|
ICMSB 2008 10th International Conference on Molecular Systems Biology |
|
tutorial SessionS
|
|
T1: Introduction to Analyzing Biological Systems with Canonical Models Speaker: Eberhard Voit (UC Davis, USA) Abstract: Mathematical modeling has great potential in biological systems analysis because, in contrast to the unaided human mind, mathematics has no problems keeping track of hundreds of interacting variables that affect each other in intricate ways. The scalability of mathematical models, together with their ability to capture all imaginable nonlinear responses, allows us to explore the dynamics of complicated biological systems, to study what happens if a gene, protein, metabolite, cell, or organism is altered, and to optimize biological systems, for instance, toward the goal of increased yield of some desired organic compound. Before we can utilize models for such purposes, we must define their mathematical structure and identify suitable parameter values. Because nature has not provided us with guidelines for selecting the best model design, the choice of the most useful model is not trivial. I show here that canonical models offer guidance for model selection, construction and analysis that is otherwise difficult to find. Specifically, focusing on power-law modeling within Biochemical Systems Theory (BST), I will demonstrate the generic modeling steps of (1) model choice, (2) model design, (3) model diagnostics, and (4) model use, and illustrate them with simple and moderately complex examples. References:
T2: Efficient Inferring Method of Genetic Interactions Based on Time Series Gene Expression Profiles: Application of Conceptual Modeling by S-System Formalism Speaker: Masahiro Okamoto (Kyushu U, Japan) Abstract: The expression profiles of hundreds and thousands of genes on a genomic scale can be measured simultaneously by recent powerful technologies such as DNA microarrays, DNA chips and so on. These observed data depending on its environment are usually obtained as snapshots, but can be generated as dense time series that indicate the dynamic behavior. The experimentally observed time-course data should contain enormous information about the regulation of genetic networks in vivo. However, since this information is entirely implicit, it requires adequate analytical and computational methods of retrieval and interpretation. This inference problem of genetic networks based on the experimentally observed time-course data is generally referred to as "inverse problem" and can be defined as function optimization of the values of parameters involved in a suitable model-representation of genetic network. In brief, we have to estimate the values of a set of system parameters in the model which can realize given experimentally observed time-course data. The key points solving such an inverse problem are how to set up canonical representation of mathematical modeling of genetic network and how to explore and exploit the values of parameters within immense huge searching space. We had first proposed a novel inferring method of genetic network by combining a dynamic network model called S-system with a computational technique of parameter estimation based on simple genetic algorithms. S-system is based on a particular type of differential equation in which the temporal (time-dependent) dynamic processes of system components are characterized by power-law formalism and S-system is suitable for the conceptual modeling and the description of organizationally complex systems involved looping or cyclic interactions between system components such as metabolic pathways and gene expression networks. The values of interrelated coefficient in this formalism are directly or indirectly related to the regulation mechanism in the modeled network, and the inferred network structure resulting from the estimation of parameters should be one of the better candidates for genetic network structure. S-system formalism, however, has a major disadvantage in that this formalism includes a large number of parameters that must be estimated; the number of estimating parameters is 2n(n+1) (where n is the number of system components). Simple genetic algorithm (SGA) is one of the well-known heuristic optimizer of such large number of parameters, however, SGA has two intrinsic problems; one is early convergence in the first stage of search and the other is evolutionary stagnation in the last stage of search. Recently real-coded genetic algorithms (RCGAs) attract attention as numerical optimizing methods instead of SGA. One of the crossover operators for RCGAs called unimodal normal distribution crossover (UNDX) has shown good performance in optimizing of various functions including multi-modal ones and benchmark functions with epistasis among parameters. Furthermore Sato et al. have proposed new generation-alternation of model called minimal generation gap (MGG) model to avoid early convergence in the first stage and to suppress evolutionary stagnation in the last stage. Using S-system modeling and RCGAs with the combination of the UNDX and MGG, we previously proposed efficient procedures for the inference of genetic interactions based on the experimentally observed time-course data of system components (mRNA). There are many network candidates of gene expression which can realize the same experimentally observed facts, however, the structures of these network candidates of gene expression are different each other. Therefore, we should propose efficient analytical method for extracting useful information from many network candidates of gene expression. We previously proposed an analyzing procedure for extracting core interactions from many network candidates, and confirmed that the sensitivity of kinetic parameters included in common core binomial interactions is significantly greater than that included in other unique interactions. The interactions with having high sensitivity much contribute to realize experimentally obtained time-course data of gene expression network. We will be able to use these interactions as clue when we investigate about organizationally complex system. It is difficult to use such common core binomial interactions to analyze the dynamic behavior on the gene expression network even if becoming a clue that investigates important relations between genes. Therefore, in this tutorial, we shall describe about the efficient method for extracting common core binomial interactions of an enough number to analyze the dynamic behavior of the gene expression network.
T3: Parameter Estimation Methods in Biochemical Systems Theory: Classical and Evolutionary Approaches Speakers: Ricardo del Rosario (UPD, Philippines and MPI Biochem, Germany) and Prospero Naval, Jr (UPD, Philippines)
T4: Digital Libraries and Workflow Processes for Systems Biology Speaker: Su-Shing Chen (PICB Shanghai, China) Abstract: We will survey the field of digital libraries and its extension to workflow processes, which has been contributed by the speaker during the last 10 years and plus. Then we will discuss the relevance of these two topics to systems biology and survey the requirements of building such systems.
T5: Extracting Network Models from Pathway Databases Speaker: Baltazar Aguda (Ohio State U, USA) Abstract: This tutorial begins with an overview of the nature of hundreds of pathway-related databases now available on the internet, and ends with an example of how a network model was derived by integrating information from these databases. Key methods involved in the extraction of network models include modularization and qualitative network analysis; these will be illustrated with a modeling problem concerned with the entry into the mammalian cell division cycle - a cellular process that is compromised in most cancers.
|