A novel QSPR study of normalized migration time for drugs in capillary electrophoresis by new descriptors: quantum chemical investigation.

Some drugs' migration time (MT) has been studied employing quantitative structure-property relationship using new descriptors that are able to predict MT value with high accuracy. MT property modeling of the drugs was established as a function of the new theoretically derived descriptors applying multiple linear regressions and partial least-squares regression. The genetic algorithm was used to select those variables that resulted in the best-fitted models. To select a set of descriptors that are most relevant to MT, illustrating the affecting degree for the affinity of different descriptors, the linear models with 1-14 variables were constructed and were then investigated based on F-value, squared regression coefficients of cross-validated (Q2), adjusted R2 (R2adj) and standard error of estimate (S) statistical parameters. Finally, the best model with ten variables was selected. Statistical parameters of the test set, such as standard deviation error in test, were 0.559 and 0.616, while relative error of test was equal to 7.648 and 8.497% for multiple linear regressions and partial least-squares models, respectively, confirming the good predictive ability of the model. Since the capillary lengths were not the same for the drugs in the data set, MT values were normalized based on a specific capillary before modeling, which is also one of the advantages of this method, enabling us to use the model for different capillary lengths.