Prediction of human clearance from animal data and molecular structural parameters using multivariate regression analysis.

The aim of the study reported here was to develop a method for predicting human clearance that can be applied to various kinds of drugs using clearance values for rats and dogs and some molecular structural parameters. The clearance data for rats, dogs, and humans of 68 drugs were obtained from literature. The compounds have various structures, pharmacological activities, and pharmacokinetic characteristics. In addition, molecular weight, c log P, and the number of hydrogen bond acceptors were used as possible descriptors related to the human clearance value for each drug. Three types of regression methods, multiple linear regression (MLR) analysis, partial least squares (PLS) method, and artificial neural network (ANN), were used to predict human clearance, and their predictive performances were compared with allometric approaches, which have been widely used in interspecies scaling. In MLR and PLS analyses, interaction terms were introduced to evaluate the nonlinear relationships. For the data sets used in the present study, MLR and PLS with quadratic terms gave the same equation and the best predictive performance. The value of the squared cross-validated correlation coefficient (q(2)) was 0.682. In conclusion, the MLR method using animal clearance data from only two species and using easily calculated structural parameters can generally predict human clearance better than allometric methods. This approach can be applied to drugs with various characteristics.