Available online at www.sciencedirect.

com

Biochemical network-based drug-target prediction

Edda Klipp1, Rebecca C Wade2 and Ursula Kummer3

The use of networks to aid the drug discovery process is a prediction of drug targets using network information
rather new but booming endeavor. A vast variety of different (Figure 1). We first give a brief overview of some of
types of networks are being constructed and analyzed for the key advances in using network analysis for drug-target
various different tasks in drug discovery. The analysis may be at identification. We then discuss the application of ‘classi-
the level of establishing connectivity, topology, and graphs, or cal’ systems biology mathematical modeling approaches
may go to a more quantitative level. We discuss here how to study signaling and metabolic networks for drug-target
computational systems biology approaches can aid the prediction. These approaches to the computational pre-
quantitative analysis of biochemical networks for drug-target diction of drug targets are increasingly supported by
prediction. We focus on networks and pathways in which the domain-specific computational tools (see Box 1) and
components are related by physical interactions or biochemical databases (see Box 2).
processes. We particularly discuss the potential of
mathematical modeling to aid the analysis of proteins for Target prediction using homogenous and
druggability. heterogeneous protein interaction networks
Addresses Recent experimental advances in high-throughput pro-
1
Humboldt-Universität zu Berlin, Germany teomics have led to a wealth of data on protein–protein
2
HITS, Heidelberg, Germany interactions. Despite the many inconsistencies in these
3
University of Heidelberg, Germany
data [16], there is great interest in mining it for drug-
Corresponding author: Klipp, Edda ([email protected]) target identification. Analyses show that known human
drug targets tend to occur at middle-degree to low-degree
nodes, that is, less connected nodes [17]. Targeting these
Current Opinion in Biotechnology 2010, 21:511–516 can be expected to result in drugs with fewer side effects
This review comes from a themed issue on (and greater synergetic efficacy in a drug made of several
Systems biology compounds). Thus, ranking of proteins by the topological
Edited by Vitor Martins dos Santos and Jiri Damborsky properties of the human protein–protein interaction net-
work is one strategy for drug-target identification [18].
Available online 14th June 2010
Another approach is to characterize the interaction prop-
0958-1669/$ – see front matter erties in protein–protein complexes, for example, by
# 2010 Elsevier Ltd. All rights reserved. identifying the domains involved in binding or by analyz-
ing the 3D structure. Comparison of domain–domain
DOI 10.1016/j.copbio.2010.05.004
interactions and interfaces across an interactome can
guide the identification of selective drug targets or drugs
targeting multiple proteins (to block parallel pathways in
Introduction a network) [19]. Structural analysis can be carried out to
Systems biology has changed the paradigm for drug-target identify pockets where drugs could bind and compare
identification from considering the roles of individual their properties with binding pockets on other proteins in
genes or proteins in diseases to studying the structure, the network [20].
properties, and behavior of various types of biochemical
networks and their changes in perturbed states such as Combining protein interaction data with other data in
cancer or apoptosis. This new approach opens the way to heterogeneous networks will provide a stronger basis for
predicting the effects of targeting distinct genes or drug-target prediction. Yildirim et al. [21] built a bipartite
proteins not only based on their individual properties, graph linking drugs and proteins and identifying topolo-
for example, binding constants and binding specificity gical features, for example, hubs in this network are target
profiles, but also based on their position and function in proteins to which many drugs bind, and differences in the
one or more networks. Such networks may comprise graph for etiological and palliative drugs. Mestres et al.
disease-relevant compounds or may cover all similar [22] combined data from 7 databases to assemble a drug-
components of the cell such as all proteins or all genes, target network with 4767 unique interactions between
or may be heterogeneous as regards compound/molecule 802 drugs and 480 targets (each drug has on average 6
type. New experimental technologies, data analysis strat- targets).
egies, and modeling approaches are being developed for
the prediction of medically relevant properties of these Such drug-target networks will increasingly be comple-
networks, most notably useful and effective drug targets. mented by and integrated with networks containing other
Here, we will focus on computational approaches to the biological and medical data (see review [23]), as well as

www.sciencedirect.com Current Opinion in Biotechnology 2010, 21:511–516

512 Systems biology

Figure 1

Illustration of different levels of computational network approaches. Protein–protein interaction networks (left) represent physical or other types of
interactions of virtually all proteins in a cell and provide means to detect candidate targets, for example hubs. Signaling networks (middle) are specific
types of protein networks involved in the transmission of information, for example, about external growth factors. Their dynamics can be described by
differential equations (right), which can be used to compute suitable targets.

