Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Nat Immunol. Author manuscript; available in PMC 2009 November 12.
Published in final edited form as:
Published online 2009 March 19. doi:  10.1038/ni.1711
PMCID: PMC2776706

Navigating the network: signaling cross-talk in hematopoietic cells


Recent studies in hematopoietic cells have led to a growing appreciation of the diverse modes of molecular and functional cross-talk between canonical signaling pathways. However, these intersections represent only the tip of the iceberg. Emerging global analytical methods are providing an even richer and more complete picture of the many components that measurably interact in a network manner to produce cellular responses. Here we highlight the pieces in this Focus, emphasize the limitations of the present canonical pathway paradigm, and discuss the value of a systems biology approach using more global, quantitative experimental design and data analysis strategies. Lastly, we urge caution about overly facile interpretation of genome- and proteome-level studies.

Past is prologue

The ability of cells to respond dynamically to their environment through engagement of cell surface receptors is a fundamental principle of biology, and a breakdown in cellular information processing involving these receptors and their linked intracellular signal transduction events underlies many diseases. This combination of basic and clinical importance has led to a substantial effort over the past 50 years to elucidate the mechanistic principles of cellular signaling.

The concept of a ‘receptor’ dates back to the first decade of the twentieth century and studies of the action of adrenergic agonists derived from “supra-renal” extract1. Classical pharmacological analyses of dose–response relationships dominated cellular research for the next 50 years (ref. 2). Biochemical methods emerging in the 1960s and 1970s led to seminal breakthroughs in the field, with the identification of cyclic AMP as the first ‘second messenger’3 and the purification and functional characterization of the components of the first G protein–coupled receptor pathways46. These studies identified reversible modification of proteins, through serine/threonine phosphorylation in this case, as a primary mechanism of signal relay and amplification, giving birth to the concept of the linear signal transduction pathway (Fig. 1a).

Figure 1
Topology of signaling networks and interpretation of experimental readouts. (a) Representation of signaling pathway schema from simple linear pathways (Rx), to branched pathways downstream of multicomponent receptor complexes (Ry), to signaling networks ...

The emergence of molecular cloning in the 1980s accelerated the identification of signaling pathways, and the discovery of tyrosine phosphorylation7 and the consequent identification of adaptor proteins with modular phosphotyrosine binding domains introduced the concept of signal propagation through modification-induced protein-protein interaction8. Receptors or noncovalent receptor complexes with many sites for stimulus-induced interaction9, severalof which were identified in hematopoietic cells10,11, were shown to initiate the formation of multicomponent complexes. Such assemblies could activate many signaling events, and their characterization revealed the prevalence of branched signaling pathways (Fig. 1a). Molecular methods to screen for protein-protein interactions12, in conjunction with fluorescent protein–based technology to observe the subcellular localization and dynamics of signaling proteins in live cells13, provided further insight into the topology of these branched pathways.

Advances in mass spectrometry and proteomics permitted a more global assessment of the protein14 and lipid modifications15,16 that result from cell stimulation. This led to the realization that single stimuli often provoke complex responses not predictable from known signaling schema, heralding the beginnings of network biology (Fig. 1a). Initial networks emerged from the addition of discrete points of cross-talk between canonical signaling pathways involving shared components1719, but the growing scale of the data sets generated by high-throughput methods and the density of interconnections20,21 revealed by these new ‘omic’ approaches have necessitated more sophisticated methods for data analysis and display2225. This has taken the field into the realm of computational systems biology, network analysis and display, and statistical inference of signal propagation routes2629.

This brings us back to the subject of the articles in this issue of Nature Immunology. They cover in substantial detail a variety of subfields in the area of immune signal transduction, with an emphasis on the frequent and diverse nature of cross-talk and component sharing among signaling pathways in hematopoietic cells (Ivashkiv30; Wilson, Dixit and Ashkenazi31; Bezbradica and Medzhitov32) and on current approaches for genetic (Saveliev and Tybulewicz33) or therapeutic (Ghoreschi, Laurence and O'Shea34) manipulation of signaling components. Here we try to place the main themes of these articles in the context of the emerging network-based view of cellular signaling, and to emphasize the growing need for careful experimental design to maximize the physiological relevance and proper interpretation of data acquired using modern data-rich technologies.

