Symbolic Reachability Analysis of Genetic Regulatory Networks using Qualitative Abstractions

Abstract

The switch-like character of gene regulation has motivated the use of hybrid, discrete-continuous models of genetic regulatory networks. While powerful techniques for the analysis, verification, and control of hybrid systems have been developed, the specificities of the biological application domain pose a number of challenges, notably the absence of quantitative information on parameter values and the size and complexity of networks of biological interest. We introduce a method for the analysis of reachability properties of genetic regulatory networks that is based on a class of discontinuous piecewise-affine (PA) differential equations well-adapted to the above constraints. More specifically, we introduce a hyperrectangular partition of the state space that forms the basis for a discrete abstraction preserving the sign of the derivatives of the state variables. The resulting discrete transition system provides a conservative approximation of the qualitative dynamics of the network and can be efficiently computed in a symbolic manner from inequality constraints on the parameters. The method has been implemented in the computer tool Genetic Network Analyzer (GNA), which has been applied to the analysis of a regulatory system whose functioning is not well-understood by biologists, the nutritional stress response in the bacterium Escherichia coli.