An QSAR Model for Predicting PBDEs Toxicity Established Based on Ridge Regression

摘要

In this paper, the ridge regression (RR) method was employed to establish the quantitative structure-activity relationships (QSAR) model for predicting toxicity with 15 polybrominated biphenyl ethers (PBDEs) and their 27 kinds of quantum descriptors. Quantum descriptors used to establish the QSAR model were filtrated out based on correlation analysis and variables importance of project (VIP) supported by partial least squares (PLS). The multicollinearity among the descriptors was removed during the calculation of RR method in order to ensure the validation of the final regression equation. The research showed that descriptors of Δα, α_(xx), α_(xy), α_(xz), α_(yz), β_(xxy) and β_(yyy) had significant effect on toxicity. The model with the simulation efficiency coefficient of 0.916 could be used to predict the toxicity of the unchecked PBDEs and as a preliminary analysis for environmental risk of organic compounds.