From pathways to networks

Canonical signaling pathways provide a fundamental organizational and intellectual framework for understanding the nature of biological responses at the cellular level. However, there are also substantial drawbacks to the dominance of the canonical pathway paradigm. The principal problem is that the experimental design of most studies is highly biased by what we presume to already ‘know’; as a result, the perceived importance of canonical pathway components is reinforced by their repeated use as the primary or only readout in many experiments. Thus, just as each of the six blind men can interpret only the part of the elephant he touches, reductionist biological approaches based narrowly on existing pathway dogma fail to provide a comprehensive account of complex cellular responses because these approaches are inherently self-reflexive.

Modern technology now allows investigators to probe the activities of a cell in a much broader, less dogmatic manner. Multiplex fluorescent technologies permit the assessment of multiple readouts from a stimulated cell35, RNA interference can perturb gene expression on a genome-wide scale36,37, new-generation sequencing technologies promise to substantially improve the accuracy and data output from transcript profiling38,39, and advances in mass spectrometry are refining our knowledge of biochemical events in stimulated cells by assessing in a highly refined manner a very large fraction of a cell's constituents and their modifications during signal transduction40.

A recurrent theme emerging from recent studies is the marked increase in the number of components that influence a phenotypic outcome that was previously linked to operation of a ‘canonical pathway’41,42. Rather than the tens of components previously thought to participate in signaling through a typical pathway, recent screens often suggest the involvement of hundreds. Moreover, these components are often identified based on arbitrary thresholds set for practical purposes, suggesting that the actual number of potential regulators could be even higher. In a prescient essay on the implications of these data41, Friedman and Perrimon recently highlighted reports showing that screens with quantitative phenotypic readouts often produce a ‘continuous’ gene–versus-phenotype relationship, one in which every tested gene has some degree of influence on the measured outcome (Fig. 1b). One interpretation of these findings is that the cell operates as a single mega-network with limited isolation between the hubs of what would traditionally be considered ‘pathways’. However, these data alone do not rule out a core signaling architecture based on canonical, relatively linear, isolated pathways.

The application of such global screens to immune cells could provide a new picture of how single and conjoint stimuli are processed within the cell to result in the measurable changes we associate with such perturbations. As just one example, the present analysis of cytokine signaling in instances that involve Jak–STAT pathways may be a classic example of ‘canonical bias,’ wherein the role of new components in cytokine responses remains unappreciated because a small, standard set of Jak–STAT components is repeatedly used as an assay readout. Such ‘hidden’ components may help to explain how particular cytokine receptors have the capacity to drive cell type–specific responses, despite their shared usage of a small set of Jak and STAT proteins. This issue is discussed in some detail in the review by Bezbradica and Medzhitov32. The argument in favor of a less ‘pathway-centric’ view of signaling is also supported by the identification of components in large-scale screens that are well established participants in ‘other’ canonical signaling pathways41. This principle is especially relevant to the review by Wilson, Dixit and Ashkenazi31, which discusses the involvement of members of the tumor necrosis factor receptor family and death domain proteins in both apoptotic and non-apoptotic pathways.

A noteworthy feature of many immune cell networks is their convergence on common downstream transcriptional modules (Fig. 1a). How do different upstream signaling modules share the same downstream components when both inputs are activated simultaneously? Do the signaling modules draw from a common pool, or are preformed complexes dedicated to a given pathway? The latter case would actually argue against the use of common components as cross-talk nodes, and certain examples of immunoreceptor tyrosine-based activation motif (ITAM)-based pathway interactions support this notion. The NF-κB and MAPK signaling complexes are activated by ITAM-based signaling, but also by tumor necrosis factor receptor family signaling and many innate immune signaling pathways. Nonetheless, as illustrated in the review by Ivashkiv30, ITAM pathway cross-talk more commonly involves components of the Ca2+ pathway activated by ITAM signaling. This situation may be somewhat specific to ITAM-based signaling, as these pathways depend heavily on co-stimulation through heterologous pathways to achieve signal flow. Further studies on pathway cross-talk will be required to determine whether shared components are more likely or less likely to represent network nodes or hubs.

