Original Articles: 2011 Vol: 3 Issue: 1
Physico-chemical parameter prediction from drug structure using multiple linear regression and artificial neural networks
A set of adamantane derivatives (AD) as drug were tested for their chromatographic behavior and Kovats retention index (RI) were determined for all the compounds. Quantitative structure Property relationship (QSPR) analysis was applied to 32 of the AD.Molecular descriptors derived solely from 3D structures of the molecular compounds. Modeling of RI of AD as a function of the theoretically derived descriptors was established by multiple linear regression (MLR) And artificial neural networks (ANNs) for the prediction of Kovats retention index . The models were constructed using 25 molecules as training set, and predictive ability tested using 7 compounds. The usefulness of the quantum chemical descriptors, calculated at the level of the Density Functional Theory(DFT) theories using 6-31+G** basis set for QSPR study of AD was examined. A multi-parametric equation containing maximum five descriptors at B3LYP/6- 31+G** method with good statistical qualities (R2 train=0.914, Ftrain=97.674, R2 test=0.770, Ftest=3.214, Q2LOO=0.895, R2adj=0.904,Q2 LGO=0.84451) was obtained by Multiple Linear Regression using stepwise method.