Spectra Prediction Tools

The verification module uses spectrum prediction tools to reduce the solution space. A reduction of up to 99% is achievable, if all spectroscopic methods are used, because each spectroscopic method has specific properties to describe structural features. 

Spectroscopic databases are a very valuable tool for the identification of known and unknown substances. In most spectroscopic laboratories they are available and frequently used. Retrieval of data and spectra similarity searches are established tools for the fast identification of unknown compounds. The spectroscopic information stored in the databases offers the generation of structure-spectra correlations, which can be used for predicting spectral features of new compounds. Effective spectrum prediction tools are available for C NMR and H NMR, and will become available for IR spectroscopy in the near future. The prediction of mass spectra is still a challenge. 

Many of the physicochemical properties of interest are dependent on the solid form and, unfortunately, successful prediction of polymorphic forms is inexact. This, in combination with the fact that prediction of physicochemical properties is also very challenging, makes ah initio prediction very difficult and imprecise. However, some discussion of predictive tools is included in this chapter. A general comment regarding ah initio prediction is that "order of magnitude" predictions may be possible once some basic physicochemical information is available. However, the complexity and diversity of the chemistry space make reliable predictions across a broad spectrum of chemical structures very difficult. It is not surprising then that physicochemical predictions across more narrowly defined chemical spaces (e.g., chemical or therapeutic classes) can be more reliable and useful. Drug delivery, formulation, and computational chemistry experts will likely be able to provide a useful perspective on opportunities to take advantage of such ah initio predictions within the chemistry space that discovery teams often operate. 

The major strength of HF is that it is relatively economical, but across the spectrum of chemically interesting problems its ability as a predictive tool is only qualitative at best. We would like to be able to retain the inexpense of HF while enjoying the quality of the more sophisticated correlated methods. Of course, this seems like asking for too much, as it would appear to contradict the sage wisdom "You get what you pay for". This may or may not be true, but in any case, it is clear that if the objective is to have a quantitatively accurate, cost-effective and widely applicable quantum mechanical method for predicting molecular energies, a radical departure from traditional correlation techniques will be necessary. 

In spectrum prediction the system receives a molecular structure that might be coded in different ways and provides the corresponding IR spectmm as output. Spectrum prediction can be a useful tool in stmcture elucidation (see Structure Determination by Computer-based Spectrum Interpretation). Even if no reference spectrum is available from a spectrum... 

A significant number of the spectra in present data collections have been abstracted From the literature. This procedure is acceptable for data from - C (or other heteronuclei) NMR, where the chemical shift information plays the central role. Reduced information of that kind can even be used as a basis for spectrum prediction algorithms, as described later. In other techniques such as IR spectroscopy and H NMR, lineshape and peak patterns play an important role and should be stored in a database, not only for enhanced search capabilities but also as a basis for prediction tools. In MS a similar situation occurs. Peaks with low intensity may contain significant structure information, therefore all peaks above a certain intensity level should be stored. Since we talk about up to 64k data-points for H NMR and around 4k for MS and IR, a manual excerption is impossible. 

EPIOS and SpecSolv include tools for spectrum interpretation, structure generation, and spectrum prediction. The latter procedure is used twice prospectively during structure generation to limit the search space, and retrospectively to rank the set of complete structures generated based on the fit between predicted and observed spectral data. 

We have already met one tool that can be used to investigate the links that exist among data items. When the features of a pattern, such as the infrared absorption spectrum of a sample, and information about the class to which it belongs, such as the presence in the molecule of a particular functional group, are known, feedforward neural networks can create a computational model that allows the class to be predicted from the spectrum. These networks might be effective tools to predict suitable protective glove material from a knowledge of molecular structure, but they cannot be used if the classes to which samples in the database are unknown because, in that case, a conventional neural network cannot be trained. 

Very powerful tools for the study of dienes and, to some extent, polyenes (in particular annular polyenes) are both H and 13 C NMR spectroscopies, which will be discussed in a separate section. As previously mentioned 1,3-butadiene is more stable in the s-trans conformation and in the H NMR spectrum both butadiene (1) and 2,3,6,7-tetramethyl-2,4,6-octatriene (3) display the vinyl proton at a low chemical shift value. In these simple examples the S value can be predicted theoretically. The 111 NMR spectrum of a C25-branched isoprenoid was examined as part of the structural determination for biomarkers and is shown in Figure l6. The other spectral and structure assignments are described later in this review. 

Since 1905, when Coblentz obtained the first IR spectrum, vibrational spectroscopy has become an important analytical research tool. This technique was then applied to the analysis of adsorbates on well-defined surfaces, subsequently moving towards heterogeneous reaction studies. Terenin and Kasparov (1940) made the first attempt to employ IR in adsorption studies using ammonia adsorbed on a silica aerogel containing dispersed iron. This led to a prediction by Eischens et al. from Beacon Laboratories in 1956 that the IR technique would prove to be extremely important in the study of adsorption and catalysis. For an excellent review article in IR spectroscopy, see Ryczkowski and references therein and for a more recent review with applications, see Topsoe. ... 

Amidst all the enthusiasm about this versatile new tool that quantum chemistry has put at the hands of practioners of IR spectroscopy in matrices, one should not forget its limitations. First, a valid prediction can only come from a calculation based on a correct structure. In the case of reactive intermediates, this is not always as evident as one might wish. A famous example is given in Chapter 16 in this volume Much of the recent discussion on the correct assignment of the IR spectrum of m-benzyne was caused by the fact that different theoretical methods predict different structures, with more or less bonding between the radical centers, for this species. The DFT methods appear to overestimate this bonding, and hence are unsuitable for the prediction of the IR spectrum of m-benzyne. 

Generally die NMR spectrum of a compound is used in conjunction witii otiier available information for identification purposes. The reactants and die reagents and reaction conditions can serve as a guide to die types of products diat might be expected. Structure identification often merely confirms die structures of products that were predicted from die chemistry employed in die syndiesis. In odier cases products are obtained whose spectta do not match die predicted products. In such cases more information is usually required to solve die structure. Thus while NMR is an extraordinarily powerful tool, it is not sufficient to solve all structural problems. This latter fact must be kept in mind. 

Naive approaches avoid theoretical assumptions and instead focus on statistics about solved RNA structures, using these to probabilistically align new sequences with solved structures. One elegant approach to this problem has used an rRNA database to generate a novel RNA-specific substitution matrix. The advantage of this approach is that it makes the whole spectrum of primary-structure sequence-analysis tools available for secondary-structure prediction (27). 

---