Control Biology, Wolkenhauersub

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Systems Biology

Looking at opportunities and challenges in applying systems theory to molecular and cell biology.

By Olaf Wolkenhauer, Hiroaki Kitano, and Kwang-Hyun Cho


ystems theory or systems science has never managed to achieve widespread and independent status in curricula, departments, and journals but instead acts as an umbrella for a number of research activities across the physical and engineering sciences. Now, with revolutionary developments in the life sciences, there is renewed interest in systems thinking. In this article, we survey opportunities and challenges for the application of systems theory to biology in the postgenomic eraa new area of research also referred to as systems biology. With the sequencing of DNA for a number of genomes, scientists now have an inventory of genes available to embark on the study of the organization and control of genetic pathways. This new phase in the biological revolution, the postgenomic era, is closely associated with the fields genomics, transcriptomics, proteomics, and metabolo-

S
38

mics (called the omics for short). These fields take us from the DNA sequence of a gene to the structure of the product for which it codes (usually a protein) to the activity of that protein and its function within a cell, the tissue, and ultimatively the organism. A series of articles in Nature [1] are recommended for an introduction to these research areas of the biomedical sciences. With the emergence of the omics, molecular biology currently witnesses a shift of focus from molecular characterization to the understanding of functional activity. The two central questions scientists investigate are What are the genes functional role? and How do genes and/or proteins interact? Answering these questions has become possible with new high-throughput technologies to take measurements at the molecular level. In the past, single genes were studied, but with DNA microarray technology we can now measure the activity levels of thousands of

0272-1708/03/$17.002003IEEE IEEE Control Systems Magazine

August 2003

EYEWIRE

genes at the same time. Thus it becomes possible to identify interrelationships between groups of genes (with respect to their functional role) and to analyze dynamic interactions among genes (gene networks). Similarly, proteomics research shows that most proteins interact with several other proteins, and it is increasingly understood that the function of a protein is appropriately described in the context of its interactions with other proteins. Most of these interactions are the consequence of dynamic and controlled processes, and it is not surprising that there is renewed interest in the application of systems thinking to biology. The rest of this article is organized as follows. We first introduce systems biology in the context of the study of complex systems, reviewing a number of related and relevant areas of research and define complexity in the context of biological systems. Systems biology has a history, and its early stages in the 1960s involved eminent researchers, including Wiener, Kalman, Bertalanffy, Rosen, and Mesarovic. We discuss why these attempts disappeared from the research agendas and why there is renewed interest in the postgenomic era of the life sciences. We then follow with two examples of systems biology before outlining current activities from groups around the world. We conclude by listing some of the challenges and hurdles for this (re)emerging field.

there have been located some spots beneath which almost surely there is pay-dirt [2]. Following Hakens synergetics, chaos theory, and fractals, the science of self-organized criticality [3], nonequilibrium physics, power laws, and emergent phenomena renewed the interest in complexity studies over the last decade or so. These studies developed mostly within the areas

Systems biologyapplying systems theory to biology in the postgenomic era.


of physics and mathematics. They are seeking general principle of phenomena that can be observed in a wide range of disciplines. Kauffmans work on genetic networks [4] paved the way for complexity studies in biology. The work of Goodwin [5], [6], Harrison [7], and Meinhardt [8] marked a trend toward an approach more focused on specific organisms but continued to investigate cellular processes and morphological development in evolutionary terms. As Harold pointed out in his recent book: Complexity studies is a fresh label for a well-known pigeonhole: general systems theory, that was pioneered by Ludwig von Bertalanffy in the 1930s [9, p. 222]. Systems biology is an emerging field that continues this research into the postgenomic era of the life sciences [10]-[12]. Complexity studies and systems biology are different in that the latter takes a signal- and systems-oriented approach to describe the dynamic processes within and between biological cells. As we shall see later, systems biology has more to do with the application of systems and control theory to cellular systems than with the application of physics to biology. Systems biology provides a vital interface between cell biology and biotechnological applications. Before we discuss this area in greater detail in subsequent sections, we note that complexity in the context of biological systems can be defined as a property of an encoding (mathematical model) (e.g., its dimensionality, order, or number of variables) an attribute of the natural system under consideration (e.g., the number of components, descriptive and organizational levels that ensure its integrity) our ability to interact with the system and to observe it (i.e., to make measurements and generate experimental data). On all three accounts, genes, cells, tissues, organs, organisms, and populations are individually and as a functional whole a complex system. It is the availability of experimental techniques, modern microscopy, laser tweezers, and nanotechnology, as well as DNA microarrays, gel technology, and mass spectrometry that drives this renewed interest in complexity studies and systems biology. While the

