Signaling Networks in Biology
Anamika Sarkar*, Aislyn Wist,* and Ravi Iyengar, Department of Pharmacology and Systems Therapeutics, Mount Sinai School of Medicine, New York, New York
Signal transduction in cells has been an area of intense study for many years. However, until recently, the focus was on interactions between individual components rather than on the global behavior of the cell signaling networks. New experimental technologies, such as protein and DNA microarrays, high-throughput screening instrumentation, and diverse compound libraries, allow for many interactions to be examined simultaneously. Along with new experimental methods, quantitative models with mathematical methods, such as deterministic, stochastic, and statistical analyses and graph theory, have been useful in the mapping and analysis of intracellular signaling networks. Here we describe current experimental and computational approaches that are used to develop a global understanding of the complex behavior of intracellular signaling networks. We also discuss important applications of signaling network analysis, including the discovery of new drug targets, the identification of the signaling components responsible for the side (off-target) effects of drugs, and the development of combination therapies.
Currently it is estimated that the human genome contains 25,000 protein-encoding genes (1), which correspond to several hundred thousand possible protein species resulting from alternative splicing and posttranslational modifications. Of these presumed protein species, the functions and interactions of only a fraction are well understood. Among the well-studied proteins, many interactions result in complex networks that often form the basis for the complex physiologic functions. Moreover, protein expression levels vary among individuals and different cell or tissue types, and the distribution of proteins varies in the subcellular regions of each cell. Therefore, it is important to develop tools to analyze these systems as a whole.
Despite the growing need for global analysis of signaling networks, mapping individual pathways has been the main approach in studying signal transduction. Studies focused on defining how these pathways interact to form networks are relatively new. By mapping out signaling networks within cells, we can get an initial description of the rules that are applied in the development of these networks and the discovery of new trends or network properties. This review article discusses the interactions that comprise intracellular networks, the mechanisms underlying them, their conserved trends, and the emerging tools being developed to understand networks as a whole.
* Both authors have contributed equally and share credit as first authors.
Cell Signaling Networks
When a cell receives stimuli, the signals are propagated from one component to another, eventually fanning out to form a network of signals that produces physiologic responses (phenotypes). Interactions within a set of interconnected signaling pathways can be represented as a network composed of components and connections. Many natural and engineered systems can be described as networks. These systems include the World Wide Web (2), networks of coauthors of scientific papers (3), and other social networks (4). Biochemical interactions within cells can also be described as networks (5-7), and in this review we consider two intracellular networks, one involved in the stimulation of cell proliferation by the epidermal growth factor (EGF) and the other involved in the initial immune response to pathogens through the toll-like receptor (TLR) (Fig. 1a and 1b, respectively).
Mechanisms of Cellular Signaling
The initial stimulus for a signaling pathway can be intracellular or extracellular. Intracellular stimuli typically are the result of changes in the cellular environment, such as the formation of reactive oxygen species or DNA damage. For extracellular signaling events, the process is initiated by a ligand binding to a receptor, often transmembrane proteins, such as the EGF receptor (EGFR) or the TLR (Fig. 1a and 1b). Ligands can either be small molecules (hormones and neurotransmitters), peptides (insulin), or small proteins (growth factors). The naturally occurring ligands that have an affinity for EGFR are EGF, transforming growth factor (TGF-α), and β-cellulin (8), all of which are small proteins, whereas the natural ligand for TLR is a macromolecule, lipopolysaccharide (LPS). Extracellular ligands can be secreted by other cells or generated by the cleavage of a larger precursor protein within the signaling cell. For example, EGF is generated by the proteolytic cleavage of the membrane-bound precursor, heparin-binding EGF-like factor (HB-EGF), by a matrix metalloproteinase (MMP) (9).
Biochemical reactions involved with signal propagation
Signal propagation within cells occurs by binding and enzymatic interactions that involve proteins, ions, and/or small molecules. Binding interactions typically occur between two or more proteins, proteins and lipids, or proteins and small molecules. Enzymatic reactions involved in signal propagation include phosphorylation, dephosphorylation, ubiquitination and sumoylation, controlled protein cleavage, and the generation of small, signaling molecules such as cyclic adenosine monophosphate (cAMP), diacylglycerol (DAG), and inositol triphosphate (IP3). Many signaling reactions in intracellular networks are reversible chemical reactions, catalyzed in the forward direction by one enzyme and in the reverse direction by a different enzyme. One example of a reversible signaling reaction is the phosphorylation of a protein by a kinase and the subsequent dephosphorylation by a phosphatase. For instance, in the TLR pathway (Fig. 1b), the activated kinase Tak1 (TGFβ activated kinase 1), phosphorylates IKK (IKB kinase), whereas the protein phosphatase 2C (PP2C) dephosphorylates phospho-IKK. In nonenzymatic signaling reactions, adaptor proteins act as mediators between the activated receptor and signaling enzymes by recruiting other signaling molecules to form a complex, which propagates the signal to downstream targets. For instance, although Grb2 (growth factor receptor bound-2) is not an enzyme itself, it acts as an adaptor protein by mediating the interaction between activated EGFR (a receptor tyrosine kinase) and Sos (Son-of-Sevenless, a guanine nuclear exchange factor, GEF) that in turn regulates Ras (R-RAS related family member, a GTPase). It is the combination and dynamic balance of these enzymatic and binding processes that make up an intracellular network.
Patterns of signal flow: propagation, amplification, and loops
Signals propagate through cells in signaling cascades. In the EGFR pathway (Fig. 1a), several cascades potentially can be activated, depending on the specific ligands that bind EGFR and on the identity of its dimerization partner. One cascade, which begins with the recruitment of Grb2 to activated EGFR, in turn activates Sos, which facilitates the exchange of GDP for GTP by the small GTPase, Ras. Activated GTP-bound Ras (a kinase) activates the protein kinase Raf-1 (v-raf-leukemia viral oncogene 1), which induces the kinase cascade (boxed in Fig. 1a) that leads to the activation of MAPK (through MAP kinase-ERK kinase, MEK), which goes on to activate transcription factors such as Elk and Myc in the nucleus (5, 10). A different signaling cascade also begins with Grb2 binding to the phosphorylated EGFR and, in turn, recruiting adaptor proteins Gab1 (growth factor receptor bound protein 2-associated protein 1). Gab1 is phosphorylated by EGFR, whereby it activates phosphatidylinositol-3 kinase (PI3K), which then activates the kinase Akt (v-akt murine thymoma viral oncogene homolog 1) that mediates cell survival signaling (5, 10).
Signal transduction from the membrane to the nucleus occurs in the TLR network via interactions between signaling components. For example, one TLR pathway (Fig. 1b) starts with the activation of TLR3 by the binding of viral dsRNA. Next, the adaptor protein, TRIF (Toll/IL-1 receptor domain-containing adaptor inducing IFNP), is recruited, and TRIF mediates the phosphorylation of TANK-binding protein kinase 1 (TBK1) by an unidentified kinase. Phospho-TBK1 in turn phosphorylates the transcription factor, IFN-regulatory factor-3 (IRF3), at five distinct sites. The phosphorylated IRF3 homodimerizes and translocates to the nucleus to regulate transcription (11, 12).
