A method for optimization of pharmacokinetic models

The authors have developed a method for optimizing nonlinear physiologically based pharmacokinetic (PBPK) models using Lagrangian-based or genetic algorithms. The optimization method is demonstrated using PBPK models for pharmacokinetics of nicotine, in which parameters that are not well established can be systematically varied to obtain optimized solutions based on experimental data. The method provides valuable guidance in the determination of these model parameters. The PBPK model for nicotine was developed in C and linked with the ordinary differential equation package, CVODE. The model can be run either stand-alone or under the control of an optimization package. The optimization is performed using SuperCode, running either a Lagrangian-based (VMCON) or a genetic algorithm-based (GALIB) optimizer. To reduce computational time, SuperCode can carry out the optimization in parallel, utilizing the Parallel Virtual Machine (PVM) message-passing software. Both the VMCON and GALIB optimizing algorithms have been used to fit the model to experimental data for humans and Sprague-Dawley rats with good results.