drug–drug networks. The latter can be built according to dependence of separately measured data such as external
chemical similarity [22,24] or from phenotypic side- stimulation, phosphorylation states, localization, protein
effect similarities [25] and can help to identify new interactions, and transcriptional activation, for example
therapeutic applications for known drugs. [33]. Signaling pathway models have been used to explain
dynamic features such as the effect of positive or negative
Pathway analysis is an approach to convert network data feedback, for example [34] or crosstalk between different
into models (and defining local networks) and its appli- pathways [35]. On top, the modeling provides a frame-
cation in drug discovery is described, for example, in work for hypothesis generation and for prediction of the
Yuryev [26]. effects of intervention.

Target prediction in signaling networks Drug-target prediction by in silico systems biology needs
Cells receive information about growth factors, nutrients, sufficiently well described networks including the network
toxic compounds, and other external changes via so-called structure and compound interaction properties. Then,
signaling pathways. The activated membrane-located re- mathematical models can be useful to predict targets for
ceptor induces a cascade of protein interactions and treatment and test the outcome of different target pos-
modifications, which eventually not only regulates tran- itions, treatment strengths, target combinations, or
scription factors and, hence, gene expression, but may temporal combination scenarios [7,36]. Having a sound
also have side effects on cell cycle and metabolism. mathematical model at hand, one can also address a num-
Within signaling pathways, information is not only line- ber of relevant questions through simulation, such as the
arly transmitted, but also processed through signal integ- effect of parameter variability (mimicking biological varia-
ration, cross-activation, positive, or negative feedback. bility) on the prediction quality, the impact of parallel
Experimental and computational systems biology have pathways bypassing the target, or saturation of the target.
intensively studied a number of human signaling path-
ways, among them the EGFR [27], Wnt [28], Jak/STAT With respect to successful application, this field is only in
[29], TGFb [30,31], and NFkB pathways [32]. its early phase since there is still a lack of sufficiently well
described networks. The major obstacle is – despite
The models are either formulated in a qualitative way massive gathering of data in all omics – to obtain data
describing how activity states of upstream compounds of the right type, amount, coverage, and quality that
influence activities of downstream compounds, or as sets covers the pathway dynamics and that is sufficient to
of ordinary differential equations (ODEs) describing the determine the model parameters. For example, high-
temporal evolution of the involved components and their throughput techniques such as microarrays or deep
activity. Models are usually formulated to solve a question sequencing are not of much help, comprehensive phos-
that cannot be answered by gathering experimental data phoproteomic data are, if we want to understand the
alone. They help to understand the network architecture protein modification dynamics in the first minutes or
and the observed dynamics and to rationalize the inter- hours after addition of growth factor.

