Application of topological and physicochemical descriptors: QSTR analysis of the toxicity of benzene derivatives

Quantitative Structure-Activity Relationship (QSAR) models are useful in understanding how chemical structure relates to the biological activity and the toxicity of natural and synthetic chemicals. In the present investigation the applicability of various topological indices and physicochemical descriptor are tested for the QSAR study on benzene derivatives. The topological indices used for the QSAR analysis were Szeged (Sz), Randic (1X) (the first order molecular connectivity), Balaban (J), HyperWiener (HW), Wiener Polarity (WP) and Harary (H) indices. The physicochemical descriptor is also used in the study (n-octanol/water partition coefficient (logP). For obtaining appropriate QSTR model we have used multiple linear regression (MLR) techniques and followed back ward regression analysis. The results have shown that best models are obtained by multi parametric analysis. The toxicities of 45 benzene derivatives are well predicted by a tri parametric model consisting of HyperWiener (HW), Wiener Polarity (WP) and partition coefficient (logP) as the correlating parameters. The predictive ability of the model is discussed on the basis of predictive correlation coefficient.