Evaluation of NMR predictors for accuracy and ability to reveal trends in 1H NMR spectra of fatty acids

Four different nuclear magnetic resonance (NMR) predictors have been evaluated for their ability to predict 600-MHz H-1 spectra of free fatty acids and fatty acid methyl esters of 20 common fatty acids. The predictors were evaluated on two main criteria: (1) their accuracy in direct prediction of the spectra (absolute accuracy) and (2) the ability to reveal trends or predict the change that occurs in the spectra as a result of a change in the fatty acid carbon chain, or by esterification of the free fatty acids to methyl esters (relative accuracy). The absolute accuracy in chemical shift prediction for fatty acids was good, compared with previous reports on a broader range of compounds. All four predictors had median prediction errors for chemical shifts of the signals in fatty acid methyl esters well below 0.1 ppm and as low as 0.015 ppm for one of the predictors. However, all predictors also had outliers with errors far above the upper interquartile range. In general, they also fail to reproduce trends of diagnostic value that were observed in the experimental data or properly predict the result of a minor change in molecular structure. All four predictors depend on experimental data from different origins. This may be a limiting factor for the relative accuracy of the predictors.

Automated summary

Four different NMR predictors were evaluated for their ability to predict 600-MHz H-1 spectra of free fatty acids and fatty acid methyl esters. The predictors were evaluated on their accuracy in directly predicting the spectra and their ability to reveal trends or predict changes. The absolute accuracy for fatty acids was good, but the predictors had outliers and failed to reproduce trends or predict minor changes in molecular structure. The reliance on data from different sources may limit the relative accuracy of the predictors.