Exact Stochastic Simulation of Chemical Reactions with Cycle Leaping
- Creators
- Riedel, Marc D.
- Bruck, Jehoshua
Abstract
The stochastic simulation algorithm (SSA), first proposed by Gillespie, has become the workhorse of computational biology. It tracks integer quantities of the molecular species, executing reactions at random based on propensity calculations. An estimate for the resulting quantities of the different species is obtained by averaging the results of repeated trials. Unfortunately, for models with many reaction channels and many species, the algorithm requires a prohibitive amount of computation time. Many trials must be performed, each forming a lengthy trajectory through the state space. With coupled or reversible reactions, the simulation often loops through the same sequence of states repeatedly, consuming computing time, but making no forward progress. We propose a algorithm that reduces the simulation time through cycle leaping: when cycles are encountered, the exit probabilities are calculated. Then, in a single bound, the simulation leaps directly to one of the exit states. The technique is exact, sampling the state space with the expected probability distribution. It is a component of a general framework that we have developed for stochastic simulation based on probabilistic analysis and caching.
Additional Information
This work is supported by the "Alpha Project" at the Center for Genomic Experimentation and Computation, a National Institutes of Health Center of Excellence in Genomic Sciences (grant no. P50 HG02370). Also available online: http://www.paradise.caltech.edu/papers/etr077.pdfFiles
Name | Size | Download all |
---|---|---|
md5:d92d86153c73749ed42f5e8901c3b3e2
|
176.2 kB | Preview Download |
Additional details
- Eprint ID
- 26108
- Resolver ID
- CaltechPARADISE:2006.ETR077
- Created
-
2006-12-22Created from EPrint's datestamp field
- Updated
-
2019-11-22Created from EPrint's last_modified field
- Caltech groups
- Parallel and Distributed Systems Group