Abstract
Comparative quantitative structure-toxicity relationship (QSTR) models of a series of saturated alcohols were constructed based on topological, constitutional, geometrical, and getaway descriptors. In the first step, all groups of descriptors were applied to 29 saturated alcohol derivatives. Molecular modeling and geometry optimization was carried out with CAChe Pro software. Calculation of descriptors was done by using Dragon software and multilinear regression analysis by using Project Leader software. Various QSTR models for each set of descriptor in different combinations were then developed, and only the top five models for each set of descriptors were chosen. Of the top five, the best QSTR model of each set of descriptors was used for comparative study. The study shows that the best QSTR model ((C)RE5 = -0.0795186 M-W + 0.110997 AM(W) + 23.0915 S-V - 4.62153 Se - 11.9932 Sp - 0.613194 Ss + 5.68544) is made from constitutional descriptors. The best model was selected on the basis of values of correlation coefficient (r (2) = 0.853) followed by other regression quality parameters such as cross-validation coefficient (r (CV) (2) = 0.453), standard error (SE = 0.453) and standard error of estimation (SEE = 0.346), F-statistic (F = 25.57), and p values (p = 0.00) that were calculated by statistical software. Secondly, it has been reported that there is a direct relationship between reported biological toxicity of the alcohols and the sum of Kier-Hall electrotopological states (Ss) descriptor of the constitutional group. Sum of Kier-Hall electrotopological states (Ss) can alone be helpful for searching for alcohols of reliable toxicities before their synthesis.