Principal Investigator Paul Barton
Biological systems have evolved to make extremely sophisticated decisions and control complex processes, while also being resilient and flexible. For example, signal transduction cascades provide an important set of pathways that sense and process extracellular signals and trigger internal cellular events. This project is developing tools to analyze and compare biological networks and classes of networks. The key hypothesis is that biological networks have evolved to function well in a complex environment and are therefore likely to be (near) optimal among a class of related networks. Given this, a key part of understanding their behavior is elucidating the precise sense in which they are optimal and the class of related networks with respect to which they are optimal. This question can, in principle, be answered by posing and solving an optimization problem and comparing the solution with the networks observed in nature. In addition, these same optimization formulations can be applied for the design of optimal biochemical networks.
The dynamics of a broad range of biochemical networks implementing functionality such as signaling, regulation, decision making and cell division and growth can be modeled with ODEs or DAEs. The class of related networks can be formulated as a superstructure of pathways, and then (in principle) the solution of a mixed-integer optimization formulation can determine the optimal network by switching off a subset of the pathways. This requires algorithms for solving mixed-integer nonlinear optimization problems with dynamic systems embedded. As with all mixed-integer optimization formulations, it is absolutely necessary to take a global optimization approach in order to avoid numerical artifacts in the results. Work on global dynamic optimization algorithms is making this approach feasible.