Using electrotopological state indices to model the depuration rates of polychlorinated biphenyls in mussels of Elliptio complanata.

A quantitative structure-activity relationship (QSAR) model on depuration rate constants (k(d)) of polychlorinated biphenyls (PCBs) in freshwater mussel Elliptio complanata was successfully constructed using electrotopological state indices (ESI) and partial least squares (PLS) regression. The cross validated Q2(cum) and the correlation coefficient R for the model were determined to be 0.845 and 0.954, respectively. The satisfactory Q2(cum) and R values indicated significantly high robustness and good predictive ability for the model. The model was tested and found acceptable for the prediction of logk(d) (the logarithm of the depuration rate constants) by validation set. According to the model, an increase in the values of S(aasC), S5', S4, S5 and S4, led to increased logk(d), and a decrease in the values of N(Cl), S2' and S6 also resulted in increased logkd. Among these descriptors, N(Cl), S(aasC), S5', S4 and S5 made significant contributions to the value of logk(d). These significant descriptors showed that the depuration of PCBs in Elliptio complanata may be mainly attributed to an equilibrium partitioning process among compartments with different lipid contents, while the reactivity of PCBs with enzymes or other molecules may play a subordinate role.