Current Opinion in Biotechnology 2010, 21:511–516 www.sciencedirect.com

 Biochemical network-based drug-target prediction Klipp, Wade and Kummer 513

Box 1 Computational tools for systems biology and target Box 2 Drug-target databases
prediction
Drugbank: http://www.drugbank.ca/, [8,9]
CellDesigner: http://www.celldesigner.org, [1]
Includes manually annotated main mode of action of drugs for over
CellDesigner is a structured diagram editor for drawing gene- 4000 drug-target pairs. Combines detailed drug data (e.g. chemical,
regulatory and biochemical networks. Networks are drawn based on pharmacological, pharmaceutical) with comprehensive target infor-
the process diagram, with graphical notation system proposed by mation (e.g. on sequence, structure, pathway). It can be searched by
Kitano, and are stored using the Systems Biology Markup Language pathway.
(SBML), a standard for representing models of biochemical and
Matador: Manually Annotated Targets and Drugs Online Resource
gene-regulatory networks [2].
http://matador.embl.de, [10]
COPASI: COmplex PAthway SImulator
Includes direct and indirect modes of action of drugs with targets
http://www.copasi.org, [3]
and can be queried with drug or target protein. The manually
COPASI is a software application for the simulation and analysis of annotated list of direct (binding) and indirect interactions between
biochemical networks and their dynamics. COPASI is a stand-alone proteins and chemicals was assembled by automated text-mining
program that supports models in the SBML standard format and can followed by manual curation.
simulate their behavior using ODEs or Gillespie’s stochastic simulation
Pdtd: Potential Drug Target Database
algorithm; arbitrary discrete events can be included in such simula-
tions. The software carries out several analyses of the network and its http://www.dddc.ac.cn/pdtd/, [11]
dynamics and has extensive support for parameter estimation and
Database providing potential drug targets of known 3D structure
optimization. COPASI provides the means to visualize data in
annotated by therapeutic area and associated diseases. Structures
customizable plots, histograms, and animations of network diagrams.
are from the Protein Data Bank (PDB).
iPATH: Interactive Pathways Explorer
STITCH: Chemical-Protein Interactions
http://pathways.embl.de/index.html, [4]
http://stitch.embl.de/, [12,13]
Web-based tool for displaying and analyzing metabolic pathways.
A resource to explore known and predicted interactions of chemicals
The pathways in specific organisms can be viewed and analyzed in
and proteins. Chemicals are linked to other chemicals and proteins
the context of a global pathways map.
by evidence derived from experiments, databases and the literature.
SABIO-RK: System for the Analysis of Biochemical Pathways - STITCH contains interactions for over 74,000 small molecules and
Reaction Kinetics over 2.5 million proteins in 630 organisms. Results are provided with
an annotated graphical view of the protein–protein, protein–chemical
http://sabio.h-its.org, [5]
and chemical–chemical interaction network, including display of
SABIO-RK is a web-based application employing the SABIO known molecular 3D structures.
relational database that contains information about biochemical
SuperTarget: http://bioinf-tomcat.charite.de/supertarget/, [10]
reactions, their kinetic equations with their parameters, and the
experimental conditions under which these parameters were Allows searching of a database of over 7000 drug-target pairs with
measured. It is able to export SBML format files of selected reactions querying by drug, target, pathway, ontology, or cytochrome P450.
sets together with kinetic information.
TDR: Targets Database
SYCAMORE: SYstems biology Computational Analysis and MOdel-
http://tdrtargets.org/, [14]
ing Research Environment
Identification and ranking of targets against neglected tropical
http://sycamore.h-its.org, [6]
diseases, part of a WHO project. Searches a genome to rank
SYCAMORE is a browser-based application that facilitates construc- potential targets according to about 10 different criteria relevant to
tion, simulation, and analysis of kinetic models in systems biology druggability. It is aimed at facilitating the identification and
projects. It provides tools for database supported modeling, basic prioritization of drug targets in pathogens causing neglected
model checking, and the estimation of unknown kinetic parameters diseases.
based on protein structures. In addition, it offers some guidance to
TTD: Therapeutic Targets Database
allow non-expert users to perform basic computational modeling
tasks. http://xin.cz3.nus.edu.sg/group/cjttd/ttd.asp, [15]
TIde: Target Identification A database that provides information about known therapeutic
protein and nucleic acid targets, the targeted disease, pathway
http://sysbio.molgen.mpg.de/tide, [7]
information, and the corresponding drugs directed at each of these
TIde is a tool for the automatic identification of optimal drug targets targets. It can be searched by giving the name of a biochemical
in kinetic models of biochemical networks based on ordinary pathway of interest and then will provide information on all known
differential equations (ODEs). Given a model in the standard Systems drug targets in the query pathway.
Biology Markup Language (SBML) format, it will identify promising
drug targets for different effective modifier concentrations. Target
combinations are also featured.

