1). The peak identification for the chromatogram shown in Figure 6.2.2 was done using MS spectral library searches only and is given in Table 6.2.2. 

The pyrolysis of poly(butyl acrylate) takes place similarly to the pyrolysis of poly(methyl acrylate). The results for a Py-GC/MS analysis of a 0.4 mg sample of poly(butyl acrylate) with M = 60,000 are shown in Figure 6.7.8. The pyrolysis and pyrolysate separation were done in the same conditions as those for other examples previously discussed (see Table 4.2.2). The MS was operated in EI+ mode and peak identification was obtained using MS spectral library searches only. The results are given in Table 6.7.6. 

The pyrogram for the sample of poly(/so-butyl methacrylate) with Mw = 300,000 Is shown in Figure 6.7.34. The pyrolysis was done in similar conditions to other examples, namely at 600° C in He at a heating rate of 20° C/ms with the separation on a Carbowax column (see Table 4.2.2) and MS detection The peak identification for the chromatogram was done using MS spectral library searches and is given in Table 6.7.23. 

A rather similar pyrogram is obtained and a similar path is followed during the pyrolysis of poly(isobutylene-a/f-maleic anhydride), CAS 26426-80-2. A sample of this polymer with M = 60,000 was pyrolyzed at 600 C in He in similar conditions as for poly(ethylene-a/f-maleic anhydride). The pyrogram is shown in Figure 6.9.5, and the peak identification obtained by MS spectral library searches only is given in Table 6.9.4. 

Another homopolymer analyzed by Py-GC/MS and discussed below Is polyisoprene (c/s), with the formula [-CH2CH=C(CH3)CH2-]n and CAS 104389-31-3. The analyzed sample (0.4 mg) had M = 38,000 (synthetic material), and the experimental conditions were similar to those for other polymers previously discussed (see Table 4.2.2). The resulting pyrogram is given in Figure 7.1.3, and the peak identification that was done using MS spectral library searches only is given in Table 7.1.4. 

Pyrolysis of poly(2,6-dimethyl-1,4-phenylene oxide) [-C6H2(CH3)20-]n, CAS 25134-01-4, is further discussed for a sample with M = 244,000. The experimental conditions used to generate the pyrogram shown in Figure 9.1.17 were similar to those for other previous examples (see Table 4.2.2). Peak identification for the pyrogram was done using mass spectral library searches only and is given in Table 9.1.10 and Table 9.1.11. 

---