An Automated Software System for Physiologically Based Pharmacokinetic Modeling Using Parallel Processing on Intel Paragon Supercomputer. Zhen Zhang, Department of Biometry and Epidemiology, Medical University of South Carolina, Charleston, SC; and Hong Zhang and Eberhard O. Voit, Department of Mathematical Sciences, Indiana University and Purdue University at Fort Wayne, IN
We have developed a software system under PC/MS Windows for physiologically based pharmacokinetic (PBPK) modeling and simulation that allows a user to specify a PBPK model and its parameters with free-hand drawing of compartments representing organs/tissue regions and the associated connecting pathways. The program, through the use of a graph theoretic method that we have derived based on the topological property of PBPK models, automatically generates all system equations and then produces the simulation results with one of several user-selectable numerical integration methods. In this paper, we extend the software system to include a parallel processing simulation module to be used on the Intel Paragon supercomputer. The MS Windows based PBPK modeling and simulation program now has an additional option for PBPK models with a large number of organs and/or tissue regions. With this option, instead of trying to solve the PBPK differential equations locally, the program generates the necessary system descriptions to be used by the parallel processing simulation module. The simulation module maps the problem to the available number of CPU nodes on the Paragon and optimizes the use of parallel computing resources to reduce the total simulation processing time. The results are then passed back to the PC based program for display and further analysis. We believe that with the widely available PC-Internet-supercomputer connections, such a hybrid software system, with its familiar MS Windows environment and graphical user interface, will bring the computational power of parallel processing to the end-user of PBPK modeling and simulation and greatly shorten model development time.