Accelerated Simulation of Hybrid Biological Models with Quasi-Disjoint Deterministic and Stochastic Subnets
- First Online:
- Cite this paper as:
- Herajy M., Heiner M. (2016) Accelerated Simulation of Hybrid Biological Models with Quasi-Disjoint Deterministic and Stochastic Subnets. In: Cinquemani E., Donzé A. (eds) Hybrid Systems Biology. HSB 2016. Lecture Notes in Computer Science, vol 9957. Springer, Cham
Computational biological models are indispensable tools for in silico hypothesis testing. But with the increasing complexity of biological systems, traditional simulators become inefficient to tackle emerging computational challenges. Hybrid simulation, which combines deterministic and stochastic parts, is a promising direction to deal with such challenges. However, currently existing algorithms of hybrid simulation are impractical for implementing real and complex biological systems. One reason for such limitation is that the performance of hybrid simulation not only relies on the number of stochastic events, but also on the type as well as the efficiency of the deterministic solver. In this paper, a new method is proposed for improving the performance of hybrid simulators by reducing the frequent reinitialisation of the deterministic solver. The proposed approach works well with models that contain a substantial number of stochastic events and higher numbers of continuous variables with limited connections between the deterministic and stochastic regimes. We tested these improvements on a number of case studies and it turns out that, for certain examples, the amended algorithm is ten times faster than the exact method.