Use of artificial neural networks for the evaluation of electrochemical signals of adenine and cytosine in mixtures interfered with hydrogen evolution

A new method for the simultaneous determination of adenine and cytosine is described. Multivariate calibration based on a suitable experimental design (ED) and soft modeling with artificial neural networks (ANNs) is used for quantitative analysis of overlapped linear scan voltammetric (LSV) and differential pulse polarographic (DPP) peaks of adenine and cytosine that occur in the region of hydrogen evolution. It is demonstrated that analysis of mixtures, even if some of the constituents undergo an irreversible reduction, can be quantified with reasonable accuracy. The average absolute error was estimated as 3.7% in LSV, 4.6% in DPP for adenine and 5.2% in LSV, 5.9% in DPP for cytosine. For the whole testing set the comparison of the added and found values of adenine and cytosine concentrations was characterized by an agreement factor (about 0.06). The method is quite general and can be used for analysis of other biologically important substances without their separation.