Cellular signals generally are not propagated through single, linear pathways. Instead, the signals have the tendency to branch or fan out into several or many pathways, which lead to one or several possible outcomes. An activated protein may have more than one protein that it activates or interacts with, such as Grb2 (Fig. 1a) and TRAF6 (tumor necrosis factor receptor-associated factor 6) and Tak1 (Fig. 1b). Nonlinearity increases the complexity of the network and introduces crosstalk between pathways or networks. Networks for the EGFR pathway (5) and the TLR pathway (6) have been constructed and have resulted in highly detailed, complicated schematics that represent the networks in their entirety; hence, a portion of each network is depicted in Fig. 1 to create a meaningful visual representation of the network. It is notable in complete networks that patterns of signal propagation and amplification exist.
Signal flow within networks often is controlled and/or amplified by regulatory loops (Fig. 1c). The connectivity within regulatory loops has multiple configurations. For example, consider an upstream protein A that activates downstream protein B that in turn activates protein C. In a feed forward loop (FFL), protein A also activates protein C. In a positive feedback loop (PFBL), protein C in turn activates protein A. In a negative feedback loop (NFBL), protein C inhibits protein A. In the TLR pathway, a FFL is found within the MAPK cascade that activates ATF2 (activating transcription factor 2) through both p38-MAPK and JNK (cJun NH(2)-terminal kinase) (13-16). A PFBL is created by PI3K activity in the EGFR pathway (represented as a thick, curved arrow in Fig. 1a). Once indirectly activated by EGFR (through phosphorylation of Gab1), PI3K generates phosphatidylinositol (3-5)-trisphosphate that facilitates the recruitment of additional Gab1 (17). In a larger PFBL, EGFR signaling causes the transcription of its ligands HB-EGF and TGFα, through the Ras-MAPK pathway (18). A PFBL is found where the ligand, interferon β (IFNβ), activates a TLR pathway that leads to its transcription (Fig. 1b) (19). An example of a NFBL is where SHP, a phosphatase, is recruited to EGFR through Gab1 and Grb2 but inhibits EGFR signaling by dephosporylation of the receptor (5, 20). Another example of a more distal NFBL is EGFR activation of Ras-GAP (Ras-GTPase activating protein), which inhibits Ras signaling by facilitating Ras-mediated GTP hydrolysis to GDP (5).
All these regulatory mechanisms play a crucial role in maintaining signal amplification and cellular homeostasis. PFBLs amplify signal and NFBLs reduce signal, which maintains a balance. Properties of signaling networks such as redundancy and robustness have been identified through the examination of reconstructed signaling networks (21, 22). Redundancy, such as that provided by FFLs, is a mechanism of maintaining homeostasis because the same outcome can be achieved by several different paths. If one path fails, another route is available to the same outcome, which leads to network robustness.
Figure 1. Signaling networks and loops. Schematic interaction maps that represent the major components in the signaling networks emanating from (a) the epidermal growth factor receptor and (b) the toll-like receptor are shown. (c) Schematics that represent several types of regulatory loops that are found in intracellular signaling networks.
Families of signaling receptor proteins most often include several isoforms that have varying affinities for ligands and adaptor proteins. Usually several to many ligands have affinity for one or more receptor. Once all possible ligand-receptor combinations and the multiplicity of recruited adaptor and signaling protein combinations are taken into account, the complexity of signaling networks becomes apparent. Take for example the ErbB signaling network. Four members exist: ErbB1/EGFR, ErbB2/Her-2/Neu, ErbB3, and ErbB4 (Fig. 1a does not depict all receptors, ErbB1-4, and ligands involved in the ErbB network). Also, several ligands are involved in the ErbB network; some are promiscuous, and others have affinity for only one receptor. For example, EGF specifically binds EGFR of all the ErbB receptors, but neuroregulin-1 has affinity for ErbB3 and ErbB4 (8).
As depicted in Fig. 1a, an abundance of adaptor proteins signal in the ErbB network. Ligand binding induces a conformational change that allows the ErbB receptors to homo- or heterodimerize and crossphosphorylate at the C-terminal region of the intracellular domains (23, 24). The potential of the receptors to homo- or heterodimerize with one another varies because of the expression level of the receptor, their affinities, and the amount and identity of ligand(s) present (8, 10). The phosphorylated region provides a docking site for adaptor proteins, which have varying affinities for the different ErbB receptors. Specifically, the Src homology-2 (SH2) or phosphotyrosine binding (PTB) domains of adaptor proteins including Vav, Shc, Grb2, small heterodimers partner (SHP), and phospholipase-Cγ (PLCγ), have affinity for the intracellular tyrosine-phosphorylated EGFR region (Fig. 1a). Many possible combinations of ligands, receptors, and recruited adaptor proteins can form the final signaling complex. Therefore, complexity in the signaling network is introduced early, at the level of the ligand. Further, the ligand governs the nature of the dimerized receptors which, in turn, modulate the mode and magnitude of the signal. At the ligand-receptor level, the TLR network (Fig. 1b) is similar to the EGFR network with respect to ligand specificity. Some TLR ligands are specific for one TLR, and others are not. In contrast to the EGFR network, one main adaptor protein MyD88 (myeloid differentiation primary response gene 88) has similar affinity for all TLR family members, with the exception of TLR3. Therefore, in the TLR pathway the nature and magnitude of the signal depends highly on the expression level of the receptors and the nature and relative amount of ligand(s) present. Because of the variation in the identity and level of expressed receptors across various cell types, different cells will respond differently to stimulation with particular ligands. The distribution of different receptors reflects the interplay of the signaling capacity of the receptor and its significance in the specialization of the cell type. For example, immune cells, such as dendritic cells, have a relatively high expression level of TLRs, whereas epithelial cells have relatively high expression levels of growth factor receptors (25, 26).
Experimental Approaches to Understanding Signaling Networks
The majority of published, reconstructed cellular signaling networks are built on the data that describe individual interactions and loops through classical biochemical techniques, such as immunoblot analysis, coimmunoprecipitation, binding, and enzymatic assays. However, to obtain global perspectives on interaction networks, new experimental approaches to obtain high-content data sets on interactions that make up cellular signaling networks recently have been developed. These fields serve two distinct but complementary purposes: First, to catalog and inventory genes, mRNAs, and proteins and their functions; second and more pertinent, to develop signaling networks, to parse, on a global scale, the properties of signaling networks in cells under different conditions, such as stress, native ligand treatment, non-native agonist/antagonist treatment, or other. Also, to observe differences in different cell types, including healthy/normal cells and those in a disease state. Often, because of the heightened possibility of false positive or negative results because of their high-content nature, results from these high-content experiments are validated by single experiments that use more classical techniques. The experimental methodologies and examples of experiments in these fields will be discussed here; whereas, the mathematical and computational approaches including the analyses of the resultant large data sets will be discussed in the section entitled “Computational approaches that facilitate the understanding of intracellular networks.”
