Prediction of retention factors in micellar electrokinetic chromatography from theoretically derived molecular descriptors

摘要

An artificial neural network (ANN) was constructed and trained for the prediction of the retention factors of some benzene derivatives and heterocyclic compounds in micellar electrokinetic chromatography (MEKC) based on quantitative structure-property relationship. The inputs of this network are theoretically derived descriptors, which were chosen by the stepwise multiple linear regressions features selection technique. These descriptors are; molecular surface area, Kier shape index, dipole moment and maximum positive charge on the Carbon atom which were used as inputs for constructed 4:2:1 ANN. By comparing of the results obtained from multiple linear regression and ANN models, it can be seen that statistical parameters (Fisher ratio, correlation coefficient and standard error of the model) of the ANN model are better than that regression model, which indicates that nonlinear model can simulate the relationship between the structural descriptors and the MEKC retention of the investigated molecules more accurately. Also the cross-validation test was used for the evaluation of the predictive power of the ANN model. The statistical parameters obtained were Q2 = 0.57 and PRESS = 0.55, which reveals the credibility of ANN model.