SIMULATION OF C-13 NUCLEAR-MAGNETIC-RESONANCE SPECTRA OF TETRAHYDROPYRANS USING REGRESSION-ANALYSIS AND NEURAL NETWORKS

The C-13 NMR spectra of tetrahydropyrans are simulated directly from t heir molecular structures. A set of 29 tetrahydropyrans is used as a t raining set to generate regression equations and to train neural netwo rks, and three additional compounds are used as a cross-validation set . The results of simulations done by regression analysis are found to be extremely sensitive to molecular geometries. To account for this, t wo different methods of descriptor manipulation, an averaging method a nd a Boltzmann-weighted averaging method, are introduced, and the mode ls generated from the descriptor sets are compared. The results for th e BolItzmann-weighted averaging method are found to be better than tho se based on descriptors derived from only the lowest energy conformati on.