Genomic and proteomic methods
Genomics entails cataloging genes and genetic mutations and observing changes in gene expression levels under various conditions. Genomics can be used to catalog genetic mutations, polymorphisms, or amplifications that may play a role in disease by comparing gene sequences or levels from disease and normal genetic samples. The overall changes in gene expression or genetic mutations in a disease can help identify a critical signaling protein or pathway as a therapeutic target. Genetic analysis of cells that have undergone treatment to activate a specific signaling pathway can lead to signaling information, which provides a basis for the construction and analysis of the cell signaling network. For example, after treatment with an agent known to affect exclusively a pathway of interest, all the genes that are upregulated or downregulated by the pathway can be analyzed by genomics. This type of analysis may lead to information on the genetic programs that are outcomes of signaling pathways.
The technology to detect gene expression, coined DNA microarrays or DNA chips, was developed over the past decade (27). Recently, methods such as ChIP-chip (chromatin immunoprecipitation-microarray) (28), DamID (DNA adenine methyltransferase ID) (29), and PBM (protein binding microarrays) (30), allow for a more mechanistic understanding of gene regulation (and chromosome structure). With transcription factor microarrays, one can detect transcription factors that are activated or inhibited under various conditions and/or different cell types.
Proteomics focuses on obtaining an inventory of proteins and their functions, as well as unfolding interaction networks and posttranslational modifications, such as phosphorylation (phospho-proteomics). Further, proteomic experiments can yield quantitative information such as rates and relative concentrations. A range of methodologies are used to generate proteomic data, ranging from low-throughput methods such as Western blotting and immunoprecipitaiton to high-throughput methods such as protein microarrays, yeast two-hybrid screens (31), and SILAC (stable isotope labeling by amino acids in culture) combined with quantitative tandem mass spectrometry (32). Important advances in the technologies involved in proteomic studies allow for a more detailed, global understanding of interaction networks.
A protein microarray study by Jones et al. (33) was designed to test the affinity of all proteins that contain SH2 and PTB domains, which specify binding to the phosphorylated domains of all four ErbB family members. The study uncovered that some ErbB family members were more promiscuous than others, which has important implications for the ErbB signaling network in general and specifically in that the promiscuous ErbB family members are much more commonly overexpressed in several cancer cell types. Furthermore, 116 new ErbB interaction partners were discovered. Another study by Schulze et al. (34) was also designed to identify all interaction partners for the phosphorylated ErbB family members but using a novel methodology that combines SILAC and LC-MS/MS. This study defined the specific ErbB sites where the interaction partners bind.
Many of these studies have resolved the temporal and spatial aspects of signaling networks. The dynamics of EGF stimulation was examined by Blagoev et al. (35) in a phospho-proteomics study by examining the phosphorylation pattern of EGFR after different periods of EGF treatment. The spatial aspects of signaling can be captured by analyzing different fractions of the cell such as the nucleus, organelles, vesicles, membrane-bound proteins, or cytoplasmic proteins. Several recent studies have analyzed holistically the signaling components of purified cellular organelles, such as synaptic vesicles (36) and the ER-Golgi apparatus (37).
Chemical genetic approaches
Chemical genetics (38, 39) can be considered a combination of chemistry and genetics or proteomics because it is based on the use of small molecules to modulate protein or gene function. This method is similar to pharmacology, except that the experiments are designed in a similar fashion to “omics” experiments, in which an entire set of genes, proteins, or phenotypes are examined simultaneously. The effects of modulating protein function with a small molecule or library of small molecules are examined either phenotypically or at the signaling level. Chemical genetic experiments are instrumental in understanding how signaling networks are effected by the down- or upregulation of one or more signaling proteins by a drug or other small molecule(s).
The two types of chemical genetic methods, “forward” and “reverse,” lead to different results. In a “forward” experiment, small molecules are used to probe for a desired phenotype, and the protein that binds the small molecule and induces the phenotype can be subsequently identified. In an example of a “forward” chemical genetic experiment, the small molecule necrostatin-1 was discovered by screening a 15,000-compound library for an inhibitor of a novel type of cell death that is unique from apoptosis or necrosis, coined “necroptosis” (40). Other forward chemical genetic experiments have led to information on the ground state of neural stem cells (41) and on the role of copper and lysyl oxidase in notochord morphogenesis during development (42). Furthermore, forward chemical genetic experiments have been used to identify unknown targets of clinically approved drugs and can lead to the identification new potential drug targets (38, 39, 43).
In a “reverse” chemical genetic experiment, a protein with a known function is modulated with small molecules, either to simply discover a small molecule modulator of the protein or to examine the downstream effects of modulating the protein to further dissect the mechanism of its function. In an example of a “reverse” chemical genetic experiment, a protein kinase of interest is mutated to be specifically inhibited by a bulky, synthetic ATP analog, which cannot bind all other protein kinases. These “reverse” chemical genetic experiments have led to the elucidation of numerous signaling pathways including the cell cycle pathway by inhibiting CDK-activating kinase 7 (cdk7) (44), cJun NH(2)-terminal kinase (JNK) signal transduction pathway (45, 46), G protein-coupled receptor kinases (GRK)-regulated receptor regulation pathway (47), the actin assembly process by inhibiting v-Src (48), and mitosis and cell division by inhibiting Plk-1 (49). The consequences or potential off-targets of modulating a protein with a small molecule or drug can be elucidated by using reverse chemical genetics.
Collaborative research initiatives for the use of chemical genetics to facilitate discovery of potential cancer (or other) therapeutics by the Initiative for Chemical Genetics at the National Cancer Institute (50) and to facilitate the discovery of new targets and drug candidates by the Chemical Genomics Center at the National Institutes of Health (51) have been initiated. These initiatives will result in databases that contain information that most likely will be valuable to the study of cellular signaling networks.
The design and synthesis of chemical compound libraries
The synthesis, collection, and development of chemical libraries are essential for the global analysis of intracellular signaling networks, chemical genetics, and drug discovery. Commercial sources of compound libraries are available and often categorized by the type of compounds they contain. Library categories include natural product, drug-like, various molecular weight ranges, various inhibitor classes, and peptide or peptide-like (38, 52). Many institutions have a screening facility where purchased libraries and/or libraries obtained from archives of compounds synthesized in-house are available for assaying.
Most commonly, a commercial library, containing a subset of compounds that matches a desired set of properties (if known), is screened in an initial study. Once hits are obtained and verified, a small library of compounds is synthesized to produce a set of compounds in the same chemical-structural space as the original hit structure. This process allows for hits with higher potency and elucidates information of the structure-activity relationship within the system of interest. Such synthetic libraries of chemically diverse compounds have been made possible through combinatorial chemistry (52-56) and diversity-oriented synthesis (52).
There continues to be is a strong motivation in the chemical biology field to synthesize new libraries and build on existing ones, as well as to improve combinatorial chemistry approaches. The data collected from chemical genetic experiments adds to the delineation of the proteome and a complete cell signaling network and promotes the discovery of new therapeutic targets. To make a compilation of the data generated, Chembank (affiliated with the Initiative for Chemical Genetics by the NCI) was created. Chembank is an initiative to create a public database containing all data obtained from screening compound libraries for various purposes.
Computational Approaches that Facilitate the Understanding of Intracellular Networks
In the past decade, computational analyses, based on mathematical methods and experimental data, have been recognized as powerful tools to understand the complexity that is inherent to biological systems. The integration of computational and experimental approaches to construct and analyze cell signaling networks generally is a multistep process involving collaborative efforts between the two fields. The major steps involved in the development of mathematical models that yield systems-level insights are outlined in Fig. 2.
