Artificial neural networks modeling in ultra performance liquid chromatography method optimization of mycophenolate mofetil and its degradation products

Abstract

The study of experimental design in conjunction with artificial neural networks for optimization of isocratic ultra performance liquid chromatography method for separation of mycophenolate mofetil and its degradation products has been reported. Experimental design showed to be suitable for selection of experimental scheme, while Kennard-Stone algorithm was used for selection of training data set. The input variables were column temperature and composition of mobile phase including percentage of acetonitrile, concentration of ammonium acetate in buffer, and its pH value. The retention factor of the most retentive component and selectivity factors were used as the dependent variables (outputs). In this way, artificial neural network has been applied as a predictable tool in solving a method optimization problem using small number of experiments. Network architecture and training parameters were optimized to the lowest root-mean-square error values, and the network with 5-4-4-4 topology has been selected as the most predictable one. Predicted data were in good agreement with experimental data, and regression statistics confirmed good ability of trained network to predict compounds retention. The optimal chromatographic conditions included column temperature of 40 degrees C, flow rate of 700 mu l min(-1), 26% of acetonitrile and 9 mM ammonium acetate in mobile phase, and buffer pH of 5.87. The chromatographic analysis has been achieved within 5.2 min. The validation of the proposed method was also performed considering selectivity, linearity, accuracy, precision, limit of detection, and limit of quantification, and the results indicated that the method fulfilled all required criteria. The method was successfully applied to the analysis of commercial dosage form. Copyright