The relative effectiveness of eight ErbB ligands with

respect to their ability to stimulate phosphorylation of
However, there are already very promising examples certain downstream proteins was tested and for the most
among signaling pathway models such as the study of effective ligands the early signaling dynamics as function
the ErbB network with sensitivity analysis [37], which of time and ligand concentration were compared with
identified ErbB3 as a key node in response to ligands. respect to target phosphorylation. The efficacy was tested

www.sciencedirect.com Current Opinion in Biotechnology 2010, 21:511–516

514 Systems biology

in mice in vivo. Another example is Logical modeling, aiming at computing subsystems able to exhibit a steady
which assigns activation or inactivation properties to state and thus represent a potentially viable system (e.g.
protein interactions, which has led to comprehensive extreme currents, elementary modes) or they use some
description EGFR/ErbB signaling network enabling the objective function to compute the fluxes in order to achieve
prediction of the effects of various ligand stimulations [27]. an optimal outcome with respect to these boundary con-
ditions. The latter is called flux balance analysis and has
Boolean modeling (assigning values of 1 or 0 to active or been employed in drug-target identification [44].
inactive protein, respectively, which are updated accord-
ing to the states of their modifiers in temporal fashion) of With the knowledge of kinetic parameters and equations,
the ErbB receptor regulated G1/S transition has revealed more detailed and powerful kinetic models can be set up.
potential new and alternative targets in case of de novo As in the case of signaling systems, these are mostly
trastuzumab resistance in breast cancer [38]. formulated as ordinary differential equations and used
to evaluate the temporal behavior of the system of in-
Chronobiology studies various rhythms followed by living terest. Examples are the model of Prostaglandin H
beings, such as cell cycle, circadian, or annual rhythms. Synthase-1 (COX-1) developed to predict inhibition
Cell cycle models can be used like signaling pathway effects of nonsteroidal anti-inflammatory drugs [45] or a
models to predict drug effects, but also to predict the model for bacterial methicillin-resistance [46].
effect of, for example, apoptosis and DNA damage [39].
Also circadian rhythms are investigated, their components Such models can be used to, for example, analyze which
identified [40] and their dynamics described with math- points in the metabolism of a pathogen are most vulner-
ematical models [41]. Chronotherapy applies the finding able and thus, probably good drug targets. Thus, a
that cancer treatment has different effects if supplied at specialized framework for the modeling of Mycobacter-
different times of the day, both shown experimentally ium tuberculosis [47] has this aim. A systematic strategy
and with mathematical models [42,43]. to scan for vulnerable points is the performance of a
sensitivity analysis. In general, such an analysis investi-
Although a major aim of systems biology is to cover gates how strong the influence of specific parameters is on
comprehensive networks there are limits to this concept an observable, for example, how reaction rate of each
when it comes to manageable, predictive dynamic enzyme in a pathway affects the steady state concen-
models. Such a model must naturally have boundaries tration of a metabolite. A specialized kind of sensitivity
discriminating between processes considered relevant for analysis is the so-called metabolic control analysis. This
the investigated problem and ‘the rest’ of the cell. has been employed in investigating vulnerable points in
Although whole cell models are envisaged for the future, Trypanosoma [48].
they are currently not at hand owing to limited data
availability and owing to lack of methods for combining One problem for sensitivity analyses is the fact that these
all knowledge about cellular life into one approach. Con- are local methods, which means the result will be different
sequently, effects that are out of the scope of that model if several parameters in the system are changed. In the
cannot be predicted (e.g. a pure cell cycle model will not absence of complete knowledge about kinetic parameters,
predict metabolic effects of a protein kinase inhibitor, this calls for more global strategies. One such strategy,
even though they might exist). which can also be used for the purpose of drug-target
identification, has been suggested recently [49]. It employs
Target prediction in metabolic networks optimization strategies to find the global maximum or
Metabolism, the breaking down and synthesis of com- minimum of sensitivity in the complete parameter space.
pounds, has long been investigated by computational
approaches. This is probably owing to the fact that very All in all, as stated above, despite promising results, for
early on in biochemical research, biotechnological as well example, for the development of drugs against certain
as pharmaceutical interests were pursued by trying to parasites/pathogens, there are still not too many success
influence the metabolism of microorganisms and differ- stories of systems biology in drug-target identification in
ent eukaryotic cell types. In systems biology, these metabolism. Again, this is mostly owing to the fact that
approaches have been intensified. The complexity of the necessary data for the development of quantitative
metabolic networks in living cells asks for computational models are often lacking and new methods have to be
models that support the analysis and understanding of the developed that deal with systems with a lot of unknown
metabolic behavior of these cells. Different techniques parameters.
are employed for this purpose and are also specifically
used to aid drug-target identification: Summary and outlook
Taken together, the combination of experimental and
In the absence of kinetic parameters, so-called stoichio- computational approaches in systems biology has already
metric network analyses can be used. These are either revealed a number of promising examples of rational and

