Hybrid Execution of Dynamic Rule-based Multi-level Models

Hybrid algorithms are a promising approach to speed-up the execution of multi-scale biochemical reaction networks, i.e., networks with reactions that operate on different time scales. The basic idea is to use the quasi-steady state distribution of the fast reactions, computed either analytically or empirically, to update the propensities of slow reactions and to apply a stochastic simulation algorithm to compute the slow reactions. We apply this approach to multi-level models. Executing multi-level models that are characterized by dynamic nested structures by these hybrid algorithms poses specific challenges. For example, all reactions, even fast reactions, can change the structure of the model and consequently the set of reactions. To evaluate our approach, we use the rule-based multi-level language ML-Rules.