Genomic Cybernetics
The understanding of causality and coping with complexity is not the holy grail of science but part of its very essence. Not surprisingly then, complexity studies have remained as elusive as inconclusive. Weaver [2] defined disorganized complexity as a problem in which the number of variables is very large and any of these variables is best described as a random process. Here we are at the molecular level, and the most successful formal methods for representing phenomena at this level derive from statistical considerations. In the context of the cell, at the cellular level, matters are complicated by the fact that organization becomes an essential feature of the processes under consideration. Weaver referred to problems in which a large number of factors are interrelated into a whole as organized complexity. The number of variables is too large to be dealt with in the Newtonian realm of physics and mathematical modeling, and the systems are too organized to allow statistical techniques. At the time, Weaver described organized complexity as the challenge for science in the coming 50 years. The enthusiasm he expressed in 1948 is very similar to how one feels today in the postgenomic era of the life sciences: It is doubtless true that we are only scratching the surface of the cancer problem, but at least there are now some tools to dig with and

August 2003

IEEE Control Systems Magazine

39

technology to generate and manage data races ahead, it becomes apparent that methodological advances in the analysis of data are urgently required if we want to turn the newly available data into information and knowledge. This need for research into new methodologies and the development of novel conceptual frameworks has been neglected in the euphoria about new technology. Problems in the post-

Systems biology has more to do with the application of systems and control theory to cellular systems than with the application of physics to biology.
genomic era of the life sciences will not only be experimental or technical, but also conceptual. The interpretation of data, turning information into knowledge, is as important for scientific and biotechnological progress as the possibility of generating and managing data. With the generation of vast amounts of data, computer scientists have been the natural allies of biologists in the management of these data. The growth of bioinformatics parallels the exciting developments in biology. However the availability of genome sequence data has led to a focus shift from molecular characterization and sequence analysis to an understanding of functional activity and now interactions of genes and proteins in pathways. Gene expression and regulation, i.e., to understand the organization and dynamics of genetic, signaling, and metabolic pathways, is the challenge for the next 50 years. The nature of the experiments and the data thereby generated requires an alliance of the biological and biomedical sciences with physical scientists (engineers, mathematicians, and physicists). The following discussion on the challenges and hurdles will clarify why such an alliance is so important.

multidisciplinary research centers that are being built around the world, gently forcing researchers to interact by confining them into purpose-built housing. Mesarovic further suggested that progress could be made by more direct and stronger interactions of biologists with system scientists: The real advance in the application of systems theory to biology will come about only when the biologists start asking questions which are based on the system-theoretic concepts rather than using these concepts to represent in still another way the phenomena which are already explained in terms of biophysical or biochemical principles. ... then we will not have the application of engineering principles to biological problems but rather a field of systems biology with its own identity and in its own right. Molecular characterization has led to very accurate spatial representations of cellular components, and biochemical modeling has been the main approach to studying cellular processes. However, the future lies in extending this knowledge to observations at higher organizational levels. There are few examples of a concerted effort to translate biological representations of gene expression and regulation into the language of the system scientist [13], [14], and all indications are that the field is going to provide the vital interface between basic cell biology, physiology, and biotechnological applications such as in metabolic engineering. Systems biology has new technologies available to generate data from the genome, transcriptome, proteome, and metabolome, in addition to the physiome. However, while bioinformatics is usually associated with vast amounts of data available in databases, the systems-biological description of cellular processes often suffers from a lack of data.

Gene Expression and Regulation


Each cell of a (multicellular) organism holds the genome with the entire genetic material, represented by a large double-stranded DNA molecule with the famous double-helix structure. Cells are therefore the fundamental unit of living matter. They take up chemical substances from their environment and transform them. The functions of a cell are subject to regulation such that the cell acts and interacts in an optimal relationship with its environment. The central dogma of biology describes how information, stored in DNA, is transformed into proteins via an intermediate product, called RNA. Transcription is the process by which coding regions of DNA (called genes) synthesize RNA molecules. This is followed by a process referred to as translation, synthesizing proteins using the genetic information in RNA as a template. Most proteins are enzymes and carry out the reactions responsible for the cells metabolismthe reactions that allow it to process nutrients, to build new cellular material, to grow, and to divide.

