Using quantum chemical descriptors and partial least squares (PLS) regression, a quantitative structure -activity relationships (QSARs) model was developed to predict the depuration rate constants (k d ) of polycyclic aromatic hydrocarbons (PAHs) for mussels, Elliptio complanata. With a high cumulative cross-validated regression coefficient value (Q 2 cum ) of 0.927 and low standard deviation (SD) of 0.065, the model obtained by the training set shows a good predictive ability, and it is validated to be robust by predicting the test set. Among 20 quantum chemical descriptors, the dielectric energy (DE), the molecular weight (M w ), and the highest occupied molecular orbital energy (E HOMO ) are the key descriptors governing the logk d values in the model. Increase in the DE or decrease in the M w values leads to the increase in logk d , indicating the van der Waals interactions and steric hindrance effect on the depuration process. Decrease in the E HOMO values results in increasing the logk d values, implying important roles the molecular orbital energies may play in the biological depuration of PAHs in mussels.