Artificial neural network modeling of pH dependent structural descriptor-mobility relationship for capillary zone electrophoresis of tripeptides

作者：Olajos M; Chovan T; Mittermayr S; Kenesei T; Hajos P; Molnar I; Darvas F; Guttman A^*

来源：Journal of Liquid Chromatography & Related Technologies, 2008, 31(15): 2348-2362.

DOI：10.1080/10826070802281935

• back-propagation artificial neural network
• capillary zone electrophoresis
• multivariable linear regression
• peptide mobility modeling

The aim of this work was to study the structural descriptor-mobility relationship of representative tripeptides in capillary zone electrophoresis (CZE) with the change of such separation parameters as pH, applied voltage and separation length in respect to their influence on electrophoretic migration properties. At the present stage of the work, the ionic charge was considered as structural descriptor. A multivariable linear regression (MLR) model and a back-propagation artificial neural network (BP-ANN) were applied to predict the electrophoretic mobilities of the model tripeptides with non-polar, polar, positively charged, negatively charged and aromatic R group characteristics. Here we present a comprehensive analysis on electrophoretic mobilities measured at pHs 2.5, 4.5, 7.5 and 9.5 at two different capillary lengths of 10cm and 30cm, as well as four applied electric field strengths ranging from 100 to 400 V/cm to teach and evaluate our mobility predicting models. The anticipated mobilities predicted by MLR and BP-ANN were compared to each other and to the experimental data, respectively. The BP-ANN model resulted in considerable higher precision in predictability that of the MLR method.

• 出版日期2008