Current Opinion in Biotechnology 2010, 21:511–516 www.sciencedirect.com

 Biochemical network-based drug-target prediction Klipp, Wade and Kummer 515

network-based drug-target prediction. Further progress 6. Weidemann A, Richter S, Stein M, Sahle S, Gauges R,
Gabdoulline R, Surovtsova I, Semmelrock N, Besson B, Rojas I
will be based on precise and reproducible data and et al.: SYCAMORE—a systems biology computational analysis
mathematical descriptions to create predictive and help- and modeling research environment. Bioinformatics 2008,
24:1463-1464.
ful models. Increasingly, it will be possible to go beyond a
static view of networks to consider their dynamics and 7. Schulz M, Bakker BM, Klipp E: TIde: a software for the
systematic scanning of drug targets in kinetic network
aim for temporally structured drug administration to first models. BMC Bioinformatics 2009, 10:344.
change the sensitivity pattern in the network and then 8. Wishart DS, Knox C, Guo AC, Cheng D, Shrivastava S, Tzur D,
attack the most vulnerable points. Gautam B, Hassanali M: DrugBank: a knowledgebase for drugs,
drug actions and drug targets. Nucleic Acids Res 2008,
36:D901-906.
The necessity for a network-based approach to drug-
target prediction is clear when one considers that many 9. Wishart DS, Knox C, Guo AC, Shrivastava S, Hassanali M,
Stothard P, Chang Z, Woolsey J: DrugBank: a comprehensive
compounds are abandoned during drug development resource for in silico drug discovery and exploration. Nucleic
because they act on a target other than the intended Acids Res 2006, 34:D668-672.
target(s). Moreover, successful drug-target prediction 10. Gunther S, Kuhn M, Dunkel M, Campillos M, Senger C, Petsalaki E,
necessitates considering not only individual cells but also Ahmed J, Urdiales EG, Gewiess A, Jensen LJ et al.: SuperTarget
and Matador: resources for exploring drug-target
effects in the human body. Therefore, future studies relationships. Nucleic Acids Res 2008, 36:D919-922.
must include cell-type specific modeling and also inte- 11. Gao Z, Li H, Zhang H, Liu X, Kang L, Luo X, Zhu W, Chen K,
grate pharmacokinetics and pharmacodynamics. With the Wang X, Jiang H: PDTD: a web-accessible protein database for
increasing speed of high-throughput analyses and rising drug target identification. BMC Bioinformatics 2008, 9:104.
computer power, it will become possible to integrate 12. Kuhn M, Szklarczyk D, Franceschini A, Campillos M, von Mering C,
network-oriented research with specific, individualized Jensen LJ, Beyer A, Bork P: STITCH 2: an interaction network
database for small molecules and proteins. Nucleic Acids Res
information and thereby bring the dream of personalized 2010, 38:D552-556.
medicine within reach [50]. 13. Kuhn M, von Mering C, Campillos M, Jensen LJ, Bork P: STITCH:
interaction networks of chemicals and proteins. Nucleic Acids
Acknowledgements Res 2008, 36:D684-688.
EK acknowledges support by the European commission (project ENFIN, 14. Aguero F, Al-Lazikani B, Aslett M, Berriman M, Buckner FS,
LSHG-CT-2005-518254), the German Ministry of Education and Research Campbell RK, Carmona S, Carruthers IM, Chan AW, Chen F et al.:
(BMBF, projects ColoNet and Drug-iPS), and the German Research Genomic-scale prioritization of drug targets: the TDR Targets
Foundation (SFB 618). RCW gratefully acknowledges the support of the database. Nat Rev Drug Discov 2008, 7:900-907.
Klaus Tschira Foundation and the German Ministry of Education and
Research (BMBF, projects Hepatosys, Virtual Liver and BioRN). UK 15. Zhu F, Han B, Kumar P, Liu X, Ma X, Wei X, Huang L, Guo Y, Han L,
likewise thanks the Klaus Tschira Foundation, the EU (NoE BioSim), the Zheng C et al.: Update of TTD: therapeutic Target Database.
NIH, and the German Ministry of Education and Research (BMBF - Nucleic Acids Res 2010, 38:D787-791.
Hepatosys and ForSys) for funding.
16. Wodak SJ, Pu S, Vlasblom J, Seraphin B: Challenges and
rewards of interaction proteomics. Mol Cell Proteomics 2009,
References and recommended reading 8:3-18.
Papers of particular interest, published within the annual period of
review, have been highlighted as: 17. Hase T, Tanaka H, Suzuki Y, Nakagawa S, Kitano H: Structure of
 protein interaction networks and their implications on drug
design. PLoS Comput Biol 2009, 5:e1000550.
 of special interest Analysis of yeast and human protein interaction networks shows that they
 of outstanding interest are scale-rich and form a highly optimized tolerance network with
advantageous drug targets at middle-degree to low-degree nodes.