A word of caution

Having emphasized the potential value of a ‘connected network’ view of cellular signaling, some caution about overinterpretation of global studies is warranted. Large-scale connectivity among signaling components may not always be the case. For example, recent studies have identified roles for death receptor (DR) signal transducers in non-DR innate signaling pathways. This may simply represent reuse of a modular signaling adaptor molecule in an alternative pathway, rather than a point of cross-talk between DR and non-DR signaling, as there is no evidence yet to demonstrate that the use of DR transducers in the non-DR pathway directly influences DR signaling. It is possible that the lack of experimental evidence of cross-talk between DR signaling and other immune cell pathways indicates a requirement for hematopoietic cells to limit influence on the apoptotic cascade because of this pathway's fundamental importance in removal of autoreactive lymphocytes.

Such findings raise questions about the notion of a fully connected network of the type pictured by Friedman and Perrimon41, suggesting instead that there remains some value in the more conventional ‘pathway’ picture of signal transduction and hinting that effective inter-pathway insulation, possibly facilitated by scaffold proteins (Fig. 1a), is needed to avoid spurious responses to particular stimuli. Continuous function graphs assign a degree of influence to every gene in a given response if the measured effect is above some statistical threshold. However, such analyses do not directly address the question of whether the effect measured is physiological, in the sense that it has evolved for a specific, necessary role in cellular behavior. It is quite possible that the effect represents only some inescapable degree of ‘spillover’ or ‘noise’ between evolved pathways, in which the energetic and other costs of establishing greater interpathway isolation are higher than the selection benefit. As a simple example, consider a kinase whose catalytic activity on two substrates differs by more than tenfold, with the more sensitive target being within a canonical signaling pathway and the less sensitive one in a distinct pathway. At modest levels of input that activate only a limited amount of the kinase, only the more sensitive target will be modified to an extent that changes cell behavior. Under most physiological circumstances, this means that only the canonical pathway will be engaged in a biologically meaningful way. However, saturating supraphysiologic inputs may induce modification of the second substrate sufficient to result in a measurable phenotypic effect (Fig. 1c). Thus, assays using supraphysiologic inputs may not necessarily reflect how the cell behaves in its normal state. Put another way, it is important to distinguish between that which can happen and that which evolution has specifically selected in ensuring proper organismal functioning under nonpathologic circumstances. In many cases, there may be no selective advantage to the evolution of strict pathway isolation of the type highlighted in the DR and death-domain pathways; indeed, the development of protein structures that enforce such isolation might actually be deleterious to the operation of the ‘canonical’ pathway to which these elements belong. The upshot is that we may be able to measure weak signal spreading within the ‘network’ in very precise experiments without such ‘signaling ripples’ being fundamental to the cell's proper interpretation of incoming stimuli.

The importance of context and state

These considerations point to a key challenge in improving our understanding of signal integration in mammalian cells—the development of enhanced experimental design practices that address the importance of cellular context, physiological conditions of stimulation, and minimal perturbation of cellular content and molecular structure. Although it is clear that immune cell signaling pathways are connected by physical and functional interactions, how do we more accurately assess the physiological relevance of these connections? Most published studies are characterized by one or all of the following experimental strategy elements: acute stimulation, (near to) saturating ligand concentration, overexpressed components, and/or use of a heterologous cell system. Because physiological stimuli are typically subsaturating with respect to available receptor pools, often transient or pulsatile in nature, and have consequences that are highly susceptible to the effects of cellular composition43, the experimental methods in common use raise the obvious question of the extent to which the results reflect the behavior of the cell under more physiologic conditions versus the outcome of extreme perturbations.

