Modeling and Symbolic Analysis of Protein Signaling Networks
Using Hybird Automata
Ronojoy Gosh, Stanford University, Computational Systems Biology
Cluster Hire Candidate
April 6, 2005, 12-1 p.m.
1441 Genetics/Biotechnology Center (new addition)
Hybrid automata models can capture observed phenomena in protein networks: the protein concentration or activation dynamics inside each cell are modeled using linear differential equations; inputs (such as genes or other proteins) activate or deactivate these continuous dynamics through discrete switches, which themselves are controlled by protein concentrations reaching given thresholds. The compelling reason to adopt this paradigm is that we are now capable of performing completely symbolic computation on certain classes of hybrid automata. That is, instead of numerical parameters, we can compute symbolic ranges for parameters such as switching thresholds and chemical rates necessary for a given process. More interestingly, we are able to compute symbolic reachable sets for experimentally observed biological steady states, using abstraction and verification algorithms. This gives us a set of initial conditions (i.e. protein concentration/ activation levels) that leads to a particular steady state, which can be used to predict which initial conditions will lead to that steady state, and to suggest experiments to study those. This talk will focus on the hybrid automata models of two signaling pathways: (i) Cell differentiation and pattern formation through Delta-Notch lateral inhibition in Xenopus embryonic skin cells, and (ii) planar cell polarity (PCP) signaling in Drosophila wing cells. Analysis results, as well as preliminary experimental model validation data for the PCP signaling model, will be discussed. They are both local feedback signaling pathways that produce robust global patterns at steady state, and their analysis give us insight into the mechanics and effects of such networks, in large cell populations.