Modeling the retention behavior of some drugs in high performance affinity chromatography

The major goal of the present study is to predict retention behavior for a series of drugs such as warfarin, diazepam and ect. in high performance affinity chromatography [1,2]. Multiple linear regression (MLR) was used as variable selection method to investigate the relationship between the selected descriptors and the retention times of the studied drugs.The molecular descriptors were generated from the molecular structure of drugs and calculated by the DRAGON software. Besides DRAGON descriptors, additional quantum mechanical descriptors were obtained using density functional theory/B3LYP/6-31G method in Gaussian 09 software. By excluding some descriptors with ignorable contributions, components of the QSRR models were simplified. The data set was divided into two groups including training set (21 molecules) and test set (7 molecules). The theoretical modeling of the retention time of studied compounds was established as a function of molecular descriptors by using linear regression method.The predictive ability of the models was evaluated using cross-validation, Y-randomization and an external test set. The results showed that MLR is a good method for predicting retention times of studied drugs.The best model is : ret=6.948+(12.656*RNCG)+(1.392*c-028)-(80.797*R6e)-(1.255*IDDE)Correlation coefficients of the selected model were R=0.982 and R= 0.99 for training and test sets, respectively.