Model Identification: A Key Challenge is Computational Systems Biology
The primary goal of computational systems biology is the integration of biological data into mathematical models. Due to rapid advances in biological techniques, these data consist more and more of cellular responses in the form of time series measurements of gene expression, protein abundances, or metabolite concentrations following some stimulus. Time series data contain enormous information, but this information is not always explicit but has to be extracted with computational methods. This inverse task faces distinct challenges. Most often discussed are purely computational difficulties. Foremost, the algorithms employed for optimizing the fit between model and data often do not converge, converge very slowly or approach a local minimum that is much inferior to the true, global optimum. Other rather evident challenges are related directly to the data,which may be overly noisy, uncertain or partially missing. Less attention has been paid to issues associated with the particular choice of a mathematical model representation, and there has almost been no discussion of the quality of data fit beyond the residual error and the efficiency of an algorithm in terms of the time required to find a satisfactory solution. Finally, there are uncounted statistical questions regarding the design of time series experiments and the assessment of model fits, most of which still await the development of new methods. This presentation discusses inverse tasks in the context of metabolic pathways and describes some advances toward a set of effective algorithms.
Biochemical Systems Theory (BST) Canonical Model Metabolic Pathway Parameter Estimation Systems Analysis System Identification
Eberhard O.Voit
Integrative BioSystems Institute, Georgia Institute of Technology, 313 Ferst Drive,Suite 4103, Atlanta, Georgia 30332-0535, USA
国际会议
The Second International Symposium(OSB08)(第二届国际优化及系统生物学学术会议)
云南丽江
英文
1-12
2008-10-31（万方平台首次上网日期，不代表论文的发表时间）