Modeling and simulation of molecular networks.
During the last decade finite dynamical systems, that is, discrete dynamical systems with a finite phase space, have been used increasingly in systems biology to model a variety of biochemical networks, such as metabolic, gene regulatory and signal transduction networks. In many cases, the available data quantity and quality and not sufficient to build detailed quantitative models such as systems of ordinary differential equations, which require many parameters that are frequently unknown. In addition, discrete models tend to be more intuitive and more easily accessible to life scientists. Boolean networks and their generalization, the so-called multistates logical models, Petri nets, and agent-based models are the main types of finite dynamical systems that have been used in this context.
Discrete models require less detailed information about the system to be modeled, so they can be used in cases where quantitative information is missing. They are also useful if qualitative predictions from the model are desired, such as whether a T cell becomes pro- or anti-inflammatory. Finally, discrete models are very intuitive compared to models based on differential equations or other more sophisticated formalisms. It is also easier to explore their dynamics, at least for reasonably small models. On the other hand, an important disadvantage of discrete models is that there are very few theoretical tools available for their analysis. Typically, discrete models are built by translating information from the literature into logical statements about the interactions of the different molecular species involved in the network. In many cases, the biological information about a particular network node might not be sufficient, however, to construct a logical function governing regulation. In the case of a continuous model, the remedy would be to insert a differential equation of specified form, e.g., mass action kinetics, with unspecified parameters. If experimental time course data are available one can then use one of several parameter estimation methods to determine those unspecified model parameters so that the model fits the given data. Data fit is determined by model simulation, using numerical integration of the equations in the model.
We have developed a software package, Polynome , which
carries out discrete analogs of parameter estimation for Boolean
networks, a particular kind of discrete model. The software integrates
several packages for network inference we have developed in recent
years. In the case of missing information about a particular node in the
network to be modeled, one can insert a general Boolean function, maybe
of a specified type, e.g., a nested canalyzing function. If
experimental time course data for the network is available, then one can
use one of several existing inference methods to estimate a function
that will result in a model that fits the data. This function in
addition satisfies a specified optimality criterion, similar to the
optimality criterion for the fitting of continuous parameters. It
furthermore couples parameter estimation with extensive simulation
Cancer systems biology.
There is increasing evidence that differences in iron metabolism play an important role in cancer risk, survival, and clinical prognosis. However, to date, why these differences in iron metabolism exist, how they contribute to malignant processes, and whether they can be successfully exploited to therapeutic advantage remains unknown. This is because the current approach to identifying whether these or other of the proteins in the complex network of iron metabolism should be targeted therapeutically is an empiric, protein-by-protein validation exercise. The intent of this project is to use the power of systems biology to identify the key regulatory points in this complex pathway, and how they change in malignancy.
In work published in 2009, our group described the
iron metabolic network in normal cells, including subnetworks for
different cell types. We distilled this complex network
into a system of feedback control loops that represent important points
of potential control of the network. The goal of this project is to build on this network to create predictive models of iron metabolism in normal and cancer cells. This will help identify key nodal points in iron metabolism and how they change as cells progress to malignancy. Ultimately, we hope to be able to leverage knowledge gained to therapeutic or diagnostic advantage.
For more information, visit http://www.vbi.vt.edu/faculty/group_overview/Laubenbacher_Research_Group.
If you are seeking a postdoctoral or PhD position and have an interest in joining our group, please send your CV including a statement of interests to email@example.com.
Current group members:
Former group members: