Formula-based structure generation

Molecular structure elucidation. Computer-aided structure elucidation (CASE) uses algorithms that construct all mathematically possible structural formulas for a given molecular formula and optional structural restrictions (often obtained from a spectrum). This has to be performed efficiently and without redundance (i.e. no duplicates allowed). Virtual spectra can be calculated for generated structures and compared with the experimental spectrum to rank the generated structure candidates. The corresponding algorithms that we need for such a formula-based structure generation will be described. 

The data structure used for molecular graphs depends on the purpose and on the problem to be solved. For example, if an efficient formula-based structure generation plays a central role, an optimal random access to the bonds is important, and so the matrix of multiplicities will be used. However, this method has rather high memory requirements. In other situations, e.g. a substructure search, fast sequential access will be favorable and only the neighborhood list is needed. A neighborhood list keeps a list of all adjacent atoms for each atom, up to three labels as well as the associated information about atoms and bonds. 

ASSEMBLE 2.0 is a structure generator taking molecular formula and fragments as input. On output, candidates can be ranked based on fragment spectra given on input. 

We first describe formula-based generation of molecular structures. This starts with a molecular formula and takes further restrictions into account, which often allow an enormous and necessary - reduction of the search space. Then we discuss the handling of restrictions, i.e. constrained generation. 

Figure 8.4 shows our workflow for structure elucidation via MS, following the plan, generate, test strategy used in DENDRAL (Section 8.2). The focus is on determination of the molecular formula and structure. For interpretation we use MS classifiers, which provide information on both element composition and structure (see Appendix B). We use classifiers described by K. Varmuza and W. Werther [324, 333, 335] and develop new classifiers based on different classification methods (Subsection 8.5.2) and new structural properties (Subsection 8.5.3). 

The ACF shortlist, an abbreviated form of which is shown in Figure 12, contains many more invalid than valid ACFs,. (Based on the molecular formula, the valid ACF set for this problem, which is ultimately produced by the structure generator COCOA, contains exactly 22 ACFs.) In the hands of an experienced chemist or spectroscopist so inclined, there may be opportunities for some simple editing of the PRUNE-generated ACF shortlist that can, at times, improve program efficiency and decrease the number of plausible candidates generated. 

Bogner, F. K., Fox, F. L. and Scliinit, L. A., 1965. The generation of interelemenl-coinpatible stiffness and mass matrices by the use of interpolation formulae. Proc. Conf. on Matrix Methods in Structural Mechanics, Air Force Institute of Technology, Wright-Patterson AF Base, OH. 

Two-Dimensional Representation of Chemical Structures. The lUPAC standardization of organic nomenclature allows automatic translation of a chemical s name into its chemical stmcture, or, conversely, the naming of a compound based on its stmcture. The chemical formula for a compound can be translated into its stmcture once a set of semantic rules for representation are estabUshed (26). The semantic rules and their appHcation have been described (27,28). The inverse problem, generating correct names from chemical stmctures, has been addressed (28) and explored for the specific case of naming condensed benzenoid hydrocarbons (29,30). 

---