Structure validation in computer-supported structure elucidation: 13C NMR shift predictions for steroids

A new approach for 13C chemical shift prediction using 3D descriptors is described. The stereochemical environment of each carbon atom of a data set of steroids with assigned chemical shifts was encoded and atoms of similar environment were grouped into clusters. Management of errors and optimization of the methodsʹ parameters in order to obtain the best results in chemical shift predictions were handled by special routines. The predicted chemical shifts of unknown compounds were matched against the values observed on the experimental spectra to calculate a degree of support index. It is shown that this index can be used to reject bad structures proposed by structure generators in the task of structure elucidation.