Abstrato

Prediction of retention times for a large set of pesticides or toxicants based on quantitative structure-retention relationships

A.H.M.Sarrafi, M.Alizadeh, F.Ahmadi

In this paper, three different multivariate calibration methods feed-forward artificial neural networks (ANN) with back-propagation learning rule, Partial Least Squares (PLS) and Multiple Linear Regression (MLR) were applied to predict the retention time of 103 diverse pesticides or toxicants in gas chromatography-mass spectrometry (GC-MS) by usingmolecular structural descriptors. Five descriptors are considered to account for the effect of solute structure on the retention time. These are (solvation connectivity index chi-1, 3D-Balaban index,Hautocorrelation of lag 3/weighted by atomic sanderson electronegativities, relative negative charge, Wiener-type index from Z weighted distance matrix (Barysz matrix). The Stepwise SPSS was used for the selection of the variables that resulted in the best-fitted models. After variables selection, 103 compounds randomly are divided into three training, validation and test sets. The mean square error (MSE) of training, test and validation sets for theANN model are 0.0008676, 0.0014 and 0.0013, respectively. Result obtained showed that nonlinear model can simulate the relationship between structural descriptors and the retention times of the molecules in data set accurately.