Comparative artificial neural network and partial least squares models for analysis of Metronidazole, Diloxanide, Spiramycin and Cliquinol in pharmaceutical preparations

• Linear and nonlinear chemometric models used for analysis of MNZ, SPY, DIX and CLQ.
• Both nonlinear PC-ANN and nonlinear GA-ANN showed best results (recovery %, S.D. and RMSEP).
• The selected methods successfully detected the analyzed drugs in their combined dosage forms.
• ANN is valid alternative to PLS and univariate methods especially in the case of nonlinear data.

Metronidazole (MNZ) is a widely used antibacterial and amoebicide drug. Therefore, it is important to develop a rapid and specific analytical method for the determination of MNZ in mixture with Spiramycin (SPY), Diloxanide (DIX) and Cliquinol (CLQ) in pharmaceutical preparations. This work describes simple, sensitive and reliable six multivariate calibration methods, namely linear and nonlinear artificial neural networks preceded by genetic algorithm (GA-ANN) and principle component analysis (PCA-ANN) as well as partial least squares (PLS) either alone or preceded by genetic algorithm (GA-PLS) for UV spectrophotometric determination of MNZ, SPY, DIX and CLQ in pharmaceutical preparations with no interference of pharmaceutical additives. The results manifest the problem of nonlinearity and how models like ANN can handle it. Analytical performance of these methods was statistically validated with respect to linearity, accuracy, precision and specificity. The developed methods indicate the ability of the previously mentioned multivariate calibration models to handle and solve UV spectra of the four components’ mixtures using easy and widely used UV spectrophotometer.

Figure optionsDownload as PowerPoint slide