Title:Probing the Effects of Stochasticity in Biochemical Reaction Networks using Multiscale Simulation

Abstract: Using our recent contribution, the "partitioned leaping algorithm" [L. A. Harris and P. Clancy, J. Chem. Phys. 125, 144107 (2006)], we investigate the effects of stochasticity in two model biochemical reaction networks. By considering situations both where "leaping" proves beneficial and where it does not, we gain valuable insight that can aid in and advance the use of leaping methods in computational biology. In particular, we identify reaction subnetworks with small populations and large rate constants as a major bottleneck for leaping. We also demonstrate the use of "model reduction" in conjunction with leaping to circumvent this problem and expose a current need in this area: a model-reduction hierarchy analogous to that on which the leaping methods are based [D. T. Gillespie, J. Chem. Phys. 113, 297 (2000); 115, 1716 (2001)]. In situations where leaping is seen to perform well, we emphasize the significant advantages to using these methods over more traditional simulation approaches. We show how leaping allows for statistical investigations beyond the reach of "exact-stochastic" methods and demonstrate how one can uncover, via such investigations, subtle stochastic effects that would be ignored in deterministic treatments. We then show, using examples taken from the literature, that conditions identified here as lending well to leaping do, in fact, arise in natural biological systems. This leads us to discuss and demonstrate how the statistical information garnered from a leaping study can be used to infer mechanistic details of a biological process and motivate experiments. Overall, we demonstrate to biologists the great potential that leaping methods hold for real-world biological research while highlighting to algorithmists areas still in need of improvement.