Mechanism Study In Bio-signaling Networks Using Random Parameter Sampling

Computer modeling is an important method in biological signaling network researches. Parameter estimation is the most important and difficult step of modeling. In common cases, in spite of the aim of modeling, parameter estimation begins with candidate generation by approximating the simulation result to the experiment result (top-down), or directly calculating from a large group of experiment raw data (bottom-up). Then the probable parameters will be tested and adjusted with another group of experimental data. Finally, one or a small group of best-fit parameters will be used in the farther experiments. However, we consider that this method is suitable for result prediction but may be not for mechanism study. To find a best-fit parameter means to find a network state that is most close to the state in real cell, or a network state that is best for the biological activity of cell. But we considered the worst states can sometimes give important information about network mechanism as the best states. This information will be lost in the traditional method basing on best-fit parameters. To solve this problem, we tried to analyze the network model with a large group of random parameters instead of a best-fit one. Changing the type of model and reducing the sampling space would be used to make the time cost of simulation acceptable. Network mechanism should be reflected by the relationship between simulation results and parameter changes. In this article, we investigated two biological signaling networks with different computer models (apoptosis network in Boolean model and MAPK pathway in ODE model). Finally we successfully reproduced the dynamics of these two networks and achieved meaningful results about their design principles and mechanisms.