Artificial neural network modeling of the electrophoretic mobility of dexamethasone and two additives in micellar electrokinetic capillary chromatography

Multiple response simultaneous optimization was used to develop a micellar electrokinetic capillary chromatography (MEKC) method for the determination of dexamethasone sodium phosphate (DEX) and the two additives, creatinine (CRE) and propylparaben (PRO) in DEX ampoules. The goal is to optimize the separation parameters to achieve maximum resolution between the three components within a reasonable run time. Buffer composition and voltage were optimized using an artificial neural network (ANN). The best neural network with a "2-10-4" structure and 150 epochs was applied. A back propagation algorithm with sigmoidal transformation function in the nodes was used for constructing the ANN. The generalization ability was studied by consulting the network with test data and calculating the performance parameters, e.g.correlation (R), mean square error (MSE), root mean square error (RMSE) and the percentage-relative error (E sub(r)%). The predicted results were in good agreement with the experimental values, indicating that a neural network is a potential method for the selection of separation conditions in MEKC. The optimum separation was achieved when using a background electrolyte of 20 mM borate buffer containing 30 mM sodium dodecyl sulfate (SDS) at pH 9.5 along with applying high voltage (30 kV) to the capillary. The total separation was reached within 4.39 min. This optimized method was applied for the analysis of commercial DEX ampoules. The method was fully validated as per ICH guidelines.