To a large extent, this experimental design paradigm has been driven by practical necessity. Such unphysiological conditions are often required to elicit a consistently measurable response from cells that invest heavily in dampening and negative regulation, a characteristic that may be an even greater factor in immortalized cell lines. Although this is helpful in generating robust data that are convincing in publications, there is great value to developing and using more sensitive multiplex methodologies with a broad range of readouts that permit experimentation under more realistic conditions of cell stimulation. Emerging sensor-based assays using 384- and 1,536-well microplates to facilitate genome-wide perturbation screens44 and the use of microfabricated devices to permit single cell measurements with high resolution45 will also allow an increase in the breadth and dynamic range of measurements involving non-transformed cells available in limited number. Application of such methods to network analysis will permit an increase in sensitivity, analysis at the single-cell level with time resolution, and an increased number of tested ligand concentrations and ligand combinations, which together should markedly enhance the likelihood that we will observe events relevant to physiology. The next step in the evolution of such analyses will be making the same sets of observations in vivo, using imaging methods that permit observation of cells interacting within the tissue environment and involving physiologic amounts of mediators in that natural setting46,47.

Assessment of cellular responses to combined rather than single stimuli is also a generally neglected but vitally important experimental design parameter in efforts to unravel complex signaling events48. For example, there is growing evidence that cytokine receptors activate non–Jak-STAT signaling effectors (reviewed by Bezbradica and Medzhitov32), but little is known about how this influences activation of these effectors through their canonical pathways when they become activated by non-cytokine stimuli also present in the cell's environment. Also, pathogens represent a complex stimulus, invariably activating several pathways. These stimuli often combine to produce a synergistic response that is not predictable from the output evoked by an individual canonical pathway downstream of a single receptor.

Careful modeling of quantitative data would also be valuable for explaining observed phenomena. For example, Ghoreschi, Laurence and O'Shea34 discuss the unexpected success of broad-spectrum kinase inhibitors such as staurosporine for treatment of various malignancies, achieved even though these drugs target a substantial proportion of the entire kinome. This success may relate to differences in network flux between normal and transformed cells. Normal cells may be capable of tolerating, or recovering from, inhibition of large portion of their cellular kinases. Tumor cells, by contrast, may require more kinase-driven pathways operating close to their maximum capacity to maintain the transformed state, and they may thus be more susceptible to a partial reduction in signaling capacity through multiple routes. A comprehensive recent assessment of the kinase dependencies of many cell types shows notable heterogeneity in kinase requirements49, pointing to a need for caution in predicting the physiological effects of kinase inhibitors. This also highlights the possibility, discussed in the review34, that combination therapies targeting multiple kinases are likely to be more fruitful, considering the interconnected nature of at least portions of the signaling network. Dual perturbation increases the likelihood that the drug treatment will inhibit signal flow (because of the branched nature of the signaling network) and lead to a significant effect, and it also may reduce the incidence of drug resistance as the network is perturbed at multiple points. The potential of this strategy has led to the recent development of screens specifically designed to identify dual-target inhibitors50.

On a related note, Saveliev and Tybulewicz33 describe the value of partial loss-of-function mutants in assessing the various functions of the ZAP70 tyrosine kinase in T cell antigen receptor (TCR) signaling, demonstrating that different amounts of kinase activity are required for different aspects of T cell development. These latter studies emphasize two key points. First is the importance of quantitative parameter recording. The contribution of ZAP70 is not a digital phenomenon, but instead resembles a rheostat in its control of signaling output from the TCR complex. Second, partial loss-of-function can provide important quantitative insights into signal flow, emphasizing the value of RNA-mediated interference (RNAi) technology—especially viral short hairpin RNA (shRNA) expression systems that permit the conditional knockdown of targets and reexpression of mutant proteins even in moderately intractable hematopoietic cells5153—as a complement to knockout and knock-in strategies.

Finally, on the qualitative side of the equation, the more sophisticated gene perturbation technologies discussed by Saveliev and Tybulewicz33 hold particular promise for improving our ability to identify the normal physiologic function of a protein and its domains: the selective ablation of a single property in a multifunctional signaling protein will more accurately describe its function within a network than the complete removal of this protein from the cell by classical knockout methods. The findings emerging from such research highlight the value of therapies that combine inhibitors of catalytic function with inhibitors of protein-protein interactions.