(Not) A New Kid on the Block


Although generally considered to be a new area of research, systems biology is not without history, and as early as the 1960s the term was used to describe the application of systems and control theory to biology [12]. At the time, Mesarovic wrote: In spite of the considerable interest and efforts, the application of systems theory in biology has not quite lived up to expectations. ... one of the main reasons for the existing lag is that systems theory has not been directly concerned with some of the problems of vital importance in biology. Today, scientists in this field are motivated by the availability of experimental data, including, for example, DNA microarray time series, and interdisciplinary collaborations are widely supported. In fact, the importance of interdisciplinary research and close collaborations between biologists and physical scientists is evident in the many

40

IEEE Control Systems Magazine

August 2003

Research conducted in the 1960s showed that most basic logical cells are not running a program but rather continucellular processes are dynamic and feedback regulated. ally sensing their environment and making decisions on the While investigating regulatory proteins and the interactions basis of that information. To determine how cells act and inof allosteric enzymes, Jacob and Monod introduced the dis- teract within the context of the organism to generate cohertinction between structural genes (coding for proteins) ent and functional wholes, we need to understand how and regulatory genes, which control the rate at which information is transferred between and within cells. Cell sigstructural genes are transcribed. This control of the rate of naling, or signal transduction, is the study of the mechaprotein synthesis was the first indication that such pro- nisms that enable the transfer of biological information. cesses are most appropriately viewed as dynamic systems. Signaling impinges on all aspects of biology, from developFigure 1 illustrates the processes of gene expression and regulation in bacterial cells. Although bacterial cells are capable of producing several thousand different proteins, not all are produced at the same time or in the same quantity. The energy consumption for protein synthesis and the relatively short half-life of the RNA molecules are reasons for the cell to control both the types and amounts of each protein. One example of a global regulatory network is the heat-shock response. When proteins are exposed to extremes of heat, they are said to undergo denaturation. Denaturation is the destruction of the folding properties of a protein leading (usually) to loss of biological activFigure 1. Gene expression and regulation in bacteria. Information, stored in ity. To counteract possible toxic effects from the DNA, is transformed into proteins via an intermediate product called mRNA. insoluble aggregates in the cell, the change in The short half-life of mRNA and the energy consumption of protein synthesis form temperature and the quantity of denatured pro- the basis for a sophisticated hierarchy of a control mechanism. teins are sensed by the cell and specific heatshock proteins are produced. Figure 2 illustrates heat-shock regulation of the DnaK operon in the bacterium Bacillus subtilis. The protein DnaK is such a chaperone, one of a group of proteins called molecular chaperones, which help other proteins to fold properly. These specialist proteins produce barrel-like structures, providing an environment for the denatured proteins to refold. The described mechanism is referred to as negative control through repressor deactivation. A repressor protein is a regulatory protein that binds to a specific site on the DNA and thereby blocks transcription. Drawings like Figure 1 are frequently used in biology textbooks to illustrate structural aspects and the spatial organization of components in the cell. Figure 2, on the other hand, also shows the signal flow and temporal effects components have. The next section introduces an area in which signal- and systems-oriented representations play an even greater role.

Intra- and Intercellular Dynamics: Cellular Weather Forecasting


The previous example described how genes act and interact within the context of the cell. Bio-

Figure 2. Negative regulation of the dnaK and GroESL operons in Bacillus subtilis. The HrcA repressor is a regulatory protein that binds at specific sites on DNA and blocks transcription. In the absence of stress, the GroESL co-repressor binds with HrcA and thereby increases repression of the dnaK operon genes. Upon heat shock, the transcription rate of a group of heat shock proteins, called chaperones, is increased. They build barrel-type structures, which help denatured protein to refold and regain its function.

August 2003

IEEE Control Systems Magazine

41

ment to disease. Many diseases, such as cancer, involve malfunction of signal transduction pathways. Downward [15] provided an excellent account of this field. Figure 3 illustrates a very basic signaling model. As indicated in the previous section, bacteria regulate cell metabolism in response to a wide variety of environmental

Environmental Signal

Cell Surface Receptor (Sensor Kinase)

Phosphorylated Protein

Cell Membrane Intracellular Protein

Response Regulator Phosphatase P RNA Polymerase DNA Promoter Operator Gene(s)

