Colloquium

Next talks

14 December, 9-11 a.m., Building 25, Room 201 (Carnot Building)
Joseph X. Zhou (Institute for Biocomplexity and Informatics, University of Calgary, Canada)
How to cure cancer without killing the tumour cells? -
Genetic mechanism of bidirectional spontaneous transition and population equilibrium of Breast cancer cells

Recent Talks

Dr. Stephan M. Feller, Biological Systems Architecture Group, Department of Oncology,
Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford University
Complexities of human cancer cell signalling -
Molecular diversity of tumours and signal processing by large protein complexes

Abstract: The heterogeneity of molecular lesions in tumours is arguably the greatest challenge in developing more effective drug therapies for solid malignancies. High-throughput sequencing of tumour DNAs is now a well-established method, but only a small percentage of the resulting data can be turned into accessible information, so many of the key players that drive tumour development in individual tumours remain unknown. Many new tumour drugs coming into clinical testing act on signalling proteins, which are thus of particular interest. Proteomic, biochemical and in particular functional studies are needed to learn more about these proteins and their behaviours. While this is still all but impossible with tumour biopsy material, large panels of tumour cell lines derived from a single tissue type are useful to gain some mechanistic insights into the highly diverse signalling networks of cancer cells. First results of our studies focused on tyrosine kinases and their signalling mediators in the world's largest panel of colorectal cancer lines will be presented.

Most intracellular signalling processes occur in large multi-protein complexes within a highly crowded environment (endogenous protein concentrations are ca. 200 mg/ml protein). Therefore, the free diffusion of signalling components that are part of these complexes is probably highly restricted. Large multi-site docking (LMD) proteins, like the members of the Gab, IRS/Dok, p130Cas and FRS families, serve as assembly platforms for signalling protein complexes, built from many components through specific protein-protein interactions. These complexes are believed to function as molecular 'computing units' that integrate multiple signalling inputs to produce several well-coordinated biological outputs (cell survival, proliferation, cell shape changes, migration, invasion etc.). Somewhat surprisingly, the LMD proteins appear to be mostly intrinsically disordered, that is lacking classical structural elements (alpha-helices, beta-sheets) and folded domains, according to current structural prediction programs. Based upon these observations, as well as important experimental data from the collaborating Schaper group and from our own work we have developed a model for co-translational 'compaction' of LMD proteins (the 'N-terminal folding nucleation hypothesis'), which would seem to explain how molecular computation of multiple intracellular signals can be mechanistically achieved. The predictions from this model can be experimentally tested, as will be discussed in the lecture.


Mirela Domijan, Warwick Systems Biology Centre at University of Warwick
Some observations on interaction graphs of mass-action reaction networks

Abstract: Recently, there has been growing interest in using graphical methods to analyse behaviour of reaction networks that are described by systems of ordinary differential equations (ODEs). Graphs have an enviable advantage that they can be used to study models of large size and with parameter uncertainty. In this talk I will focus on the interaction graph; a graph that is defined by the signs of the Jacobian matrix entries. Its structures such as signed circuits and the nucleus (or Hamiltonian hooping) have been linked to a variety of network dynamics. I will talk about some of our recent observations* about these structures and showcase how they further, or in some cases do not further, our understanding of the underlying dynamics.

Domijan M and Pecou E. On the interaction graphs of mass-action reaction networks, in review for J. Math. Biol., 2011.


Prof. Nadav Skjøndal-Bar (Department Chemical Engineering, NTNU, Trondheim, Norway)
Gastrin regulated transcriptional network : Network Component Analysis

Abstract: Gastrin is a hormone, mainly produced by G-cells in response to food, and is under feedback regulation. Gastrin activates and stimulates processes in the stomach and gut, including growth of specific cells in these tissues. Wrong regulation of Gastrin can result for example in cancer tumours. The study of processes involved Gastrin can assist in a solution for many cancer types. We aim at understanding the topology, the dynamics and the mechanisms of the gastrin network using a large microarray dataset we generated. Many decomposition methods are available to extract hidden regulatory signals from the high-throughput data sets, such as Principal component analysis (PCA), Independent component analysis (ICA), Singular value decomposition method (SVD). These methods can reduce the dimensionality of data but fails to extract the biologically significant information. Network Component Analysis (NCA) makes use of available information on the system from experiments and the extracted signals are biologically significant. Using the NCA we manage the estimate kinetic data of many TF's involved in the process and characterize them according to early-late response in gene expression. Through an extension of the NCA method, we manage to constract a network topology of the key TFs in the process.


