We have constructed a group contribution method for estimating Kov ts retention indices by using observed data from a set of diverse organic compounds. Our database contains observed retention indices for over 35,000 different molecules. These were measured on capillary or packed columns with polar and non-polar (or slightly polar) stationary phases under isothermal or non-isothermal conditions. We neglected any dependence of index values on these factors by analyzing the median of multiple observations. Using 84 groups we determined two sets of increment values, one for non-polar and the other for polar column data. For non-polar column data the median absolute prediction error was 46 index units (3.2%). For data on polar columns, the median absolute error was 65 index units (3.9%). While accuracy is insufficient for identification based solely on retention, it is suitable for rejection of certain classes of false identifications made by gas-chromatography/mass spectrometry.
Citation: Journal of Chemical Information and Modeling
Pub Type: Journals
group methods, property prediction, retention index