Figure 2. The process of integrating mathematical modeling and experimental data for analyzing intracellular networks. A flowchart that describes steps involved in the use of computational modeling and experimental data to obtain a systems level understanding of cellular signaling networks.
The first step in constructing a cell signaling network model is to generate an in silico interaction network. Signaling components of interest are identified, and data on binary interactions are extracted from the experimental literature. Often times, the creation of an entire interaction map for a large network is beyond the scope of one laboratory; therefore, public databases have been created in which newly discovered proteins and/or protein interactions are deposited (57-63). However, often data are included from studies that cover a broad range of protein interactions, such as proteomic studies or yeast two-hybrid screens, and frequently contain many potential false positives or negatives. Thus, it becomes necessary to critically examine each reference to verify the interaction data reported, as this in silico network is the basis of further study.
Once the reaction network is created, an appropriate mathematical approach is identified according to the information available and the biological question(s) to be addressed. In this section, some mathematical approaches that have been applied successfully in the analysis of biological signaling networks will be discussed. These methods are summarized in Table 1 (64-85).
Table 1. Various computational methods that have been used in analyzing signaling networks
Coupled biochemical systems
• Reaction kinetics are represented by sets of ordinary differential equations (ODEs).
• Rates of activation and deactivation of signaling components are dependent on activity of upstream signaling components.
Spatially specified systems
• Reaction kinetics and movement of signaling components are represented by partial differential equations (PDEs).
• Useful for studies of reaction-diffusion dynamics of signaling components in two or three dimensions.
• Based on Monte Carlo theory and probabilities of reactions.
• Facilitates modeling signaling components in small volumes and capturing stochastic fluctuations within the network.
• Describes topology of networks and subnetworks, based on quantification of the number of nodes (signaling components) and links between them.
• Dynamic properties of networks through Boolean analysis.
• Network analysis based on probabilities (Markov chain and Bayesian) to identify paths and relationships between different nodes in the network.
• Clustering based on correlations to identify group relationships.
• Partial least square regression analysis to track pathways of signal flow.
• Principal component analysis to identify significant signaling components.
High-throughput experiments such as those involved with genomic and proteomic studies (e.g., microarray studies and SILAC-mass spectrometry studies, discussed in the section entitled “Experimental approaches to understanding signaling networks”) are useful for understanding the interdependence of signaling components in the network. However, these types of experiments generally yield large-scale data sets that often are analyzed by computational algorithms, based on statistics. The basic concept behind this computational statistical analysis is to identify clusters or groups of signaling components that show similar trends. This analysis is used to extract the relationships between different signaling components in a network from these large-scale data sets. These kinds of analyses have been used for analyzing microarray data to identify patterns that correlate with distinct physiologic and pathophysiologic states. For example, an hierarchical algorithm has been used to identify different types of cancers in human soft tissue tumors (86), and K-mean clustering has been used to identify molecular subtypes of brain tumors (87). However, these types of statistical clustering analysis do not exclude the noise or artifacts induced by the experimental techniques. To overcome this limitation, singular value decomposition (a modified type of principal component analysis) has been used to identify which clusters of signaling components have a significant amount of influence in changing patterns of biological behaviors (88-91). Another statistical technique, partial least square regression analysis on multidimensional experiments (signaling components, time, and ligands), has been used to predict new interactions between components in a network, regulating apoptosis (92).
Statistical analysis is a powerful tool for understanding interdependence between groups of self-similar behaviors, especially when a large data set is being analyzed. However, it does not yield information about the topology of the interaction network. Graph theory-based approaches have been developed to study the topological properties of a large biological network. In graph theory, each signaling component in the network is represented as a node and identified by its degree, which represents the number of interacting partners. Within the context of the entire signaling network, the relationship between the degree of individual nodes and the degree of distribution of the whole network can be characterized. This kind of analysis has shown many intrinsic topological properties, such as scale-free configuration, small world configuration, and modular organization of networks across different cell types and species (2, 7, 22, 93, 94).
Although it is possible to analyze the topological properties of the network using graph theory, the states of the signaling components (phosphorylated/dephosphorylated, bound/unbound) often are not accounted for in such analyses. Boolean analysis can be used to understand such dynamics. In Boolean analysis, the signaling components in the network are identified by strings of 0’s and 1’s. The activated state of the protein is denoted by ‘1’, and the inactivated state is denoted by ‘0’. Changes in states of components can be computed as a function of various perturbations to the network.
To illustrate the concept of Boolean analysis in a biological context, an example is taken from the TLR pathway (Fig. 1c). To initiate the induction of the IFNP gene in TLR signaling pathway, it is necessary for three transcription factor complexes, ATF2-cJun heterodimer, two IRF3 homodimers, and NF-KB, to bind simultaneously at the promoter region. This biological condition can be represented in the Boolean form in the following way: Each of the four transcription factor complexes, ATF2-cJun, two IRF3 homodimers and NF-KB, and the gene IFNP is expressed as strings of five 0s and 1s that lead to 25 (or 32) combinations. For instance, the representation ‘11111’ denotes that all transcription factor complexes are bound to the promoter regions and IFNP gene expression occurs; ‘01000’ denotes that only one of two IRF3 homodimers are bound to the promoter region and, hence, IFNβ gene expression does not occur. A simulation is performed to analyze the effect of the change of state of an upstream signaling component on the downstream component in the signaling network, and rules, based on Boolean operators, are applied (95, 96).
Boolean analysis has been used to understand the effect of different states of signaling components on their downstream signaling components under various conditions. For example, Albert and Othmer (64) used Boolean analysis of a Drosophila melanogaster network to validate results from previously published microarray data and predicted the crucial role of the two genes, wingless and sloppy paired, in segment polarity. Gonzales et al. (65) mutated different genes in silico and used Boolean analysis to predict that the delay in the activation of two genes Apterous and Notch is essential to maintain the dorsal-ventral boundary of Drosophila melanogaster. Sarkar and Franza (97) predicted that the costimulation of T-cell receptors (TCR) and CD28 in T cells influences cell proliferation by providing a greater diversity of paths in the network. Maayan et al. (67) has used a combination of graph theory and Boolean analysis to obtain a distribution of redundant pathways in a cell signaling network.
Although large-scale network analysis using graph theory or Boolean analysis can be very powerful in understanding the overall topological properties of the network, these tools do not take into consideration the rate of change of states of the signaling components with respect to time, explicitly. To understand the dynamics of activation or deactivation of a component in a signaling network, the most commonly used approach is deterministic analysis, which involves solving ordinary differential equations (ODEs). For deterministic analysis of biological systems, the initial concentrations and the kinetic parameters (KM, kcat, and Vmax) for enzymatic reactions and binding reactions (Kd and kon or koff) are required and need to be determined experimentally.
Numerous examples exist in which deterministic analysis has provided critical insights into the dynamic behavior of a protein in a network under various conditions. One early effort in computational modeling, which uses a system of ODEs, is the EGF-EGFR reaction kinetics and internalization of the receptors (98-102). Bhalla and Iyengar (103) predicted the importance of feedback loops in a signaling network by illustrating bistable behavior of output proteins. Moreover, the sensitivity of the MAPK activity in response to a stimulus has been studied in detail by using systems of ODEs (104, 105). Bentele et al. (106) have studied apoptosis induced by the CD95 receptor by using deterministic analysis.
