C-13 NMR CHEMICAL-SHIFT PREDICTION OF THE SP(3) CARBON-ATOMS IN THE ALPHA-POSITION RELATIVE TO THE DOUBLE-BOND IN ACYCLIC ALKENES

The C-13 NMR chemical shift of sp(3) carbon atoms situated in the ex p osition relative to the double bond in acyclic alkenes was estimated w ith multilayer feedforward artificial neural networks (ANNs) and multi linear regression (MLR), using as structural descriptors a topo-stereo chemical code which characterizes the environment of the resonating ca rbon atom. The predictive ability of the two models was tested by the leave-20%-out cross-validation method. The neural model provides bette r results than the MLR model both in calibration and in cross-validati on, demonstrating that there exists a nonlinear relationship between t he structural descriptors and the investigated C-13 NMR chemical shift and that the neural model is capable to capture such a relationship i n a simple and effective way. A comparison between a general model for the estimation of the C-13 NMR chemical shift and the ANN model indic ates that general models are outperformed by more specific models, and in order to improve the predictions a possible way is to develop envi ronment-specific models. The approach proposed in this paper can be us ed in automated spectra interpretation or computer-assisted structure elucidation to constrain the number of possible candidates generated f rom the experimental spectra.