Artificial neural network aided estimation of the electrochemical signals of monosaccharides on gold electrode

Artificial neural networks were used to predict the oxidation peaks potentials of 7 monosaccharides under linear sweep voltammetry regime. Two sets of descriptors, one based on molecular properties calculated through DFT and another based on simple geometric distributions of hydroxyl groups and asymmetric carbon atoms along molecular chains, were employed to introduce the molecules to networks. Relatively, simple networks of (3,3,1) and (3,3,3,1) structures with the number of epochs not exceeding 15 through training process were capable of correctly predicting the peaks positions with R values in the range of 0.97–0.99.

Two sets of descriptors derived through DFT calculations and geometric considerations are fed to artificial neural networks to predict electrochemical signals of monosaccharides.