Lumpability abstractions of rule-based systems (Q418843)

The goal of this paper is to speed up the analysis of complicated networks of chemical reactions in molecular biology. For instance, the speed of one reaction occurring at one end of a big molecule may be independent of whether another reaction at the opposite end of the molecule has occurred. This paper exploits such situations by splitting a stochastic model of a biomolecular system into equivalence classes and performing the analysis on the resulting quotient model instead of the original model. After discussing the motivation, the paper illustrates the main biomolecular concepts and shows how they map to weighted labelled transition systems (WLTS) and their trace density semantics. Then it presents a language called Kappa for representing dynamic stochastic biomolecular systems. A Kappa agent represents a molecule or sub-molecule in some form that it can be transformed to (e.g., bound or not bound to another sub-molecule). A Kappa mixture is a multiset of agents satisfying some well-formedness conditions. A Kappa system consists of a set of Kappa rules and a probability distribution on an abstracted mixture (the initial distribution). The abstraction only removes syntactical differences. The evolution of a Kappa system is determined by the rules and induces a WLTS. The paper continues by defining the equivalence used in computing the quotient model. Then it recalls and slightly develops concepts and results from Markov chain and WLTS literature, and cites a result (proven elsewhere) that they can be applied to Kappa systems.