Prediction of 13C NMR chemical shifts by artificial neural network. I. Partial charge model as atomic descriptor

Abstract

Mulliken population analysis (MPA), Hirshfeld population analysis (HPA), Charge Model 5 (CM5) and Hu Lu Yang charge fitting method (HLY) were considered in order to reveal influence of atomic partial charges on the 13C NMR chemical shifts. The test set included seven classes of organic molecules. Partial charges of carbon atoms were obtained from quantum-chemical calculations at DFT/HISS level. Linear regressions were constructed as estimators of accuracy of each model. The best approach was shown by multivariate regression with MPA, HPA, and CM5 charges as predictors in a linear model with mean value of R2=0.8917.