1. Kitano H, Funahashi A, Matsuoka Y, Oda K: Using process 18. Zhu M, Gao L, Li X, Liu Z, Xu C, Yan Y, Walker E, Jiang W, Su B,
diagrams for the graphical representation of biological Chen X et al.: The analysis of the drug-targets based on the
networks. Nat Biotechnol 2005, 23:961-966. topological properties in the human protein–protein
interaction network. J Drug Target 2009, 17:524-532.
2. Hucka M, Finney A, Sauro HM, Bolouri H, Doyle JC, Kitano H,
 Arkin AP, Bornstein BJ, Bray D, Cornish-Bowden A et al.: The 19. Keskin O, Gursoy A, Ma B, Nussinov R: Towards drugs targeting
systems biology markup language (SBML): a medium for multiple proteins in a systems biology approach. Curr Top Med
representation and exchange of biochemical network models. Chem 2007, 7:943-951.
Bioinformatics 2003, 19:524-531.
Using unambigous descriptions of models and of the underlying networks 20. Sugaya N, Ikeda K, Tashiro T, Takeda S, Otomo J, Ishida Y,
is a major necessity to make them comprehensible to a wider audience Shiratori A, Toyoda A, Noguchi H, Takeda T et al.: An integrative in
and to enable systematic work, exchange, and interpretation. Formula- silico approach for discovering candidates for drug-
tion of SBML as a standard for model encoding in computer-readable targetable protein–protein interactions in interactome data.
scripts was a first step in a long series of community efforts for common BMC Pharmacol 2007, 7:10.
standards in model description. 21. Yildirim MA, Goh KI, Cusick ME, Barabasi AL, Vidal M: Drug-
3. Hoops S, Sahle S, Gauges R, Lee C, Pahle J, Simus N, Singhal M, target network. Nat Biotechnol 2007, 25:1119-1126.
Xu L, Mendes P, Kummer U: COPASI—a COmplex PAthway
22. Mestres J, Gregori-Puigjane E, Valverde S, Sole RV: The topology
SImulator. Bioinformatics 2006, 22:3067-3074.
 of drug-target interaction networks: implicit dependence
4. Letunic I, Yamada T, Kanehisa M, Bork P: iPath: interactive on drug properties and target families. Mol Biosyst 2009,
exploration of biochemical pathways and networks. Trends 5:1051-1057.
Biochem Sci 2008, 33:101-103. Mining of 7 different databases for drug-target interactions showed that,
on average, each drug has 6 known targets. Data completeness, the
5. Rojas I, Golebiewski M, Kania R, Krebs O, Mir S, Weidemann A, physicochemical properties of drugs, and the types of target families
Wittig U: Storing and annotating of kinetic data. In Silico Biol were found to be important factors in determining the topologies of drug-
2007, 7:S37-44. target networks.

