First-principle, structure-based prediction of hepatic metabolic clearance values in human.

The first-principle, quantitative structure-hepatic clearance relationship for 50 drugs was constructed based on selected molecular descriptors calculated by TSAR software. The R(2) of the predicted and observed hepatic clearance for the training set (n=36) and test set (n=13) were 0.85 and 0.73, respectively. The average fold error (AFE) of the in silico model was 1.28 (n=50). The prediction accuracy of in silico model was superior to in vitro hepatocytes' model in literature (n=50, AFE=2.55). It is attractive to predict human hepatic clearance based on molecular descriptors merely. The structure-based model can be used as an efficient tool in the rapid identification of hepatic clearance of new drug candidates in drug discovery.