The deterministic analysis provides the average behavior of signaling components in a network but does not include the results of fluctuations in the activation states of the signaling components (noise). Such analysis becomes essential when the volume of interest becomes in the scale of femtoliters because of the limitingly small number of molecules contained in such a small volume (1 fL of a 0.1 μM solution is equivalent to 60 molecules). Small volumes such as these are biologically relevant and present in cellular systems quite often, for example, in the spines of dendrites and endosomal regions. Moreover, stochastic analysis becomes important in studying gene regulation where many genes have less than 10 copy numbers.
Stochastic analyses, based on Monte Carlo theory and probabilities, have been used to capture the effects of fluctuations of an upstream signaling component on the kinetics of a downstream signaling component in the network. The most popular method for stochastic analysis is the Gillespie method (107), which is based on the probability of which reaction will take place within a defined time period. This type of simulation can give insights into the characteristics of signaling events at the single-molecule level. For example, in the analysis of an individual cell, noise introduced in enzymatic futile cycles can be strong enough to switch the system from one state of the cell to another (108). This technique has not gained much popularity over deterministic approaches, even though it is the only appropriate mathematical representation for certain cellular phenomenon. This lack of popularity primarily is because of the following reasons: 1) Stochastic models need information at the molecular level as building blocks to construct the model properly and to validate simulation results obtained from the model. Experimental methods that yield molecular-level data have not been developed for biological systems, and 2) Stochastic simulations of a relatively small model, which contains less than 10 signaling components and their interactions, can pose large computational demands. However, application of the Gillespie method in small volume can reduce computation time to be comparable to that of deterministic simulation.
The Importance of Signaling Networks in Understanding Mechanisms that Underlie Disease and Drug Discovery
Cellular signaling networks are useful for drug discovery in three ways: 1) To develop an understanding of the biochemical mechanisms that underlie disease, which aids the discovery of novel drug targets, 2) To explain how current drugs affect the on-target (leading to therapeutic effect) and off-target (causing side effects and/or drug resistance) signaling pathways, which may lead to more informed prescribing methods and improved therapies, and 3) To lay out the interplay between signaling pathways in such a way that explains or facilitates the success of combination therapies.
A dynamic relationship exists between the understanding of the biochemical mechanisms that underlie disease and drug discovery (Fig. 3). The compilation of data into a detailed, signaling network is central in properly choosing a drug target (109). Many network models are in place, such as those described for the EGFR and TLR networks in this review, which can be applied to drug discovery (110). Before pursuing a drug target, it is important to understand as thoroughly as possible the consequences of regulating the target of choice because of the sheer expense involved with drug development and clinical trials. This understanding may be gained by analyzing a computational model of the signaling network involved, generating a knockout mouse, examining the chemical genetics of the target, and critically examining the current data on the drug target in the literature and databases (i.e., Chembank).
Figure 3. Intracellular signaling networks, disease, and drug discovery. A schematic is shown that represents the ways in which intracellular signaling networks play a role in the drug discovery and development process.
Of the total number of proteins that have a function we currently understand, only a small fraction is drug targets. The pool of drug targets is evolving with the introduction of new drug discovery technologies, such as the cell signaling network analyses described in this review, combinatorial chemistry, chemical genetics, molecular informatics, and advanced high-throughput screening technologies. With a new direction in the medical field toward personalized therapy regiments, it is increasingly important to expand the “druggable” network, and the enhanced global knowledge of signaling networks and the new drug discovery technology are indispensable in achieving this goal.
Although signaling network analysis is essential to uncover new drug targets and aid in subsequently validating their potential, it also plays a role in studying roadblocks for ap- proved/pipeline drugs, such as developed drug resistance or off-target effects. Network analysis combined with carefully designed proteomic experiments aid in the understanding of the mechanisms of resistance and in developing methods to overcome the resistance. For example, a study by Chen et al. (111) combined signaling network analysis and a carefully designed proteomic experiment to illuminate the signaling network involved with the resistance of ovarian cancer cells to cisplatin treatment.
In the light of the abundant data on drug resistance in disease, many therapies are dosed in combinations to affect several pathways at once and/or gain a therapeutic advantage that stems from synergistic effects between drugs (112). An assembly of the data on the pathophysiology and biochemistry of cancer will allow treatment regimens to evolve toward targeting a network rather than a single protein (113). Efforts are currently underway to use a network analysis approach in understanding the effects of treatment with more than one drug (114). Alternatively, the combination of effects of two or more drugs can be used to uncover information about how the networks that each drug targets are connected (114). Combined with advances in high-throughput screening technologies, these approaches will be the powerhouse in the discovery of numerous novel therapeutic drug combinations and the most effective dosing schedules.
Perspective and Future Directions
Cell signaling networks are a rich source of targets for drug. A vast majority of drugs today are targeted to receptors or signaling components inside the cell such as protein kinases. The understanding that individual signaling pathways come together to form networks is very useful in looking for new drug targets not solely as individual components but also as pairs and triplets in interacting pathways for combination therapy for complex diseases. The convergence of computational approaches with high-throughput experimental analyses and chemical libraries offers substantial new opportunities not only to understand in depth the regulatory complexities within cells but also to design and identify drugs that can be used as therapeutic agents for diseases that arise from alterations in the regulatory capabilities of cell signaling networks.
This work was supported by National Institutes of Health Grants numbers GM54508, DK38761 and a contract from National Institute of Allergy and Infectious Diseases.
1. The Human Genome Program of the U.S. Department of Energy Office of Science. http://www.ornI.gov/sci/techresources/Human_Genome/home.shtml.
2. Albert R, Jeong H Barabasi AL. Diameter of the World-Wide Web. Nature 1999; 401:130-131.
3. Newman MEJ. The structure of scientific collaboration networks. Proc. Natl. Acad. Sci. U.S.A. 2001; 98:404-409.
4. Wasserman S, Faust K. Social Network Analysis. 1994. Cambridge Univeristy Press, Cambridge.
5. Oda K, Matsuoka Y, Funahashi A, Kitano H. A comprehensive pathway map of epidermal growth factor receptor signaling. Mol. Syst. Biol. 2005; 1:1-17.
6. Oda K, Kitano H. A comprehensive map of the toll-like receptor signaling network. Mol. Syst. Biol. 2006; 2:1-20.
7. Maayan A, Jenkins SL, Neves S, Hasseldine A, Grace E, Dublin-Thaler B, Eungdamrong NJ, Weng G, Ram PT, Rice JJ, Kershenbaum A, Stolovitzky G, Blitzer RD, Iyengar R. Formation of regulatory patterns during signal propagation in a mammalian cellular network. Science 2005; 309:1078-1083.
8. Jones JT, Akita RW, Sliwkowski MX. Binding specificities and affinities of egf domains for ErbB receptors. FEBS Lett. 1999; 447:227-231.
9. Nanba D, Higashiyama S. Dual intracellular signaling by proteolytic cleavage of membrane-anchored heparin-binding EGF-like growth factor. Cytokine & Growth Factor Reviews. 2004; 15: 13-19.