Figure 3. Cell signaling (signal transduction). Intracellular dynamics (gene expression) can be affected by extracellular signals. Receptors spanning the cell membrane receive signals and transmit the information to activate intracellular proteins (the response regulator). In the figure, the response regulator binds to the operator region of a gene and prevents the RNA polymerase from transcription of the adjacent gene. A phosphatase ensures that the process is continuous.

fluctuations, including the heat-shock example above. Thus, there must be mechanisms by which the cells receive signals from the environment and transmit them to the specific target to be regulated. Receptors are proteins that span the membrane, with a site for binding the signaling compound on the outer surface. Binding of the extracellular signaling compound to the outer surface of the receptor results in an activation of an intracellular protein (the response regulator), for example, by phosphorylation. Signal transduction pathways commonly consist of many more cascaded modules between receptor and genome. There can be numerous intermediate steps before the signal transduction process ends, often with a change in the gene expression program of the cell. In the figure, the phosphorylated response regulator is a DNA binding protein, which serves as a repressor, preventing the RNA polymerase from transcribing the adjacent gene(s). In addition to crosstalk between pathways, negative feedback systems can occur, and the time course of a signal transduction pathway can be critical. It is therefore important to develop experimental techniques that allow quantitative measurements of proteins and protein interactions. Mathematical modeling and simulation in this field has the purpose to help and guide the biologist in designing experiments and generally to establish a conceptual framework in which to think. The article by Hasty [16] provides a survey of such in numero molecular biology (see also [17]). Smolen [18] surveys mathematical modeling of transcriptional control and future directions. The signal-oriented approach to cellular models by Kremling et al. [13] is an example of systems and control theory bridging the gap between cellular biology and metabolic engineering. Progressing from merely descriptive models to predictive models will require the integration of data analysis and mathematical modeling with information stored in biological databases. The data we currently have available do not allow parametric systems identification techniques to build predictive models. Instead, it is the systems thinking, the modeling process itself, that often proves useful.

Example: Modeling of Ras/Raf-1/MEK/ERK Signal Transduction Pathway


The Ras/Raf-1/MEK/ERK module in Figure 4 is a ubiquitously expressed signaling pathway that conveys mitogenic and differentiation signals from the cell membrane to the nucleus [19]-[22]. This kinase cascade appears to be spatially organized in a signaling complex nucleated by Ras proteins. The small G protein Ras is activated by

Figure 4. Biologists drawing for the Ras/Raf-1/MEK/ERK signal transduction pathway.


42 IEEE Control Systems Magazine

August 2003

many growth factor receptors and binds to the Raf-1 kinase with high affinity when activated. This induces the recruitment of Raf-1 from the cytosol to the cell membrane. Activated Raf-1 then phosphorylates and activates MAPK/ERK kinase (MEK), a kinase that in turn phosphorylates and activates extracellular signal regulated kinase (ERK), the prototypic mitogen-activated protein kinase (MAPK). Activated ERKs can translocate to the nucleus and regulate gene expression by the phosphorylation of transcription factors. This kinase cascade controls the proliferation and differentiation of different cell types. The specific biological effects are crucially dependent on the amplitude and history of ERK activity. The adjustment of these parameters involves the regulation of protein interactions within this pathway and motivates a systems biological study. Figures 5 and 6 describe the circuit diagrams of the biokinetic reactions for which a mathematical model is used to simulate the influence of ligand variations on the pathway k1 S + E SE k 3 E + P.
k2

can be dissociated into E and S with a rate constant k2 or it can further proceed to form a product P with a rate constant k3 . It is required to express the relations between the rate of catalysis and the change of concentration for the substrate, the enzyme, the complex, and the product. Based on this reaction kinetics [23], we first consider a basic modeling block

Signal transduction pathways can be represented as sequences of enzyme kinetics reactions which turn a substrate S into a product P via an intermediate complex SE and regulated by an enzyme E. The rate by which the enzyme- substrate complex SE is formed is denoted by k1 . The complex SE holds two possible outcomes in the next step. It

Figure 5. Basic pathway modeling block. The pathway model is constructed from basic reaction modules like this enzyme kinetic reaction for which a set of four differential equations is required.

