List operations on chemical graphs. 4. Using edge models for prediction of retention index data

Previously developed edge models1 for fitting retention index data are validated by the prediction of this property for three compound classes: monoterpenes, diand tricyclic methyl esters and alcohols, and monocyclic ketones and alcohols. As a necessary condition, the predicted compounds must have the same bond types as the calibration data set of model development. The resulting prediction errors can be divided into systematic and statistical parts. Systematic errors are based mainly on different experimental conditions between model development and prediction. The statistical part of the prediction error ranges from 15 to 22 retention index units.