A way to predict retention indices (RI) from component structures was investigated.In contrast to work already published components with different functional groups and different stationary phases were chosen.Molecular descriptors were used to code the structural information.Multi-linear Regressions on a variety of descriptor combinations vs.retention factors and retention indices failed, so an Artificial Neural Network (ANN) approach was used.The results were analyzed on precision of single retention indices and on the elution order of all possible 2 component sets.The ANN was trained with 75 % of the measured values.Errors were calculated for the remaining 25 % of the data set and the whole set.The experimental data set contained the retention indices of 109 components from 8 different chemical groups injected onto 14 different stationary phases.