Dr. John Goutsias, Professor of Electrical and Computer Engineering
North Carolina State University
The functional properties of biochemical reaction networks are determined by their structural and kinetic parameters. An important problem pertaining to the analysis of such networks is to determine which parameters are most responsible for shaping system behavior and identify those parameters that only weakly affect that behavior. This can be achieved by sensitivity analysis. Among other things, results obtained by sensitivity analysis can produce valuable insights into network robustness and fragility, provide important information for improving experimental design, and lead to effective techniques for selective perturbation and optimal intervention.
Many sensitivity analysis techniques have been proposed in the literature. Most techniques have been applied in a variety of problems of molecular and cellular biology with mixed results. In general, there is a tendency to use the simplest techniques, without paying much attention to their limitations. Inappropriate use of sensitivity analysis may lead to erroneous results and biologically false conclusions. Therefore, it is very important to understand the advantages, disadvantages and limitations of each technique and focus our effort on developing a sensitivity analysis method that can effectively deal with the high complexity and nonlinear nature of biochemical reaction systems relevant to cell biology.
In this talk, we first provide an overview of most popular sensitivity analysis techniques. By using a mathematical model of the MAPK signaling cascade, we clearly show that the use of these techniques may not be appropriate. Then, we discuss a probabilistic approach to sensitivity analysis, which investigates the effects of random parameter fluctuations on the statistical behavior of network response. This method shows great promise for addressing most problems hampering traditional methods. Among other things, it can result in derivative-free sensitivity analysis, can be applied to a wide range of parameter values, may lead to rigorous statistical/information-theoretic approaches for sensitivity analysis, and result in novel methodologies for optimal hypotheses formulation and testing. Unfortunately, the biggest challenge associated with a probabilistic sensitivity approach is the need of extensive numerical Monte Carlo simulations, which, for large biochemical reaction networks, is not practical from a computational point of view. We conclude our talk with a discussion of future directions and possible solutions.
John Goutsias received the Diploma degree in Electrical Engineering from the National Technical University of Athens, Greece, in 1981, and the M.S. and Ph.D. degrees in Electrical Engineering from the University of Southern California, Los Angeles, CA, in 1982 and 1986, respectively. In 1986, he joined the Department of Electrical and Computer Engineering at The Johns Hopkins University, Baltimore, MD, where he is currently a Professor of Electrical and Computer Engineering, a Whitaker Biomedical Engineering Professor, and a Professor of Applied Mathematics and Statistics. His research interests include signal processing and analysis, computational systems biology, and bioinformatics.