Prediction of the retention factor in micellar electrokinetic chromatography using computational descriptors and an artificial neural network

Abstract

An artificial neural network (ANN) method is proposed to calculate retention factor of analytes using structural features computed using HyperChem software. The absolute average relative deviation (AARD) and individual deviation (ID) are calculated as accuracy criteria. The accuracy of the proposed method is compared with that of previously reported least square models. The proposed method was tested on eight experimental data sets and mean +/- standard deviation of AARDs for ANN was 10.7 +/- 2.1 and those of previous models were 48.5 +/- 20.4 and 130.1 +/- 79.7, in which the mean differences were statistically significant (p < 0.001). The distribution of IDs sorted in three subgroups, i.e. <= 10, 10-30 and >30%, shows the superiority of the ANN over the previous models.