Algorithms for Multiscale Discrete Stochastic Simulation | ||
|
Traditional differential equation-based models for simulating chemically reacting systems fail to capture the randomness inherent in intracellular biochemical processes. The stochastic simulation algorithm (SSA) derived by Gillespie [1] faithfully captures exact behavior of these systems by simulating every reaction event. However, because it simulates every reaction, the SSA is inherently slow. It is possible to speed up the simulations considerably with minimal loss of accuracy by taking advantage of timescale separation. In recent years, with close collaboration with Dan Gillespie, we have developed the underlying theory and implementation of several multiscale stochastic simulation algorithms. "Tau-leaping" takes large time-steps and simulates potentially many reactions in each time step. The "slow-scale SSA" (ssSSA) accelerates simulations where fast reactions involve chemical species that are present in small numbers. And most recently, the "slow-scale tau-leaping" method incorporates aspects of both the tau-leaping and ssSSA methods. In our current work, we're continuing to build the theoretical understanding of multiscale simulation methods and applying the theory to create new algorithms that adaptively select the appropriate simulation method to achieve ultimate speed and accuracy. This research is funded by the DOE, NIH, and the UCSB Institute for Collaborative Biotechnologies.
[1] D.T Gillespie. J. Phys. Chem., 81:2340-2361, 1977. [2] F. Schlogl and R.S. Berry. Phys. Rev. A, 21(6):2078-2081, 1980. |