www.sciencedirect.com Current Opinion in Biotechnology 2010, 21:511–516

516 Systems biology

23. Vogt I, Mestres J: Drug-target networks. Mol Inf 2010, 29:10-14. targeting ErbB3: a key node in ligand-induced activation of the
ErbB receptor-PI3K axis. Sci Signal 2009, 2:ra31.
24. Keiser MJ, Setola V, Irwin JJ, Laggner C, Abbas AI, Hufeisen SJ, Mathematical modeling of the ErbB receptor network together with new
Jensen NH, Kuijer MB, Matos RC, Tran TB et al.: Predicting new data is used to elucidate the effect of different drugs under normal and
molecular targets for known drugs. Nature 2009, 462:175-181. tumor conditions. An impressive example for clinically relevant results of
systems biology.
25. Campillos M, Kuhn M, Gavin AC, Jensen LJ, Bork P: Drug target
identification using side-effect similarity. Science 2008, 38. Sahin O, Frohlich H, Lobke C, Korf U, Burmester S, Majety M,
321:263-266. Mattern J, Schupp I, Chaouiya C, Thieffry D et al.: Modeling
26. Yuryev A (Ed): Pathway Analysis for Drug Discovery: ERBB receptor-regulated G1/S transition to find
Computational Infrastructure and Applications. Weinheim: Wiley; novel targets for de novo trastuzumab resistance. BMC Syst
2008. Biol 2009, 3:1.

27. Samaga R, Saez-Rodriguez J, Alexopoulos LG, Sorger PK, 39. Zhang T, Brazhnik P, Tyson JJ: Computational analysis of
Klamt S: The logic of EGFR/ErbB signaling: theoretical dynamical responses to the intrinsic pathway of programmed
properties and analysis of high-throughput data. PLoS Comput cell death. Biophys J 2009, 97:415-434.
Biol 2009, 5:e1000438. 40. Ueda HR, Hayashi S, Chen W, Sano M, Machida M, Shigeyoshi Y,
28. Lee E, Salic A, Kruger R, Heinrich R, Kirschner MW: The roles of Iino M, Hashimoto S: System-level identification of
 APC and Axin derived from experimental and theoretical transcriptional circuits underlying mammalian circadian
analysis of the Wnt pathway. PLoS Biol 2003, 1:116-132. clocks. Nat Genet 2005, 37:187-192.
Mathematical model of a cancer-relevant network highlighting potential
41. Brown SA, Kunz D, Dumas A, Westermark PO, Vanselow K,
drug targets. Control theory is applied to detect robustness and sensi-
Tilmann-Wahnschaffe A, Herzel H, Kramer A: Molecular insights
tivity of b-catenin and axin against changes in parameter values.
into human daily behavior. Proc Natl Acad Sci USA 2008,
29. Swameye I, Muller TG, Timmer J, Sandra O, Klingmuller U: 105:1602-1607.
Identification of nucleocytoplasmic cycling as a remote
sensor in cellular signaling by databased modeling. Proc Natl 42. Levi F, Schibler U: Circadian rhythms: mechanisms and
Acad Sci USA 2003, 100:1028-1033.  therapeutic implications. Annu Rev Pharmacol Toxicol 2007,
47:593-628.
30. Zi Z, Klipp E: Constraint-based modeling and kinetic analysis of The authors establish a relation between temporal regulation of cellular
the Smad dependent TGF-beta signaling pathway. PLoS ONE networks (daily rhythms, cell cycle) and effectiveness of therapies at
2007, 2:e936. different time windows, depending on the activity of the targeted pro-
cesses.
31. Schmierer B, Tournier AL, Bates PA, Hill CS: Mathematical
modeling identifies Smad nucleocytoplasmic shuttling as a 43. Levi F, Okyar A, Dulong S, Innominato PF, Clairambault J:
dynamic signal-interpreting system. Proc Natl Acad Sci USA Circadian timing in cancer treatments. Annu Rev Pharmacol
2008, 105:6608-6613. Toxicol 2010, 50:377-421.

