In this work the 50% effective inhibition concentration after 48 h (log(1/EC50)) of 39 substituted benzenes to the algae Scenedesmus obliquus was predicted by quantitative structuretoxicity relationship (QSTR) approaches. Descriptors which appeared in the best model are; highest occupied molecular orbital, molecular weight, lowest unoccupied molecular orbital, chemical potential, zero point energy, and hardness parameters. These descriptors were obtained from density functional theory (DFT) calculation at B3LYP level together with 6-31G* basis set. Results of this modeling provide the statistics of R = 0.91, F = 19, and SE = 0.30 for training set and R = 0.91, F = 18, and SE = 0.34 for test set. Also the average error and average absolute error in the prediction of log(1/EC50) are ¹0.00041 and 0.198 for training set and for prediction set are ¹0.056 and 0.217, respectively. Moreover, the credibility of the model was tested by cross-validation and Y-scrambling, the results of these tests indicated the reliability of the model. The results of this investigation show that it was possible to predict the molecular toxicity of organic compounds from their DFT-calculated molecular descriptors.