10. Yarden Y. Sliwkowski M. X. Untangling the ErbB signalling network. Nat. Rev. Mol. Cell Biol. 2001; 2:127-137.
11. Takeda K, Kaisho T, Akira S. Toll-like receptors. Annu. Rev. Immunol. 2003; 21:335-376.
12. Kawai T, Akira S. TLR signaling. Cell Death Differ. 2006;13: 816-825.
13. Rahaus M, Desloges N, Wolff MH. ORF61 protein of Varicella- zoster virus influences JNK/SAPK and p38/MAPK phosphorylation. J. Med. Virol. 2005; 76:424-433.
14. Chu WM, Ostertag D, Li ZW, Chang L, Chen Y, Hu Y, Williams B, Perrault J, Karin M. JNK2 and IKKbeta are required for activating the innate response to viral infection. Immunity. 1999; 11:721-731.
15. Iordanov MS, Paranjape JM, Zhou A, Wong J, Williams BR, Meurs EF, Silverman RH, Magun BE. Activation of p38 mitogen- activated protein kinase and c-Jun NH(2)-terminal kinase by double-stranded RNA and encephalomyocarditis virus: involvement of RNase L, protein kinase R, and alternative pathways. Mol. Cell Biol. 2000; 20:617-627.
16. Du W, Thanos D, Maniatis T. Mechanisms of transcriptional synergism between distinct virus-inducible enhancer elements. Cell 1993; 74:887-898.
17. Rodrigues GA, Falasca M, Zhang Z, Ong SH, Schlessinger J. A novel positive feedback loop mediated by the docking protein Gab1 and phosphatidylinositol 3-kinase in epidermal growth factor receptor signaling. Mol. Cell. Biol. 2000; 20:1448-1459.
18. Schulze A, Lehmann K, Jefferies HB, McMahon M, Downward J. Analysis of the transcriptional program induced by Raf in epithelial cells. Genes Dev. 2001; 15:981-994.
19. Honda K, Yanai H, Takaoka A, Taniguchi T. Regulation of the type I IFN induction: a current view. Int. Immunol. 2005; 17:1367-1378.
20. Agazie YM, Hayman MJ. Molecular mechanism for a role of SHP2 in epidermal growth factor receptor signaling. Mol. Cell. Biol. 2003; 23:7875-7886.
21. Papin JA, Hunter T, Palsson BO, Shankar S. Reconstruction of cellular signaling networks and analysis of their properties. Nat. Rev. Mol. Cell Biol. 2005; 6:99-111.
22. Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D, Alon U. Network motifs: simple building blocks of complex networks. Science 2002; 298:824-827.
23. Kohda D, Odaka M, Lax I, Kawasaki H, Suzuki K, Ullrich A, Schlessinger J, Inagaki F. A 40-kDa epidermal growth factor/transforming growth factor alpha-binding domain produced by limited proteolysis of the extracellular domain of the epidermal growth factor receptor. J. Biol. Chem. 1993; 268:1976-1981.
24. Burgess AW, Cho H-S, Eigenbrot C, Ferguson KM, Lemmon MA, Sliwkowski MS, Ward CW, Yokoyama S. An open-and-shut case? Recent insights into the activation of EGF/ErbB receptors. Mol. Cell. 2003; 12:541-552.
25. Hornung V, Rothenfusser S, Britsch S, Krug A, Jahrsdorfer B, Giese T, Endres S, Hartmann G. Quantitative expression of toll-like receptor 1-10 mRNA in cellular subsets of human peripheral blood mononuclear cells and sensitivity to CpG oligodeoxynucleotides. J. Immunol. 2002; 168:4531-4537.
26. Zarember KA, Godowski PJ. Tissue expression of human Tolllike receptors and differential regulation of Toll-like receptor mRNAs in leukocytes in response to microbes, their products, and cytokines. J. Immunol. 2002; 168:554-561.
27. Schena M, Shalon D, Davis RW, Brown PO. Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 1995; 270:467-470.
28. Buck MJ, Lieb JD. ChIP-chip: considerations for the design, analysis, and application of genome-wide chromatin immuno- precipitation experiments. Curr. Opin. Genet. Dev. 2006; 16:157- 164.
29. Orian A. Chromatin profiling, DamID and the emerging landscape of gene expression. Curr. Opin. Genet. Dev. 2006; 16:157-164.
30. Bulyk ML. Analysis of sequence specificities of DNA-binding proteins with protein binding microarrays. Methods Enzymol. 2006; 410:279-299.
31. Parrish JR, Gulyas KD, Finley RL. Yeast two-hybrid contributions to interactome mapping. Curr. Opin. Biotechnol. 2006; 17:387-393.
32. Ong SE, Foster LJ, Mann M. Mass spectrometric-based approaches in quantitative proteomics. Methods 2003; 29:124-130.
33. Jones RB, Gordus A, Krall JA, MacBeath G. A quantitative protein interaction network for the ErbB receptors using protein microarrays. Nature 2006; 439:168-174.
34. Schulze W.X., Deng L., Mann M. Phosphotyrosine interactome of the ErbB-receptor kinase family. Mol. Syst. Biol. 2005; 11-13.
35. Blagoev B, Ong SE, Kratchmarova I, Mann M. Temporal analysis of phosphotyrosine-dependent signaling networks by quantitative proteomics. Nat. Biotechnol. 2004; 22:1139-1145.
36. Takamori S, Holt M, Stenius K, Lemke EA, Gronborg M, Riedel D, Urlaub H, Schenck S, Brugger B, Ringler P, et al. Molecular anatomy of a trafficking organelle. Cell 2006; 127:831-846.
37. Gilchrist A, Au CE, Hiding J, Bell AW, Fernandez-Rodriguez J, Lesimple S, Nagaya H, Roy L, Gosline SJ, Hallett M, et al. Quantitative proteomics analysis of the secretory pathway. Cell 2006; 127:1265-1281.
38. Spring DR. Chemical genetics to chemical genomics: small molecules offer big insights. Chem. Soc. Rev. 2005; 34:472-482.
39. Knight ZA, Shokat KM. Chemical genetics: where genetics and pharmacology meet. Cell 2007; 128:425-430.
40. Degterev A, Huang Z, Boyce M, Li Y, Jagtap P, Mizushima N, Cuny GD, Mitchison TJ, Moskowitz MA, Yuan J. Chemical inhibitor of nonapoptotic cell death with therapeutic potential for ischemic brain injury. Nat. Chem. Biol. 2005; 1:112-119.
41. Diamandis P, Wildenhain J, Clarke ID, Sacher AG, Graham J, Bellows DS, Ling EK, Ward RJ, Jamieson LG, Tyers M, Dirks PB. Chemical genetics reveals a complex functional ground state of neural stem cells. Nat. Chem. Biol. 2007; 268-273.
42. Anderson C, Bartlett SJ, Gansner JM, Wilson D, He L, Gitlin JD, Kelsh RN, Dowden J. Chemical genetics suggests a critical role for lysyl oxidase in zebrafish notochord morphogenesis. Mol. Biosyst. 2007; 3:51-59.
