Nuclear Magnetic Resonance is the most common tool used for the structure elucidation of new compounds. The used 2D NMR experiments like COSY, HSQC, and 13C–HMBC deliver correlation information between atoms that can be translated into connectivity information. Out of these, correlation information from COSY and HSQC experiments can be transcribed directly into connectivity between atoms. But the 13C–HMBC correlations need more attention because of their ambiguity and complexity. Hence the difficulty of the structure elucidation problem depends more on the type of the investigated molecule than on its size. This ambiguity has driven the development of different software packages like COCON to aid in the interpretation of the 13C–HMBC correlation data [1–19] as much as the development of additional correlation experiments [20,21].
In the case of unsaturated molecules COCON [3, 21-24] will usually generate a very large number of possible solutions. Since the solutions will then have to be checked manually for their chemical feasibility and sense, different efforts have been made to reduce the number solutions. The statistical filter shown compares the suggested constitutions against existing molecules, like the ones contained in the PubChem database. For each Cocon suggested constitution all 1 sphere elements of the constitutions are checked for corresponding elements in PubChem. This comparison is done indirectly, by generating molecular dynam- ics parameters in smi23d, which has been used to generate 3D coordinates for almost 13M compounds contained in PubChem and succeeded on generating coordinates for 99.6% of the molecules contained in the Database. The filtering application eliminates those constitutions for which smi23d fails because of lacking MD parameters.
Since smi23d has successfully been used on so many well known compounds, this means that the structural element for which parameters were missing has hardly ever been observed and therefore might not exist in natural products. Due to the nature of the filter, no ranking of the remaining constitutions is carried out and further methods might be necessary to improve the results.
The filter has been used with several different molecules on the WebCocon Server, and a webpage with some results has been made available on the server: http://cocon.nmr.de/StatisticalFilter/ . Several molecules have been analized, for some, like Ascomycin, the filter showed no reduction in the number of solutions. For others, like Aflatoxin B1, the number of suggested constitutions when using only COSY and 13C-HMBC data, dropped from 970 to 539, a reduction of 45%.
The following literature references are used for this poster: