|
|
Departamento de Gravitación y Teoría de CamposSeminarios 2012 |
|
Instituto de Ciencias Nucleares, UNAM |
|
|
Departamento de Gravitación y Teoría de CamposSeminarios 2012 |
|
Instituto de Ciencias Nucleares, UNAM |
Viernes 30 de marzo, a las 14:00 hrs en el Auditorio Moshinsky
Urs Gerber (Max Planck Institute for Biochemistry, Munich)
Modeling of Protein Networks
| A model for protein networks is presented. Proteins can activate or inhibit each other. This is represented by an interaction network. However, the formulation of these interaction networks is usually done by experience and not in a systematic way. Often interactions are assumed which in reality do not take place or are irrelevant. However, a precise understanding of the protein network is crucial for understanding the difference between healthy and diseased cells. When medicaments are added to a network a diseased cell might turn into a healthy one. The protein network is formulated as a signed and directed graph. The proteins can only take the states on (1) or off (0). We use the rules of Boolean logic to determine the resulting graphs in the next time steps. The system either converges to a logical steady state or has a periodic behavior. For a given logical steady state of a protein network it is the aim to find the graph which describes the situation of an experiment the best. Starting with an initial graph we calculate the action of the system in the logical steady state. The Markov chain is built by modifying either logical gates or interactions between the proteins. For the new resulting graph we calculate the action again and determine in a Metropolis step if it is accepted or not. The temperature is not fixed however. It is decreased systematically by using the Simulated Annealing algorithm. This allows us to find the optimal graph to describe the experimental data. |