Figure 6. Graphical representation of the Ras/Raf-1/MEK/ERK signal transduction pathway (the shadowed part represents the suppression by RKIP): a circle represents a state for the concentration of a protein and a bar a kinetic parameter of the reaction to be estimated. The directed arc (arrows) connecting a circle and a bar represents a direction of a signal flow. The bidirectional thick arrows represent an association and a dissociation rate at the same time. The thin unidirectional arrows represent a production rate of products.
August 2003 IEEE Control Systems Magazine 43

of signal transduction pathways. This basic modeling block is illustrated in Figure 5 and can be described by the following set of nonlinear ordinary differential equations: dm1( t ) = k1 m1( t )m2( t ) + k2 m3( t ) dt dm2( t ) = k1 m1( t )m2( t ) + k2 m3( t ) + k3 m3( t ) dt dm3( t ) = k1 m1( t )m2( t ) k2 m3( t ) k3 m3( t ) dt dm4 ( t ) = k3 m3( t ). dt From these we have m2( t ) + m3( t ) = C 1 , m1( t ) + m3( t ) + m4 ( t ) = C 2. Hence we can describe the basic reaction module by two nonlinear equations subject to two algebraic conditions.

In general, for a given signal transduction system, the whole pathway can be modeled by a set of nonlinear differential equations and a set of algebraic conditions in the following form: dm( t ) = f( m( t ), k( t )), dt
i { 1, p }

m (t ) = C
i

where m( t ) = [m1( t ), m2( t ),K , mp ( t )], k( t ) = [k1( t ), k2( t ),K , kq ( t )], p is the number of proteins involved in the pathway,q is the required number of parameters, and j {1,K , J } with the number of algebraic conditions J < p. Parameter estimation is widely regarded as a major problem in dynamic pathway modeling [25], [26]. A simple method first discretizes the nonlinear differential equations into algebraic difference equations that are linear with re-

Figure 7. Illustration of parameter estimation from time series data: Each parameter is determined from the value to which the estimates converge (shown by the horizontal line). (Note that any experimental noise can be further eliminated by regression techniques if multiple experimental replicates at each time point are available.)

44

IEEE Control Systems Magazine

August 2003

spect to the parameters and then solve the transformed linear algebraic difference equations to obtain the parameter values at each sampling time point. We can then estimate the required parameter values by employing curve fitting, calculation of steady state values, and regression techniques. For this purpose of parameter estimation, the previous equations are transformed into dm( t ) k( t ) = g m( t ), , dt and this can be further transformed into a set of algebraic difference equations by approximating the differential operator vector g via a difference operator vector h as k( t ) h ( m( t ), m( t 1),K , m( t r )) where r depends on the order of approximation. Without loss of generality, k( t ) can be approximated by k since most

of the signal transduction systems can be regarded as slowly time varying systems compared with the measurement windows in time scale. Hence we have k h ( m( t ), m( t 1),K , m( t r )), which implies the parameter estimates based on time course measurements. The entire model, as shown in Figure 6, is constructed in this way, leading to what usually becomes a relatively large set of differential equations for which parameter values have to be identified. As illustrated in Figure 7, in the estimation of parameters from western blot data, the parameter estimates usually appear as a time dependent profile since the time course data include various uncertainties. However, since the signal transduction system itself can be considered as time invariant, the estimated parameter profile should converge to a constant value at steady state. Figure 7 illustrates this estimation procedure.

Figure 8. The simulation results for fixed initial conditions: (a) shows the binding of RKIP to Raf-1*, (b) shows the binding of MEK-PP to ERK-P, (c) shows the binding of ERK-PP to Raf-1*/RKIP, and (d) shows the binding of RP to RKIP-P.
August 2003 IEEE Control Systems Magazine 45

If a reasonable model is constructed, this can then be used in a variety of ways to validate and generate hypotheses, or to help experimental design [22], [24]. Based on the mathematical model illustrated in Figure 6 and the estimated parameter values as, for example, obtained using a discretization of the nonlinear ordinary differential equations (as illustrated in Figure 7), we can perform simulation studies to validate the signal transduction mechanism as illustrated in Figure 5 and also to analyze the signal transduction system with respect to the sensitivity for the ligand (via simulation of variable initial conditions) as illustrated in Figure 9.