43. Sato S, Kwon Y, Kamisuki S, Srivastava N, Mao Q, Kawazoe Y, Uesugi M. Polyproline-rod approach to isolating protein targets of bioactive small molecules: isolation of a new target of indomethacin. J. Am. Chem. Soc. 2007; 129:873-880.
44. Larochelle S, Merrick KA, Terret ME, Wohlbold L, Barboza NM, Zhang C, Shokat KM, Jallepalli PV, Fisher RP. Requirements for Cdk7 in the assembly of Cdk1/cyclin B and activation of Cdk2 revealed by chemical genetics in human cells. Mol. Cell. 2007; 25:839-850.
45. Jaeschke A, Karasarides M, Ventura JJ, Ehrhardt A, Zhang C, Flavell RA, Shokat KM, Davis RJ. JNK2 is a positive regulator of the cJun transcription factor. Mol. Cell. 2006; 23:899-911.
46. Ventura JJ, Hubner A, Zhang C, Flavell RA, Shokat KM, Davis RJ. Chemical genetic analysis of the time course of signal transduction by JNK. Mol. Cell. 2006; 21:701-710.
47. Kenski DM, Zhang C, von Zastrow M, Shokat KM. Chemical genetic engineering of G protein-coupled receptor kinase 2. J. Biol. Chem. 2005; 280:35051-35061.
48. Shah K, Shokat KM. A chemical genetic screen for direct v-Src substrates reveals ordered assembly of a retrograde signaling pathway. Chem Biol. 2002; 9:35-47.
49. Burkard ME, Randall CL, Larochelle S, Zhang C, Shokat KM, Fisher RP, Jallepalli PV. Chemical genetics reveals the requirement for Polo-like kinase 1 activity in positioning RhoA and triggering cytokinesis in human cells. Proc. Natl. Acad. Sci. U.S.A. 2007; 104:4383-4388.
50. Tolliday N, Clemons PA, Ferraiolo P, Koehler AN, Lewis TA, Li X, Schreiber SL, Gerhard DS, Eliasof S. Small molecules, big players: the National Cancer Institute’s Initiative for Chemical Genetics. Cancer Res. 2006; 66:8935-8942.
51. Garber K. Big plans for small molecules: NIH launches chemical genomics initiative. J. Natl. Cancer Inst. 2003; 95:1740-1741.
52. Tan DS. Diversity-oriented synthesis: exploring the intersections between chemistry and biology. Nat. Chem. Biol. 2005; 1:74-84.
53. Ede NJ, Wu Z. Beyond Rf tagging. Curr. Opin. Chem. Biol. 2003; 7:374-379.
54. Linclau B, Sing AK, Curran DP. Organic-Fluorous phase switches: a fluorous amine scavenger for purification in solution phase parallel synthesis. J Org Chem. 1999; 64:2835-2842.
55. Parlow JJ. Polymer-assisted solution-phase chemical library synthesis. Curr. Opin. Drug Discov. Devel. 2005; 8:757-775.
56. Parlow JJ, Devraj RV, South MS. Solution-phase chemical library synthesis using polymer-assisted purification techniques. Curr. Opin. Chem. Biol. 1999; 3:320-336.
57. AfCS. http://www.signaling-gateway.org/.
58. BIND. http://bond.unleashedinformatics.com/index.jsp?pg=0.
59. BioCarta. http://www.biocarta.com/.
60. DIP. http://dip.doe-mbi.ucla.edu/.
61. HPRD. http://www.hprd.org/.
62. STKE. http://stke.sciencemag.org/.
63. Swiss-Prot. http://expasy.org/sprot/.
64. Sarkar A, Beard DA, Franza BR. Effect of binding in cyclic phosphorylation-dephosphorylation process and in energy transformation. Math. Biosc. 2006; 202:175-193.
65. Eungdamrong NJ, Iyengar R. Compartment-specific feedback loop and regulated trafficking can result in sustained activation of Ras at the golgi. Biophys. J. 2007; 92:808-815.
66. Hayer A, Bhalla US. Molecular switches at the synapse emerge from receptor and kinase traffic. PLOS Comp. Biol. 2005; 1:137-87.
67. Hoffmann A, Levchenko A, Scott ML, Baltimore D. The IkappaB-NF-kappaB signaling module: temporal control and selective gene activation. Science 2002; 298:1241-1245.
68. Lander AD, Nie Q, Wan FYM. Do morphogen gradients arise by diffusion? Dev. Cell. 2002; 2:785-796.
69. Lee S E, Hori Y, Chakraborty A. Low T cell receptor expression and thermal fluctuations contribute to formation of dynamic multifocal synapses in thymocytes. Proc. Natl. Acad. Sci. U.S.A. 2003; 100:4383-4388.
70. Ma L, Wagner J, Rice JJ, Hu W, Levince AJ, Stolovitzky G. A plausible model for the digital response of p53 to DNA damage. Proc. Natl. Acad. Sci. U.S.A. 2005; 102:14266-14271.
71. Bhalla US. Signaling in small subcellular volumes. I. Stochastic and diffusion effects on individual pathways. Biophys. J. 2004; 87:733-744.
72. Bhalla US. Signaling in small subcellular volumes. II. Stochastic and diffusion effects on synaptic network properties. Biophys. J. 2004; 87:745-753.
73. Resat H, Ewald JA, Dixon DA, Wiley HS. An integrated model of epidermal growth factor receptor trafficking and signal transduction. Biophys. J. 2003; 85:730-743.
74. Albert R, Othmer HG. The topology of the regulatory interactions predicts the expression pattern of the segment polarity genes in Drosophila melanogaster. J. Theor. Biol. 2003; 223:1-18.
75. Gonzalez A, Chaouiya C, Thieffry D. Dynamical analysis of the regulatory network defining the dorsal-ventral boundary of the Drosophila wing imaginal disc. Genetics 2006; 174:1625-1634.
76. Mayaan A, Lipshtat A, Iyengar R. Topology of resultant networks shaped by evolutionary pressure. Phys Rev E Stat Nonlin Soft Matter Phys. 2006; 73(6 Pt 1):061912.
77. Friedman N. Inferring cellular networks using probabilistic graphical models. Science 2004; 303:799-805.
78. Sachs K, Perez O, Peer D, Lauffenburger DA, Nolan GP. Causal protein-signaling networks derived from multiparameter single-cell data. Science 2005; 308:523-529.
79. Sarkar A, Franza BR. A logical analysis of the process of T cell activation: different consequences depending on the state of CD28 engagement. J. Theor. Biol. 2004; 226:455-466.
80. Sanchez L, Thieffry D. A logical analysis of the Drosophila gap-gene system. J Theor Biol. 2001; 211:115-141.
81. Sanchez L, Thieffry D. Segmenting the fly embryo: a logical analysis of the pair-rule cross-regulatory module. J Theor Biol. 2003; 224:517-537.
82. Belacel N, Wang Q, Cuperlovic-Culf M. Clustering methods for microarray gene expression data. OMICS. 2006; 10:507-531.
83. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of genome-wide expression patterns. Proc. Natl. Acad. Sci. U.S.A. 1998; 95:14863-14868.
84. Tavazoie S, Hughes JD, Campbell MJ, Cho RJ, Church GM. Systematic determination of genetic network architecture. Nat. Gen. 1999; 22:281-285.
