Quantitative structure-retention relationship study of some phenol derivatives using artificial neural networks

The gas Chromatographic retention times (RT) of 37 derivatives of phenols on a DB-5 column have been predicted using an artificial neural network (ANN). Molecular descriptors, including topological, getaway and 2D autocorrelations descriptors, were calculated by the use of Hyperchem and Dragon programs. To select the descriptors as inputs for the artificial neural network the multiple linear regression (MLR) technique was used. The neural network is a fully connected Back- Propagation model which is used in the present study with a 4-4-1 architecture. The results obtained using the neural networks were compared with those obtained using the MLR technique. Standard error of training and standard error of prediction for the MLR model and for the ANN model were compared. The results of which showed the better prediction power of the neural network. Moreover, the mean effect of descriptors shows that Kier symmetry index descriptor is most important parameter affecting the retention behavior of molecules.