Using theoretical molecular descriptors as well as partial least squares(PLS) regression,two quantitative structure-activity relationship(QSAR) models were developed for depuration rate constants(kd) of polychlorinated biphenyls(PCBs) in two species of mussels,Perna viridis and Dreissena polymorpha.The cross-validated Q2cum(an indicator of fitting of goodness) values for the two models are 0.501 and 0.756,and the standard deviation(SD) is 0.084 and 0.076,respectively.The achievement of satisfactory Q2cum and low SD values indicates good predictive ability and precision of the two models.The significant descriptors governing lgkd include polarizability(α),molecular volume(MV),molecular weight(Mw),molecular surface area(S),and total energy(TE).The key descriptors in the models reflect that van der Waals interactions play a dominant role in the depuration of PCBs.The depuration of PCBs in the two mussel species may be mainly attributed to the biota-water phase partitioning processes.