Evaluation of calibration data in capillary electrophoresis using artificial neural networks to increase precision of analysis

Increase of precision in capillary electrophoresis can be achieved applying suitable markers and evaluating calibration curves and data analysis with artificial neural networks. They are able to account for errors in both x- and y-axes, nonlinear response of detector and non-linearity of calibration curves eventually. A comparison of the artificial neural networks approach with ordinary least-squares (OLS) and bivariate least-squares regression (BLS) was done. While OLS and BLS give similar results, the method proposed and tested in analysis of several pharmaceutical products yields lower prediction errors than traditional linear least-squares methods and the precision of analysis was found in the range 0.5–1.5% relative.