Prof. Martha Grover (Georgia Institute of Technology, Atlanta, USA)
Control of Macromolecular Assembly

Abstract: Atoms and molecules interact dynamically via local forces, and these interaction rules can be manipulated through macroscopic system inputs such as temperature, pressure, and electric field. Moreover, the nature of the individual molecules can be designed to achieve a desired macromolecular assembly of the entire system. Biological systems achieve great complexity and robustness via this bottom-up molecular self-assembly, although human-designed systems are usually manufactured with a top-down approach.

The molecular structure of a material strongly impacts its mechanical, electrical, and optical properties, and ultimately the performance of the system in which it is incorporated. Often a perfectly ordered crystalline structure is desired, but defects reduce performance from this ideal case. In other systems the intended material structure is amorphous, but the details of the nanocrystalline ordering and molecular orientation strongly impact the material properties. A material may exist in its thermodynamic equilibrium structure, but often materials are locked into non-equilibrium meta-stable configurations during their processing. Even though the perfect crystalline state may be the thermodynamic equilibrium, the dynamics of nucleation and growth of crystalline domains during temperature annealing may create distinct domains that intersect at grain boundaries. Dislocations and vacancies may also be locked in during processing. Non-equilibrium structures vastly increase the space of possible structures, and these dynamics can be intentionally exploited to achieve novel properties via time-varying process inputs.

Stochastic simulations provide a quantitative framework in which to predict the overall dynamic organization of millions of atoms, based on local pair-wise interactions between individual atoms or small molecules. The events included in these kinetic Monte Carlo simulations may be selected using first-principles calculations, experimental measurements, or a combination of both. Two case studies will be described in this presentation. The first example focuses on surface morphology evolution during thin film deposition. Here the stochastic simulations provide the starting point to derive reduced-order coarse-grained models, which are subsequently used in a dynamic optimization to control the deposition process. In the second case study, the design of the local interaction rules between particles is considered, such that a desired assembly can be achieved in minimum time. Markov chain theory is employed to bound the convergence rate of the assembly process.


Martin Schuster, Ph.D., Associate Professor (Department of Microbiology, Oregon State University, USA)
Cooperation and cheating in bacterial quorum sensing

Abstract


PD Dr.-Ing. Niels Grabe (Universitätsklinikum Heidelberg)
Towards Modelling of Epidermal Skin Homeostasis

Abstract


Dr. Carito Guziolowski (Universitätsklinikum Heidelberg)
Analysis of large-scale biological networks with constraint based approaches

Abstract


Dr. Hans-Michael Kaltenbach (Computational Systems Biology Group at ETH Zürich)
Monotone decomposition of biochemical reaction networks


Joseph X. Zhou (Institute for Biocomplexity and Informatics, University of Calgary, Canada)
Cancer Attractor: A Systems Biology Understanding And Differentiation Therapy Of Cancer

Abstract: Based on clone evolution theory of cancer, cancer will be initiated if a cell receives multiple-hit mutations in at least five biological pathways of cancer hallmarks. However, the discovery of cancer stem cells (CSC) seriously challenges this view. It is proved, by xenograft experiments in both Leukemia and solid cancers, that cancer cells are, like other cells[1], intrinsically heterogeneous and only small portion of them can regenerate all cancer cell lineages. If cancer cells need multiple-hit oncogenetic mutations to obtain stemness with higher proliferation rate and evolutional advantages over normal cells, why large number of cancer cells do not have the infinite self-renew ability? To solve this puzzle, we propose a cancer attractor hypothesis based on theory of complex gene regulatory networks (GRN) that control cell fates and that the CSC is trapped in a pathological cancer attractor state of GRN. Multiple-hit oncogenetic mutations only make cells to lose the stability of development hierarchy (stem cells->progenitor cells->differentiated cells) and gain more cellular plasticity. These cells do not automatically gain the stemness. They need further intercellular signalling and environmental cues to enter the cancer stem cell state as a meta-stable cancer attractor in GRN[2,3]. This leads to a new approach to cancer drug development that departs from the existing paradigm of identifying critical cancer or cancer stem cell specific pathways and blocking them with target-selective drugs. Instead, we try to find small drug molecules which destroy the stemness of cancer stem cell and decrease the propensity of cancer cells to enter the CSC state. Simply put, instead of using target-selective drugs which may apply selection pressure upon cancer and trigger drug resistance, we use a differentiation therapy to cure cancer by reprogramming CSCs to non-dividing cells with much less selection pressure[4,5]. In past five years, with a new multi-step combinatorial screening scheme of FDA-approved small drug molecule library (JHCLL) on breast cancer line MCF7, we have already found 16 non-cytotoxic drug molecules which have a high efficiency of reprogramming MCF7 to non-dividing states. Our new paradigm for cancer drug discovery and preliminary results are present here.