Current Research: An International Perspective


The interest in systems biology is documented by the increasing number of conferences, research groups, and institutes dedicated to this area. Funding initiatives in the United States, Japan, South Korea, and Germany provide new opportunities to the engineering sciences. Systems biology provides evidence of the growing involvement of control engineers, not just at the technological level, but also playing a vital role in the development of novel methodological ap-

proaches in mathematical modeling, simulation, and data analysis [27]. To allow large scale simulations, international projects are developing a systems biology markup language (SBML) [28] and Systems Biology Workbench to allow the integration of models and simulation tools (visit http://www.sbw-sbml.org for further information). The principal challenge for the biomedical sciences is to answer the following questions [6]: How do cells act and interact within the context of the organism to generate coherent and functional wholes? How do genes act and interact within the context of the cell to bring about structure and function? For systems biology we can summarize the challenges as follows: methodologies for parameter estimation experimental and formal methods for model validation identification of causal relationships, feedback, and circularity from experimental data modular representations and simulation of large scale dynamic systems investigations into the stability and robustness of cellular systems

Figure 9. The simulation results for variable initial conditions: (a) shows the variation of Raf-1*, (b) shows the variation of ERK, (c) shows the variation of RKIP, and (d) shows the variation of RKIP-P.
46 IEEE Control Systems Magazine August 2003

visualization and fusion of information, integration of models and simulators scaling models across time scales and description levels (from genes, transcripts, and proteins to cells and organisms). Gene expression takes place within the context of a cell a n d b e t w e e n c e l l s , o rg a n s , a n d o rg a n i s m s . T h e reductionist approach is to isolate a system, conceptually closing it off from its environment through the definition of inputs and outputs. We inevitably loose information using this approach since conceptual closure amounts to the assumption of constancy for the external factors and the fact that external forces are described as a function of something inside the system. Different levels of cellular systems may require different modeling strategies, and ultimately we require a common conceptual framework that integrates different models, for example, differential (mass action or rate) equations provide a realistic modeling paradigm for a single-gene or single-cell representation, but cell-to-cell and large-scale gene interaction networks could, for example, be represented by finite-state models or using agent-based simulation.

One should not jump to the conclusion, however, that systems and control theory could provide all answers to the challenges given by dynamic systems in molecular biology. In dynamic systems theory, one can often ignore spatial aspects. However, time and space are essential for explaining the physical reality of gene expression. The same component of a pathway may have a different functional role depending on its location within the cell. Although components of cells have specific locations, these locations lack exact coordinates. Not only signals are being transmitted but components also move around in a nonrandom fashion. Without spatial entailment there can be no living cell, and for systems biology it is thus necessary to integrate a topological representation of this organization with models of the dynamic behavior. The biologist Frank Harold presented an excellent discussion of the complexity of cellular processes and provided a compelling argument for the need for more research in complexity studies: From the chemistry of macromolecules and the reactions that they catalyze, little can be inferred regarding their articulation into physiological functions at the cellular level, and nothing whatever can be said regarding the form of development of these cells. It therefore seems to me self-evident that the quest for the nature of life cannot be conducted exclusively on the biochemists horizon. We must also inquire how molecules are organized into larger structures, how direction and function and form arise, and how parts are integrated into wholes [9]. In general, causation is a principle of explanation of change in the realm of matter. In systems biology causation is defined as a (mathematical) relationship, not between material objects (molecules), but between changes of states

within and between elements of a system. Instead of trying to identify genes as causal agents for some function, role, or change in phenotype we relate these observations to sequences of events. In other words, instead of looking for a gene that is the reason, explanation, or cause of some phenomenon we seek an explanation in the dynamics (sequences of events ordered by time) that led to it. It is systems dynamics, not a genetic program that gives rise to biological forms and function [9, p. 199]. In analyzing experimental data, we usually rely on assumptions made about the ensemble of samples. A statistical or average perspective, however, may hide short-term effects that are the cause for a whole sequence of events in a genetic pathway. What in statistical terms is considered an outlier may just be the phenomenon the biologist is looking for. It is very difficult to obtain sufficiently large and reliable data sets for pathway modeling; it is therefore important to compare different methodologies, their implicit assumptions, and the consequences of the biological questions asked. To allow reasoning in the presence of uncertainty, we have to be precise about uncertainty, and if we cannot be precise about uncertainty, modeling (generating hypotheses), and model validation (for hypothesis testing) become complementary aspects of an iterative process. It is thus of paramount importance that we strive to bridge the gap between data and models. In the words of Bertalanffy: Thus even supposedly unadulterated facts of observation already are interfused with all sorts of conceptual pictures, model concepts, theories, or whatever expression you choose. The choice is not whether to remain in the field of data or to theorize; the choice is only between models that are more or less abstract, generalized, near or more remote from direct observation, more or less suitable to represent observed phenomena [29].

