Acceleration of sensitivity analysis by distributed computation

Abstract

We have been developing a comprehensive mathematical model of cardiac myocyte based on molecular functions using a biological simulator, simBio. In this approach, expanding computing power is needed as the model becomes more intricate. Sensitivity analyses, which provide dependency of the model behavior on specific parameters, are also inevitable for developing and utilize models. Meanwhile, distributed computation by a cluster of personal computers (PC) became available using an open source package. A free software package named Jini orchestrates computers on a network, which we coupled with simBio. We connected 11 PCs and found that the time required for computing 504 models became 13 times shorter than that with a single PC. This method was proved efficient for sensitivity analysis, because calculations are independent and a linear decrease of computation time was obtained by adding PCs to the cluster. The visualization feature gives a researcher an instant feed-back, thus this system accelerates model driven study. The whole system with source code is available at www.sim-bio.org. © 2007 IEEE.