Computer-asslsted structure elucidation by the DARCIEPIOS system Is descrlbed through practical examples. The search for structures compatible wlth the spectrum, the molecular formula of an unknown compound, and eventual supplementary information Is lnitlalized by dlscrlmlnating lnterpretatlon of these data. Substructures, fully assigned with multlresonance subspectra, are Inferred from the EPIOS knowledge base (EKB). This Interpretatton ellmlnates most of the substructures usually selected by classlcai slngle-resonance methods. Both structural and spectral overlaps between these asslgned substructures lead to a generation graph that contains the candMate structures and their asslgned spectra as subgraphs. The system dynamkally controls the graph search, maklng It heuristlc, by synchronlzlng structure generation with assignment of the query spectrum. Only vaild spectrum/structure solutions are produced by this search.Computer-assisted structure elucidation of unknown organic compounds from their spectroscopic properties is an old but still current project ( I ) . More and more, the analytical potential of 13C nuclear magnetic resonance (13C NMR) and the increasing mass of experimental data available provide a suitable support for the development of computer-based systems (2). The ' 9 N M R spectrum of a compound expresses implicitly the constitution and the configurational stereochemistry of the local environment of each carbon atom therein. The basic strategy for determining a molecular structure from its spectrum consists in inferring these substructural environments and combining them (according to eventual specific constraints) to produce candidate structures. However, the many-to-many relationships between chemical shifts and substructures and the resulting combinative problems make it difficult to automate this procedure.Of the systems currently proposed, two approaches can be distinguished (2). These concern the method by which the 13C NMR spectra are initially interpreted single-resonance subspectrum/substructural interpretation ( Figure 1); multiresonance subspectrum/substructural interpretation ( Figure 2).The best known elucidation strategy is that defied by the DENDRAL project (3). Elucidation here is carried out in three steps: PLAN-GENERATE-TEST. Extension of this approach to 13C NMR (4) is based on single-resonance interpretation. Substructural constraints are deduced during PLAN step and then submitted to the isomer generator GE-NOA (5) at the heart of the system.Other systems also use this strategy, for example CASE (6, 7) which consists of three modules, INTERPRET-ASSEM-BLE-SIMULATE. The combinative problems encountered motivated the development of a new generator (8), more specialized than a modified version of ASSEMBLE (9).In Sasaki's CHEMICS system (IO) the interpretation of various spectroscopic data ('H and 13C NMR, infrared (IR), mass spectrometry (MS)) eliminates, from a set of fundamental fragments, those that are inappropriate. Constraints can be introduced (11) to restrict the number ...

No abstract

Using a topological parameter, A,, to characterize the effects of a variable mono-or polyalkyl environment, R, on widely varied types of "C sites gives rise to a homogeneous description of their chemical shifts over the entire range of 613C variation (220 ppm/TMS). The proposed model, 613C= w,A,+q, expresses the chemical shift in terms of this parameter together with the sensitivity, o , , of the perturbed 13C site, bonded to a constant environment in a given family, and with qc as the reference shift. In the case of polyalkyl environments, R=Zi=, Ri (nc4); since the use of A, presupposes that groups R, coexist in R without interacting, it is expressed as the sum of the intrinsic effects of each of these groups. Thus, this model demonstrates the general character of most of the common alkyl environment effects and the specific behaviour of the various types of 13C atoms. The homogeneity of this description relies on an assessment of the Ri contribution by the DARC/PELCO topological analysis of alkyl sites in a series of reference alkynes. The 613C values calculated for 138 compounds, belonging to the alkane, alkene, allene, alcohol, amine, epoxide, ester and ketone families, are highly satisfactory (F=0.979, S=0.98).

In the EPIOS system, the elucidation strategy is guided by the progressive intersection of ordered substructures. Pairs combining substructural and spectral information are extracted from a suitable DARC-W-NMR databank. They constitute the basic units for the overlapping process and preserve it from a blind combinatorial matching used to build a molecule and/or its spectra. Enriched knowledge, needed by the system to define a larger solution space, is achieved by investigating the topological relations governing the substructural overlapping. A comparison of the experimental space and its virtual extension through these topological relations is presented.