1. Chang HH, Hemberg M, Barahona M, Ingber DE, Huang S (2008) Transcriptome-wide noise controls lineage choice in mammalian progenitor cells. Nature 453: 544-547. doi:10.1038/nature06965
2. Huang S, Ernberg I, Kauffman S (2009) Cancer attractors: a systems view of tumors from a gene network dynamics and developmental perspective. Seminars in cell & developmental biology 20: 869-76. Available here. Accessed 16 Jul 2010.
3. Huang S, Ingber DE (2007) A non-genetic basis for cancer progression and metastasis: self-organizing attractors in cell regulatory networks. Breast disease 26: 27-54. Available here.
4. S. Huang (2011) Systems biology of stem cells : three useful perspectives to help overcome the paradigm of linear pathways. Philosophical Transactions of the Royal Society B Biological Sciences Transactions Society, Royal. doi:10.1098/rstb.2011.0008
5. J. X. Zhou (2011) Understanding gene circuits at cell-fate branch points for rational cell reprogramming . Trends in Genetics. Available here.


Prof. Dr. Fred Schaper (Chair for Systems Biology, University of Magdeburg)
More than JAKs and STATs: New insights into the regulation of interleukin-6 signal transduction


Abstract: Inflammation is a physiological response of the organism to cope with infections by microrganisms and chemical, physical or thermal induced traumata. Interleukin-6 (IL-6) is a crucial cytokine which regulates the outcome of inflammatory events. Disregulation of IL-6 signal transduction is associated with many inflammatory as well as proliferative diseases.

IL-6 activates the JAK/STAT pathway but also the MAPK- and PI3K-cascade. In the past major focus was on IL-6-induced STAT3 activation and its negative regulation. Recently, it became obvious that also IL-6-induced MAPK activation is crucial for promoter activation of a set of IL-6-dependent genes.

An overview on IL-6 signal transduction will be presented. A special focus will be on the negative regulation and the balance of the pathways initiated by IL-6. Additionally, cross-regulation of IL-6 signalling by other hormones will be discussed.


Dr. Ronny Straube (Max Planck Institute Magdeburg)
Bistable Network Motifs in Signal Transduction Networks


Dr. Jörg Schaber (University of Magdeburg)
Automated ensemble modeling with modelMaGe: technology and application


Dr. Utz-Uwe Haus (University of Magdeburg)
Discovering All Associations in Discrete Data Using Frequent Minimally Infrequent Attribute Sets

Abstract: Associating biological categories with measured or observed attributes is a central challenge for discrete mathematics in life sciences. We propose a new concept to formalize this question: Given a binary matrix of objects and attributes, determine all attribute sets characterizing object sets of cardinality t1 that do not characterize any object set of size t2 > t1. We determine how many such attribute sets exist, give an output-sensitive quasi-polynomial time algorithm to determine them, and show that k-sum matrix decompositions known from matroid theory are compatible with the characterization. (joint work with Dr. Elke Eisenschmidt, University of Magdeburg)


Prof. Stefan Schuster, Department of Bioinformatics, Faculty of Biology and Pharmacy, Friedrich Schiller University Jena
Theoretical Systems Biology in Biotechnology: Promises, success stories and limitations

Abstract


Dr. Steffen Waldherr, Institut for Systems Theorie and Automatic Control, Universität Stuttgart
Parametric uncertainty analysis of biochemical signal transduction models

