Study on aromatic hydrocarbons toxicity to <i>chlorella vulgaris</i> based on QSAR model

Abstract

Predicting ability based on the quantitative structure–activity relationships (QSAR) model of unknown aromatic hydrocarbons toxicity is one of the tasks of security precaution. To establish the QSAR model between the physical and chemical properties of aromatic hydrocarbons and the inhibited activity of Chlorella vulgaris(C. Vulgaris), the optimized geometries, based on the 96 hr-EC50 of 25 aromatic hydrocarbons with C. Vulgaris were carried out at the B3LYP/6-311G** level by density functional theory (DFT) calculation. With matlab2010(a) software, multiply linear regression(MLR) and three types principal components regression (PCR) methods were used to develop the QSAR model. Three methods were introduced to select the PCs, namely k-values(K), eigenvalue ranking(EV) and correlation ranking(CR) procedures. The different resutls then were compared. After eliminated one of the collinear descriptor and constants, 6, 1 and 4 PCs were select by K, EV, CR for PCR method respectively. LOO cross-validated coefficient (R²_CV) of training set to MLR,K-PCR,EVPCR,CRPCR were 0.673,0.817,0.874,0.907,the R² of prediction were 0.569,0.678,0.519,0.79,respectively.