In this chapter, a computational formalism for modeling and reasoning about the control of biological processes is explored. It comprises five main sections: a survey of related work, a background on methods (including discussion of the Wnt5a gene regulatory network, the coefficient of determination method for deriving gene regulatory network models, and the partially observable Markov decision process model and its role in modeling intervention planning problems), a main section on the approach taken (including algorithms for solving the intervention planning problems and techniques for representing components of the problems), an empirical evaluation of the intervention planning algorithms on synthetic and the Wnt5a gene regulatory networks, and a conclusion and future directions section. The techniques described present a promising avenue of future research in reasoning algorithms for improved scalability in planning interventions in gene regulatory networks.
On periodic reference tracking using batch-mode reinforcement learning with application to gene regulatory network control