Conclusions
Systems biology marks a shift away from an often obsessively reductionist (molecular) approach to providing a causal and dynamic account of cellular form and function. The biggest if not the principal hurdle for the systems approach is Zadehs uncertainty principle, which states that as the complexity of a system increases, our ability to make precise and yet significant statements about its behavior diminishes until a threshold is reached beyond which precision and significance (or relevance) become almost exclusive characteristics. The cell is a self-controlled and self-regulating dynamic system consisting of components that are interacting in space and time. The relationships that prevail between structure, function, and regulation in cellular networks are still largely unknown. Systems biology aims to identify and explain these relationships through an integrated effort of both experimental and theoretical methodologies. For this we require scientists who are prepared to invest time and effort into more than one discipline and scientific culture. This will necessitate a

August 2003

IEEE Control Systems Magazine

47

change in the education, training, and career prospects of interdisciplinary scientists. For those who persist, the reward is a better understanding of life itself.

References
[1] O.G. Vukmirovic et al., InsightFunctional genomics, Nature, vol. 405, no. 15, pp. 819-846, June 2000. [2] W. Weaver, Science and complexity, Amer. Scientist, vol. 36, pp. 536-544, Oct. 1948. [3] P. Bak, How Nature Works: The Science of Self-Organized Criticality. Oxford, U.K.: Oxford Univ. Press, 1997. [4] S. Kauffman, At Home in the Universe: The Search for Laws of Self-Organization and Complexity. Oxford, U.K.: Oxford Univ. Press, 1995. [5] B. Goodwin, How the Leopard Changed Its Spots: The Evolution of Complexity. Princeton, NJ: Princeton Univ. Press, 2001. [6] R.V. Sole and B.C. Goodwin, Signs of Life: How Complexity Pervades Biology. New York: Basic Books, 2002. [7] L.G. Harrison, Kinetic Theory of Living Pattern, Cambridge, U.K.: Cambridge Univ. Press, 1993. [8] H. Meinhardt, The Algorithmic Beauty of Sea Shells. Berlin, Germany: Springer-Verlag, 1988. [9] F.M. Harold, The Way of the Cell: Molecules, Organisms and the Order of Life. Oxford, U.K.: Oxford Univ. Press, 2001. [10] H. Kitano, Ed., Foundations of Systems Biology. Cambridge, MA: MIT Press, 2000. [11] H. Kitano, Systems biology: A brief overview, Science, vol. 295, no. 5, pp. 1662-1664, Mar. 2002. [12] O. Wolkenhauer, Systems biology: The reincarnation of systems theory applied in biology?, Briefings Bioinform., vol. 2, no. 3, pp. 258-270, 2001. [13] A. Kremling, K. Jahreis, J.W. Lengeler, and E.D. Gilles, The organization of metabolic reaction networks: A signal-oriented approach to cellular models, Metabolic Eng., vol. 2, no. 3, pp. 190-200, July 2000. [14] Caltech ERATO Kitano, Systems Biology Workbench, [Online]. http://www.cds.caltech.edu/erato/ [15] J. Downward, The ins and outs of signaling, Nature, vol. 411, no. 14, pp. 759-762, June 2001. [16] J. Hasty, D. McMillen, F. Isaacs, and J.J. Collins, Computational studies of gene regulatory networks: In numero molecular biology, Nature Reviews Genetics, vol. 2, no 4, pp. 268-279, Apr. 2001. [17] J.J. Tyson and M.C. Mackey, Molecular, metabolic, and genetic control, Chaos, vol. 11, no. 1, pp. 81-282, Mar. 2001. [18] P. Smolen, D.A. Baxter, and J.H. Byrne, Modeling transcriptional control in gene networksMethods, recent results, and future directions, Bull. Math. Biol., vol. 62, pp. 247-292, 2000. [19] K. Yeung, P. Janosch, D.W. Rose, H. Mischak, J.M. Sedivy, and W. Kolch, Mechanism of suppression of the Raf/MEK/Extracellular Signal-Regulated Kinase pathway by the Raf Kinase Inhibitor Protein, Mol. Cell. Biol., vol. 20, pp. 3079-3085, May 2000. [20] K. Yeung, T. Seitz, S. Li, P. Janosch, B. McFerran, C. Kaiser, F. Fee, K.D. Katsanakis, D.W. Rose, H. Mischak, J.M. Sedivy, and W. Kolch, Suppression of Raf-1 kinase activity and MAP kinase signaling by RKIP, Nature, vol. 401, pp. 173-177, Sept. 1999. [21] C.J. Marshall, Specificity of receptor tyrosine kinase signaling: Transient versus sustained extracellular signal-regulated kinase activation, Cell, vol. 80, pp. 179-185, Jan. 1995. [22] K.-H. Cho, S.-Y. Shin, H.-W. Kim, O. Wolkenhauer, B. McFerran, and W. Kolch, Mathematical Modeling of the Influence of RKIP on the ERK Signaling Pathway (Lecture Notes in Computer Science). Milan, Italy: Springer-Verlag, 2003. [23] D.P. Robert and M. Tom, Kinetic modeling approaches to in vivo imaging, Nature Rev.: Molecular Cell Biol., vol. 2, pp. 898-907, Dec. 2001. [24] K.-H. Cho, S.-Y. Shin, H.-W. Lee, and O. Wolkenhauer, Investigations into the analysis and modeling of the TNF mediated NF- B signaling pathway, in Proc. 3rd Int. Conf. Systems Biology, Stockholm, Sweden, p. 87, 2002 and Genome Research (Special Issue on Systems Biology), to be published. [25] H.G. Bock, Numerical treatment of inverse problems in chemical reaction kinetics, in Modeling Chemical Reaction Systems, K.H. Ebert, Ed. Springer-Verlag: New York, vol. 18, pp. 102-125, 1981.