32. Radulescu O, Gorban AN, Zinovyev A, Lilienbaum A: Robust 44. Raman K, Chandra N: Flux balance analysis of biological
simplifications of multiscale biochemical networks. BMC Syst systems: applications and challenges. Brief Bioinform 2009,
Biol 2008, 2:86. 10:435-449.
33. Chen WW, Schoeberl B, Jasper PJ, Niepel M, Nielsen UB, 45. Goltsov A, Maryashkin A, Swat M, Kosinsky Y, Humphery-Smith I,
Lauffenburger DA, Sorger PK: Input-output behavior of ErbB Demin O, Goryanin I, Lebedeva G: Kinetic modelling of NSAID
signaling pathways as revealed by a mass action model action on COX-1: focus on in vitro/in vivo aspects and drug
trained against dynamic data. Mol Syst Biol 2009, 5:239. combinations. Eur J Pharm Sci 2009, 36:122-136.
34. Bluthgen N, Legewie S, Kielbasa SM, Schramme A, Tchernitsa O, 46. Autiero I, Costantini S, Colonna G: Modeling of the bacterial
Keil J, Solf A, Vingron M, Schafer R, Herzel H et al.: A systems mechanism of methicillin-resistance by a systems biology
biological approach suggests that transcriptional feedback approach. PLoS ONE 2009, 4:e6226.
regulation by dual-specificity phosphatase 6 shapes
extracellular signal-related kinase activity in RAS- 47. Fang X, Wallqvist A, Reifman J: A systems biology framework for
transformed fibroblasts. FEBS J 2009, 276:1024-1035. modeling metabolic enzyme inhibition of Mycobacterium
tuberculosis. BMC Syst Biol 2009, 3:92.
35. Borisov N, Aksamitiene E, Kiyatkin A, Legewie S, Berkhout J,
Maiwald T, Kaimachnikov NP, Timmer J, Hoek JB, 48. Barrett MP, Bakker BM, Breitling R: Metabolomic systems
Kholodenko BN: Systems-level interactions between biology of trypanosomes. Parasitology 2010, 17:1-6.
insulin-EGF networks amplify mitogenic signaling. Mol Syst
Biol 2009, 5:256. 49. Sahle S, Mendes P, Hoops S, Kummer U: A new strategy for
assessing sensitivities in biochemical models. Philos Trans A
36. Fitzgerald JB, Schoeberl B, Nielsen UB, Sorger PK: Systems Math Phys Eng Sci 2008, 366:3619-3631.
biology and combination therapy in the quest for clinical
efficacy. Nat Chem Biol 2006, 2:458-466. 50. Aebersold R, Auffray C, Baney E, Barillot E, Brazma A, Brett C,
Brunak S, Butte A, Califano A, Celis J et al.: Report on EU-USA
37. Schoeberl B, Pace EA, Fitzgerald JB, Harms BD, Xu L, Nie L, workshop: how systems biology can advance cancer research
 Linggi B, Kalra A, Paragas V, Bukhalid R et al.: Therapeutically (27 October 2008). Mol Oncol 2009, 3:9-17.

Current Opinion in Biotechnology 2010, 21:511–516 www.sciencedirect.com