Rule-based Modeling of Cell Signaling: Advances in Model Construction, Visualization and Simulation

[摘要] Rule-based modeling is a graph-based approach to specifying the kinetics of cell signalingsystems. A reaction rule is a compact and explicit graph-based representation of a kinetic process,and it matches a class of reactions that involve identical sites and identical kinetics. Compact rule-based models have been used to generate large and combinatorially complex reaction networks,and rules have also been used to compile databases of kinetic interactions targeting specific cellsand pathways. In this work, I address three technological challenges associated with rule-basedmodeling. First, I address the ability to generate an automated global visualization of a rule-basedmodel as a network of signal flows. I showed how to analyze a reaction rule and extract a set ofbipartite regulatory relationships, which can be aggregated across rules into a global network. Ialso provide a set of coarse-graining approaches to compress an automatically generated networkinto a compact pathway diagram, even for models with 100s of rules. Second, I resolved anincompatibility between two recent advances in rule-based modeling: network-free simulation(which enables simulation without generating a reaction network), and energy-based rule-basedmodeling (which enables specifying a model using cooperativity parameters and automatedaccounting of free energy). The incompatibility arose because calculating the reaction rate requirescomputing the reaction free energy in an energy-based model, and this requires knowledge of bothreactants and products of the reaction, but the products are not available in a network-freesimulation until after the reaction event has fired. This was resolved by expanding each energy-based rule into a number of normal reaction rules for which reaction free energies can be calculatedunambiguously. Third, I demonstrated a particular type of modularization that is based on treatinga set of rules as a module. This enables building models from combinations of modular hypothesesand supplements the other modularization strategies such as macros, types and energy-basedcompression.

[发布机构] the University of Pittsburgh