[26] R. Hegger, H. Kantz, F. Schmuser, M. Diestelhorst, R.P. Kapsch, and H. Beige, Dynamical properties of a ferroelectric capacitor observed through nonlinear time series analysis, Chaos, vol. 8, pp. 727-736, 1998. [27] H. Kitano, Computational systems biology, Nature, vol. 420, pp. 206-210, 14 Nov. 2002. [28] M. Hucka, A. Finney, H. Sauro, H. Bolouri, J. Doyle, and H. Kitano, The ERATO systems biology workbench: An integrated environment for multiscale and multitheoretic simulations in systems biology, in Foundations of Systems Biology, H. Kitano, Ed. Cambridge: MIT Press, 2001, chap. 6, pp. 125-143. [29] L. Bertalanffy, General Systems Theory. New York: Braziller, 1969.

Olaf Wolkenhauer received the Dipl.-Ing. and BEng. degree in control engineering in 1994 from the University of Applied Sciences, Hamburg, Germany and the University of Portsmouth, U.K., respectively. He received the Ph.D. from the University of Manchester Institute of Science and Technology (UMIST). From 1997 to 2002 he held a research lectureship at the Control Systems Centre, UMIST, and since 2002 he has held a joint senior lectureship between the Department of Biomolecular Sciences and the Department of Electrical Engineering and Electronics. In 1999 and 2000 he was a visiting research fellow at Delft University of Technology, The Netherlands. He has authored two books, Possibility Theory with Applications to Data Analysis (RSP, 1998) and Data Engineering (Wiley, 2001). His research interest is in the application of systems and control methodologies to molecular and cell biology. Since July 2003 he has held the chair in Bioinformatics and Systems Biology at the University of Rostock. He can be contacted at the Department of Biomolecular Sciences and Department of Electrical Engineering and Electronics, Control Systems Centre, University of Manchester Institute of Science and Technology , P.O. Box 88, Manchester M60 1QD, U.K., o.wolkenhauer@umist.ac.uk. Hiroaki Kitano is a director at Sony Computer Science Laboratories, Inc., director of ERATO Kitano Symbiotic Systems Project, JST, and president of The Systems Biology Institute. He received his B.A. in physics from International Christian University, Tokyo, in 1984 and a Ph.D. in computer science from Kyoto University in 1991. He received the Computers and Thought Award in 1993 and the Prix Ars Electronica in 2000. Kwang-Hyun Cho received his B.S., M.S., and Ph.D. degrees in electrical engineering from the Korea Advanced Institute of Science and Technology in 1993, 1995, and 1998, respectively. He joined the School of Electrical Engineering at the University of Ulsan, Korea, in 1999 as an assistant professor. From 2002 to 2003, he was a research fellow at the Control Systems Centre, Department of Electrical Engineering and Electronics at the University of Manchester Institute of Science and Technology, U.K. His research interests cover the areas of systems science and control engineering including analysis and supervisory control of discrete event systems, nonlinear dynamics, hybrid systems, and applications to complex systems such as communication networks and biological systems.

48

IEEE Control Systems Magazine

August 2003

You might also like