85. Wang J, Delabie J, Aasheim H, Smeland E, Myklebost O. Clustering of the SOM easily reveals distinct gene expression patterns: results of a reanalysis of lymphoma study. BMC Bioinformatics. 2002; 3:36.
86. Nielsen TO, West RB, Linn SC, Alter O, Knowling MA, O’Connell JX, Zhu S, Fero M, Sherlock G, Pollack JR, Brown PO, Botstein D, Rijn Mvd. Molecular characterization of soft tissue tumors: a gene expression study. Lancet 2002; 359:1301-1307.
87. Shai R, Shi T, Kremen T, Horvath S, Liau L, Cloughesy T, Mischel P, SF N. Gene expression profiling identifies molecular subtypes of gliomas. Oncogene 2003; 22:4918-4923.
88. Alter O, Brown PO, Botstein D. Singular value decomposition for genome-wide expression data processing and modeling. Proc. Natl. Acad. Sci. U.S.A. 2000; 97:10101-10106.
89. Calvano SE, Xiao W, Richards DR, Felciano RM, Baker HV, Cho RJ, Chen RO, Brownstein BH, Cobb JP, Tschoeke SK, et al. A network-based analysis of systemic inflammation in humans. Nature 2005; 437:1032-1037.
90. Holter NS, Maritan A, Cieplak M, Fedoroff NV, Banavar JR. Dynamic modeling of gene expression data. Proc. Natl. Acad. Sci. U.S.A. 2001; 98:1693-1698.
91. Yeung MK, Tegner J, Collins JJ. Reverse engineering gene networks using singular value decomposition and robust regression. Proc. Natl. Acad. Sci. U.S.A. 2002; 99:6163-6168.
92. Janes KA, Albeck JG, Gaudet S, Sorger PK, Lauffenburger DA, Yaffe MB. A systems model of signaling identifies a molecular basis set for cytokine-induced apoptosis. Science 2005; 310:1646-1653.
93. Barabasi AL, Oltvai ZN. Network biology understanding the cell’s functional organization. Nat. Rev. Gen. 2004; 5:101-113.
94. Shen-Orr S, Milo R, Mangan S, Alon U. Network motifs in the transcriptional regulation network of Escherichia coli. Nat. Gen. 2002; 31:64-68.
95. Gregg JR. Ones and Zeros: Understanding Boolean Algebra, Digital Circuits and the Logic of Sets. 1998. John Wiley & Sons, Inc., New York.
96. Shiva SG. Introduction to Logic Design. 1998. CRC Press, Boca Raton, FL.
97. Sarkar A, Franza BR. A logical analysis of the process of T cell activation: different consequences depending on the state of CD28 engagement. J. Theor. Biol. 2004; 226:455-466.
98. Haugh JM, Lauffenburger DA. Analysis of receptor internalization as a mechanism for modulating signal transduction. J. Theor. Biol. 1998; 195:187-218.
99. Lauffenburger DA. Receptors: models for binding, trafficking, and signaling. 1993. Oxford University Press, Oxford.
100. Schoeberl, B., Eichler-Jonsson C. Gilles, E.D. and Muller, G. Analysis of receptor internalization as a mechanism for modulating signal transduction. Nat. Biotech. 2002; 2037-375.
101. Starbuck C, Lauffenburger DA. Mathematical model for the effects of epidermal growth factor receptor trafficking dynamics on fibroblast proliferation responses. Biotechnol. Prog. 1992; 8:132-143.
102. Waters CM, Oberg KC, Carpenter G, Overholser KA. Rate constants for binding, dissociation, and internalization of EGF: effect of receptor occupancy and ligand concentration. Biochemistry 1990; 29:3563-3569.
103. Bhalla U.S. Iyengar R. Emergent properties of networks of biological signaling pathways. Science. 1999; 283:381-387.
104. Ferrel J.E.a.M. E.M., The biochemical basis of an all-or-none cell fate switch in xenopus oocytes. 1998 Science.; 1280: 895-898.
105. Kholodenko BN. Negative feedback and ultrasensitivity can bring about oscillations in the mitogen-activated protein kinase cascades. Eur. J. Biochem. 2000; 267:1583-1588.
106. Bentele M, Lavrick I, Ulrich M, Stober S, Heermann DW, Kalthoff H, Krammer PH, Elis R. Mathematical modeling reveals threshold mechanism in CD95-induced apoptosis. J. Cell. Biol. 2004; 166:839-851.
107. Gillespie DT. A general method for numerically simulating the stochastic time evolution of coupled chemical reactions. J. Comp. Physics. 1976; 22:403-434.
108. Samiolov M, Plyasunov S, Arkin AP. Stochastic amplification and signaling in enzymatic futile cycles through noise-induced bistability with oscillations. Proc. Natl. Acad. Sci. U.S.A. 2005; 102:2310-2315.
109. Strong M, Eisenberg D. The protein network as a tool for finding novel drug targets. Prog. Drug. Res. 2007;64:193-215.
110. Kumar N, Hendriks BS, Janes KA, de Graaf D, Lauffenburger DA. Applying computational modeling to drug discovery and development. Drug Discov. Today. 2006; 11:806-811.
111. Chen JY, Shen C, Yan Z, Brown DP, Wang M. A systems biology case study of ovarian cancer drug resistance. Comput Syst Bioinformatics Conf. 2006; 389-398.
112. Greco WR, Bravo G, Parsons JC. The search for synergy: a critical review from a response surface perspective. Pharmacol. Rev. 1995; 47:331-385.
113. Alberti S, Parodi S. Signaling protein networks as targets of new antineoplastic drugs. Int. J. Biol. Markers. 2003; 18:57-61.
114. Lehar J, Zimmermann GR, Krueger AS, Molnar RA, Ledell JT, Heilbut AM, Short GF 3rd, Giusti LC, Nolan GP, Magid OA, Lee MS, Borisy AA, Stockwell BR Keith CT. Chemical combination effects predict connectivity in biological systems. Mol. Syst. Biol. 2007; 380.
Alon U. An Introduction to Systems Biology: Design Principles of Biological Circuits. 2006. Chapman & Hall/CRC Press, Boca Raton, FL.
Helmreich E. Biochemistry of Cell Signaling, Oxford University Press, 2001.
Ma’ayan A, Blitzer RD, Iyengar R. Toward predictive models of mammalian cells. Annu. Rev. Biophys. Biomol. Struct. 2005; 3431934349.
Murray JD. Mathematical Biology: I. An Introduction. Springer, New York. 2002.
Murray JD. Mathematical Biology: II. Spatial Models and Biomedical Applications. Springer, New York. 2003.
Rangamani P, Iyengar R. Modelling spatio-temporal interactions within the cell. J. Biosci. 2007; 32:157-167.
Computational Chemistry in Biology
Peptides Chemistry of
Cyclic AMP in Signaling
Cell Cycle, Computation and Modeling of
G-Protein Coupled Receptor (GPCR) Signaling
Receptor Tyrosine Kinases
Array-Based Techniques: Proteins
Chemical Libraries: Screening for Biologically Active Small Molecules
Synthetic Networks in Living Systems: Topics in Chemical Biology