A study on systems modeling frameworks and their interoperability

'What is in a model?' is a question that any systems modeler has asked at least once when working with a biological system. To answer this question we must understand how every species and reaction in a model are biologically related to each other, which in the case of cellular systems requires an understanding of local protein and gene interactions at the domain level. Typical modeling approaches like reaction-networks do not explicitly encode this information, which can obfuscate the assumptions made by a model and limit their further analysis, comparison and re-use. Additionally, the multi-state nature of the elements in a model can lead to a combinatorial explosion of the interactions in the model, pushing reaction-networks to their limits. Rule-based modeling is a modeling paradigm that addresses these issues through the use of a graph-based representation of each species's interaction domains and their local interactions. In this paradigm, reaction \emph{rules} describe a class of reactions and are represented as graph rewriting operations that encode biological processes like bond formation and post-translational modification. This allows for a concise, explicit encoding of how species interact, modify and are related to each other while at the same time managing the enormous number of potential molecular interactions in a multi-state system. In this work I present two different approaches that aim to bring the benefits of rule-based modeling's graph representation to other modeling paradigms: \emph{Atomizer} and \emph{MCell-R}. Atomizer is an algorithm for the extraction of structural and process information from reaction-network models, which I use to encode the model in a rule-based format. Once we have an explicit understanding of what is in a model and how entities in a model interact, we can then perform meta-modeling operations like model analysis, alignment, comparison and visualization, which I demonstrate through the application of Atomizer to a large dataset of reaction network models. MCell-R is a framework for the efficient modeling and simulation of multi-state, multi-component spatial systems. The framework consists of an integration of the NFsim rule-based simulation engine together with the MCell spatial modeling system, which highlights the utility of bringing rule-based paradigms into reaction based platforms.