Q-bio:Exploring the (hidden) assumptions of signal transduction models

Brief description

Lecture 2 of the Signal Transduction Unit, which provides an introduction to rule-based modeling and the BioNetGen language, is a prerequisite for this project.

1. Select a published kinetic model of a signal transduction network from the literature. The network should consist primarily of mass action reactions describing the processes of complex formation and 'activation', which usually means phosphorylation, and 'deactivation', which means ubiquitination followed by internalization and proteolysis. A good place to find candidate models is the Biomodels database (biomodels.net), which contains a curated set of published kinetic models that have been manually validated. These are available in a variety of formats including SBML and something that is close to Matlab called Scilab (it is easy to translate the latter into the former).
2. Assemble a list of the proteins and other molecules that comprise the model. In some cases a 'protein' could be a multi-subunit polypeptide complex, but the key point is that each molecule should behave as an indivisible object.
3. Determine all possible interactions among the proteins as specified by the reactions in the model. For each distinct interaction, add a binding site to the protein. For each modification state of the protein, add a modification site to the protein. In some cases a modification and a binding site may be combined into a single site, e.g. a site of protein that must be phosphorylated in order for an interaction with a second protein to take place.
4. If you have chosen a model with sufficiently detailed interactions, you will notice that many possible complexes and modification states have been omitted from the model for simplicity. Which states have been excluded from the model and what justification to the authors give for the exclusion? If justification has been given, is it persuasive?
5. One way to demonstrate 4 is to rewrite each reaction in the network in the BioNetGen notation making explicit the binding and modification state of each component of each molecule in the reaction. Reactions can be generalized by eliminating components that do not play a role in determining whether the reaction takes place or at what rate. Try generalizing some of the reactions in the model, particularly those that you feel do not require all of the specified context. What happens to the size of the model? Do the results of simulating the original model change when one or more reactions are generalized? Do the discrepancies increase when you change the parameters of the model or change the structure of the model in some way, e.g. 'knocking-out' one of the interactions in the model? As a hint, note that the stoichiometry of complex formation often depends strongly on the concentrations of the proteins involved.
6. [hard] Does the model contain any thermodynamic cycles to which the detailed balance constraint should apply? Were the authors aware of this fact and did they include detailed balance constraints in the model? If not, determine the correct constraint relationships among the parameters of the model. Do these change the behavior of the model?
7. [hard] Read up on the structure and biochemistry of one or two of the proteins in the model to learn about their allosteric regulation and newly discovered interactions with other proteins. Expand the model to include these new interactions. How do they alter the predictions of the model? Design an experiment to test the differences. See if you can find any experiments in the literature that are similar to yours or provide another means of distinguishing the models.

[hard] indicates more open-ended and difficult tasks that are optional but could be starting points for significant research projects.

Contact instructor

Jim Faeder

Additional Materials

Prerequisite

Lecture 2, Signal Transduction Unit

Back to The First q-bio School.