|
Internet Electronic Journal of Molecular Design - IEJMD, ISSN 1538-6414, CODEN IEJMAT
| ABSTRACT - Internet Electron. J. Mol. Des. December 2004, Volume 3, Number 12, 759-770 |
Use of an Expert System in Lignan Skeleton Prediction from 1H NMR Data
Mara B. Costantin, Marcelo J. P. Ferreira, Gilberto V. Rodrigues, and Vicente P. Emerenciano
Internet Electron. J. Mol. Des. 2004, 3, 759-770
|
Abstract:
The identification of natural products is a task that demands skilled
spectroscopists in this area. The number of isolated substances has
been enhanced significantly in the last years. Therefore, trying to
reduce the time spent during the identification this process, we
developed a program able to predict the carbon skeleton for a
compound from 1H NMR data. For that, the natural product class
selected in this study was the lignans, products showing some
complex structures. The H1MACH program was developed to assist
the process of skeleton prediction of organic compounds from the 1H
NMR chemical shift data. Thus, a database containing 760 1H NMR
spectra data was established. From the data, the program can predict
the most probable skeleton type for a new compound under analysis
and show several structures that have a high similarity index with the
supplied data. The program was evaluated with 30 lignan structures
not stored yet in the database. The results show that the program was
able to predict, in 70% of the studied cases, the correct skeleton of
the compound. Analyzing with more details the results, one can
verify that in 90% of the tests, the correct skeleton was predicted
among the three first skeletons by the program. Regarding the
obtained results, it can be concluded that the tests carried out with
the program H1MACH showed good results, once that the signal
multiplicity was not included in the database. The procedure here
described can be applied for other classes of compounds. This new
tool will increase the power of spectral data interpretation of the
expert system SISTEMAT.
|