Abstract: Dynamical models of biochemical signal transduction pathways are often affected by large uncertainties on the parameter values. Robustness analysis is an efficient tool to quantify the effects of model uncertainty on qualitative properties of the model. This talk addresses in particular the robustness analysis problem for qualitative dynamical behaviour in the pathway, such as sustained oscillations or bistability. The level of parametric uncertainty not affecting the dynamical behaviour is thereby quantified by a suitably defined robustness measure. In the robustness analysis of biochemical networks, there are several challenges which obstruct the direct application of control engineering methods to this problem. These challenges include nonlinearity of the equations, dependence of the steady state on uncertain parameters, and the need to consider a nominally unstable system. This talk presents a novel solution to the robustness analysis problem, overcoming the mentioned challenges within a control engineering point of view. To this end, parameter values yielding a change in the dynamical behaviour are characterised via a feedback loop breaking approach. Based on this approach, two methods are proposed: one to compute robustness certificates, yielding a lower bound on the robustness measure, and one to search for nearby bifurcations, yielding an upper bound. To illustrate the proposed methods, an analysis of the NF-kB pathway is presented. This pathway is a central player in the mammalian immune system and of high biomedical relevance. The uncertainty analysis yields novel biological insights into the oscillatory behaviour of this pathway.


Björn Heynisch, University of Magdeburg
Influence of Host Cell Defense during Influenza Vaccine Production in MDCK Cells

Abstract: For cell culture-based influenza vaccine production virus yield optimization is of crucial importance. In particular, with the recent threat of the new H1N1 pandemic, not only seasonal vaccines but also pre-/pandemic vaccines have to be supplied in large quantities. In vivo influenza replication is limited by the immune system, but for production cell lines the impact of cellular defense mechanisms on virus yield is unknown. In influenza-infected adherent Madin-Darby canine kidney (MDCK) cells the interferon (IFN) response and subsequent induction of the antiviral state was monitored. Virus yield and host cell signaling intensity were strain-dependent. By over-expression of viral antagonists IFN-signaling could be reduced up to 90%. However, maximum virus titer determined by real-time PCR and HA-assay was not altered significantly. Stimulation of the antiviral state by conditioned medium led to enhanced IFN-signaling, which initially slowed down virus replication but had only minor effects on final virus titers. Interestingly, minireplicon assays revealed that canine Mx proteins are lacking the antiviral activity against influenza of their human or mouse counterparts. In summary, for MDCK cell culture-based influenza virus production host cell defense mechanisms seem to play only a minor role for final virus yields. Antiviral mechanisms of these epithelial cells may slow down influenza replication, which in vivo gains time for the immune system to be activated, but do not reduce maximum virus titers obtained in the bioprocess.


Dr. Birgit Schöberl, Merrimack Pharmaceuticals (Cambridge, US)
Applying engineering principles to the development of novel cancer therapies

Abstract: Combining quantitative biology and computational modeling provides a powerful toolkit to design novel therapies in a context dependent manner. We will provide multiple examples where we translated the insights gained from modeling and simulation into practice by engineering and testing novel, antibody-based therapeutics in the context of the computer simulations. Through this iterative process between computational modeling and antibody engineering, we gain a deeper understanding of the drug's mechanism of action which allows us to design therapeutics with a specific tumor type in mind. This context specific design can subsequently be translated into the clinic.


Dr. Joseph Xu Zhou, Centre for Information Services and High Performance Computing (ZIH), TU Dresden
Predicting pancreas cell fate decisions and reprogramming with a gene regulatory network model


Prof. Friedrich Srienc, Department of Chemical Engineering and Materials Science, and BioTechnology Institute, University of Minnesota, Minneapolis/St.Paul, MN
From Metabolic Pathway Analysis to a Theory of Evolution

Abstract: Complex metabolic networks can be decomposed into a set of discrete fundamental pathways or elementary modes that support cell function under the constraint of mass conservation. The identification of these modes enables the analysis of the pathway capabilities and the rational design of efficient networks. This analysis approach is of immense value in biotechnology as metabolic networks can be engineered on a completely rational basis. The analysis predictions have been confirmed in several experimental systems related to biofuels production and to the production of secondary metabolites. Moreover, the set of discrete elementary modes in a network can be interpreted with the tools of statistical thermodynamics. The usage probabilities of individual elementary modes are expected to be distributed according to Boltzmann's distribution law such that the rate of entropy production is maximized. Adaptive evolution experiments support the idea that metabolic networks evolve towards such state. Ultimately, evolution of metabolic networks appears to be driven by forces that can be quantified by the distance of the current metabolic state from the state of maximum entropy production that represents the unbiased, most probable selection of fundamental pathway choices.


