Estimation of bioconcentration factors using molecular electro-topological state and flexibility

Bioconcentration assessment is important in the scientific evaluation of risks that chemicals may pose to humans and environment and is a current focus of regulatory effort. In this work, a new QSAR model by adopting electronic topological properties and flexibility of chemicals to predict the bioconcentration factor (BCF) in fish was established based on a large number of diverse compounds. Multiple linear regression (MLR) and partial least squares (PLS) were used to build reliable QSARs, which were evaluated with internal five cross-validations (Q(cv)(2)) and an external validation (Q(ex)(2)). The proposed MLR model showed reasonable predictivity of BCF (Q(cv)(2) = 0.79, Q(ex)(2) = 0.79) and included seven molecular descriptors, namely SsCl, SaasC, SaaaC, SsNH2, Hmin, SssO, and Phia. The PLS model (Q(cv)(2) = 0: 83, Q(ex)(2) = 0.80) was shown to be slightly better than the MLR one in prediction accuracy, using six PLS latent components. In addition, the relationship between the log BCF and the theoretical calculated log K-ow was extensively investigated. These studies may help to understand the factors influencing the bioconcentration process of chemicals and to develop alternative methods for prescreening of environmental toxic compounds.