Application of Correlation Ranking Procedure and Artificial Neural Networks in the Modeling of Liquid Chromatographic Retention Times (t_R) of Various Pesticides

Abstract

In this research, the study of a quantitative structure-property relationship (QSPR) is described for the examination of the retention behavior of some pesticides on a nonpolar stationary phase. The QSPR analysis included the utilization of the genetic algorithm multiple linear regressions (GA-MLR), the artificial neural network (ANN), the eigenvalue (EV) ranking, and the correlation (CR) ranking procedure techniques. In the MLR model, the employment of five descriptors (one physicochemical and four topological descriptors) was illustrated. A 5-1-1 ANN was generated with the use of these descriptors as inputs. For the test set, the mean relative errors between the ANN calculated and the experimental values of the retention times (t_R) were 1.2%.

Keywords:

• Artificial neural networks
• correlation ranking
• eigenvalue ranking GA-MLR
• pesticides
• QSAR

Notes

^a Desethyldesisopropylatrazine.

^a Statistics of the model: n = 55, R² = 0.804, Q² _L10O = 0.784, SE = 5.56, and F = 68.8.