Prof. David Fell, School of Life Sciences, Oxford Brookes University
Building and analyzing genome-scale metabolic networks

Abstract: My group has been building structural metabolic models from annotated genome sequences for bacteria, such as Saccharopolyspora erythrea and Streptococcus agalactiae, and a plant, Arabidopsis thaliana, and further models are under construction. By a structural metabolic model, we mean a list of connected, stoichiometrically-balanced reactions catalyzed by the complement of enzymes considered to be encoded in the genome. Analysis of the model then consists of mathematical and computational operations on the stoichiometry matrix derived from the reaction list, with the assumption that all internal metabolites are at steady state as nutrients are converted into end products such as biomass. The techniques include, but are not limited to, linear programming (also referred to as Flux Balance Analysis in this context).

In fact, analysis cannot be separated from model construction, since errors, uncertainties and incompatibilities in the information used to construct the models mean that there are many problems to identify and correct during model construction. These slow down network construction, so part of our research has been to develop methods that can assist in identifying the sources of errors in large networks.

Much analysis of genome-scale metabolic networks has revolved around optimizing the rate of biomass production (or yield) and preicting the effect of mutations. We have been concerned to develop other analyses that provide insight into properties and potential behaviours of the network, and this will be illustrated with reults from our Arabidopsis model.


Prof. Jörg Stelling, ETH Zürich
Computational Engineering of Synthetic Genetic Circuits

Abstract: Ultimately, synthetic biology aims at establishing novel, useful biological functions by suitably combining well-characterized parts. Especially when complex circuits -- in terms of the number of components and interactions involved, or with respect to the dynamic behavior -- are to be designed, computational engineering methods have to be an integral part of the approach.

This talk will focus on engineering concepts to achieve scalability and robustness (relative insensitivity to external or internal perturbations of the designed circuits). Both are important concerns for the field because the biology-based parts employed are not (yet) well-characterized, the circuits have to operate in a noisy (cellular) environment, and they cannot be completely isolated from, e.g., a cellular context.

More specifically, major open issues exist regarding (i) principles of circuits design with standardized parts, and (ii) principles for the design of robust performance of synthetic circuits. 'Classical' synthetic genetic circuits as well as novel systems such as a tunable synthetic oscillator in mammalian cells will be discussed as prototypical examples to illustrate our current capabilities. In perspective, synthetic approaches do not only have the potential of major impacts in different application areas, but also present challenging problems for engineering design.


Dr. Verena Wolf
Universität Saarbrücken

Stochastic Modelling of Biochemical Networks


Dr. Lars Küpfer
Systems Biology and Computational Solutions
Bayer Technology Services GmbH, Leverkusen
Application of integrated multi-scale models to pharmaceutical research and development

Abstract: Computational models play an increasing role in pharmaceutical research and development, since they offer an efficient way for storing, representing and analyzing experimental data at each stage of (pre-)clinical development. Physiologically-based pharmacokinetic (PBPK) models are a special form of pharmacokinetic models which represent the processes underlying the distribution of a substance within the human body mechanistically at a high level of detail. Based on generic drug distribution models and extensive collections of physiological parameters they thus enable a comprehensive simulation of drug pharmacokinetics at the whole-body scale. Moreover, structural refinements such as metabolization processes or active transport can easily be introduced into the basic PBPK models such that structural hypotheses can be evaluated. Using different exemplary case studies we show here how computational models allow the investigation of mechanisms governing a specific pharmacokinetic behavior. Simultaneous consideration of a drug's modes of action at the cellular level enables construction of integrated multi-scale models based on which dose-effect relationships can be investigated from a systems perspective. Such multi-scale models have amongst others been used for the correlation of genetic predisposition in a patient subgroup with clinical endpoints. Computational models thereby significantly support assessment of crucial points in drug R&D such as (1) a rigorous evaluation of drug efficacy at an early phase of clinical development, (2) avoidance of adverse effects and (3) development of individualized therapeutic designs.

Biographical Sketch: Lars Kuepfer is a scientist in the group "Systems Biology and Computational Solutions" of Bayer Technology Services GmbH in Leverkusen. He studied chemical engineering at the TH Karlsruhe, RWTH Aachen and Carnegie Mellon University, Pittsburgh, and received his Ph.D. degree from ETH Zurich. In his thesis, Lars Kuepfer analyzed metabolic and regulatory principles in yeast based on computational models. In his current position, he works on pharmacokinetic and pharmacodynamic modeling of novel drug candidates thereby supporting decision making along the pharma development process. Lars Kuepfer's main research interests are in the area of pharmacogenomics, metabolic modeling and system identification.