ENVR 71 |
| The rate by which chemical substances are absorbed into the body affects the magnitude of their pharmacodynamic impact. We have created a predictive model of the oral absorption rate constant (Ka) combining a 2-way neural network (NN) classifier with multivariate linear regression (MLR) models, within each class, to account for the non-linear nature of the absorption process. The range of Ka in the models was from 0.01/hr to 4.69/hr. A set of 20 descriptors was used to build the NN classifier while 18 and 25 descriptors were employed in the MLR models for the individual classes. The final model had a squared correlation coefficient of 0.77 for the train set of 193 compounds and 0.66 for the external test set of 30 compounds. This is first report of a quantitative structure activity relationship (QSAR) model for Ka, which can be used in physiologically based pharmacoKinetic (PBPK) models for drugs or toxins. |
|
General Topics
8:30 AM-8:20 PM, Tuesday, March 27, 2007 McCormick Place South -- Room S100B/C, Level 1, Oral
Division of Environmental Chemistry |