Spectrum-substructure relationships

Data from the TOMS instrument are used in two different ways 1) to develop a library of spectrum/substructure correlations from studies of known compounds and 2) to use the developed correlations to determine the substructures and thence the overall structures of unknown compounds. The data base required for this process is a library of the spectral characteristics of many substructures, rather than a library of the spectra of all known compounds. In principle, millions of compounds could be identified using a library of only a few thousand spectrum/substructure relationships. 

The procedure for obtaining the spectrum/substructure relationships is as follows. For a selected known compound, a daughter spectrum is acquired for every mass value greater than 1 relative intensity that appears in the primary spectrum of that compound. These... 

From the daughter spectra of di-n-octylphthalate, we were able to determine two spectrum/substructure correlations the 149+ daughter spectrum to structure I in Figure 3 and the 105+ daughter spectrum to structure III in Figure 3. In order to obtain spectrum substructure relationships for the alkyl portions of the reference molecule di-n-octylphthalate, we would then match other portions of the complete MS/MS spectrum against those of compounds containing alkyl substructures. However, this portion of the reference library has not yet been developed. Thus, to complete the structure elucidation we have used standard methods of spectral interpretation (11). As will be shown, these methods can also lead to useful spectrum/substructure relationships. 

Although rather formal, this m ij) provides extremely valuable information about the 2D spectrum/substructure relationship within the set of generated extensions at a given level. 

Expert systems are based on spectral feature-substructure relationship rules that comprise the knowledge base for IR spectral analysis. This is the main difference from neural network techniques, where no prior knowledge about the structure-spectrum relationship is necessary because the network learns inductively from examples. For expert systems a knowledge base has to be established and transformed into a computer operable form. This expert knowledge is expressed in terms of substructure-subspectra relationships. 

Artificial neural networks do not require any information about the relationship between spectral features and corresponding substructures in advance. The lack of information about complex effects in a vibrational spectrum (e.g., skeletal and harmonic vibrations, combination bands) does not affect the quality of a prediction or simulation performed by a neural network. 

If T(A, S2) is the Boolean function which describes all the mutual relationships between the characteristic frequencies (characteristic intervals) and the structural elements in a given problem, and R(fi) is the Boolean function describing the IR spectrum, then the logical analysis of this situation lies in the joint processing of the functions T(A, Q) and R(J2), As a result, the function, f(A), that enumerates the probable sets of substructures which may be present in the molecule under investigation is established Symbolically, this is expressed as... 

---