Concluding remarks

There is now a debate about whether we should abandon the canonical pathway concept for a network picture of cellular function in which nearly all molecular elements influence one another to some measurable degree. However, caution is needed to avoid confusing what our techniques are now capable of measuring and what actually guides cell behavior under physiologically relevant conditions.

Refined experimental approaches need to be combined with recent advances in mathematical tools and computer software that permit the generation of quantitative computer models of complex biological systems across scales of organization and with a high degree of correspondence to biological reality—from molecular interactions underlying intracellular signaling pathways, to gene regulatory networks interfacing with these signaling events and also regulated by epigenetic modifications, to the cross-talk between cells as is critical in the immune system in particular, to organ behavior, and finally, to the overall response of an organism5457. The resulting simulations will allow us to make predictions about the anticipated results of experiment and to refine our understanding in an iterative manner. An exciting path lies ahead for ‘systems immunology’, but we must tread carefully.


Supported by the Intramural Research Program of the US National Institutes of Health, National Institute of Allergy and Infectious Diseases.


Reprints and permissions information is available online at


1. Langley JN. Observations on the physiological action of extracts of the supra-renal bodies. J. Physiol. (Lond.) 1901;27:237–256. [PubMed]
2. Furchgott RF. Receptor mechanisms. Annu. Rev. Pharmacol. 1964;4:21–50.
3. Rall TW, Sutherland EW. Formation of a cyclic adenine ribonucleotide by tissue particles. J. Biol. Chem. 1958;232:1065–1076. [PubMed]
4. Walsh DA, Perkins JP, Krebs EG. An adenosine 3′,5′-monophosphate-dependant protein kinase from rabbit skeletal muscle. J. Biol. Chem. 1968;243:3763–3765. [PubMed]
5. Rodbell M, Birnbaumer L, Pohl SL, Krans HM. The glucagon-sensitive adenyl cyclase system in plasma membranes of rat liver. V. An obligatory role of guanylnucleotides in glucagon action. J. Biol. Chem. 1971;246:1877–1882. [PubMed]
6. Ross EM, Gilman AG. Resolution of some components of adenylate cyclase necessary for catalytic activity. J. Biol. Chem. 1977;252:6966–6969. [PubMed]
7. Hunter T, Sefton BM. Transforming gene product of Rous sarcoma virus phosphorylates tyrosine. Proc. Natl. Acad. Sci. USA. 1980;77:1311–1315. [PubMed]
8. Pawson T, Gish GD. SH2 and SH3 domains: from structure to function. Cell. 1992;71:359–362. [PubMed]
9. Ullrich A, Schlessinger J. Signal transduction by receptors with tyrosine kinase activity. Cell. 1990;61:203–212. [PubMed]
10. Reth M. Antigen receptors on B lymphocytes. Annu. Rev. Immunol. 1992;10:97–121. [PubMed]
11. Weiss A. T cell antigen receptor signal transduction: a tale of tails and cytoplasmic protein-tyrosine kinases. Cell. 1993;73:209–212. [PubMed]
12. Fields S, Song O. A novel genetic system to detect protein-protein interactions. Nature. 1989;340:245–246. [PubMed]
13. Tsien RY. The green fluorescent protein. Annu. Rev. Biochem. 1998;67:509–544. [PubMed]
14. Olsen JV, et al. Global, in vivo, and site-specific phosphorylation dynamics in signaling networks. Cell. 2006;127:635–648. [PubMed]
15. Liscovitch M, Cantley LC. Lipid second messengers. Cell. 1994;77:329–334. [PubMed]
16. Forrester JS, Milne SB, Ivanova PT, Brown HA. Computational lipidomics: a multiplexed analysis of dynamic changes in membrane lipid composition during signal transduction. Mol. Pharmacol. 2004;65:813–821. [PubMed]
17. Jordan JD, Landau EM, Iyengar R. Signaling networks: the origins of cellular multitasking. Cell. 2000;103:193–200. [PubMed]
18. Pawson T. Specificity in signal transduction: from phosphotyrosine-SH2 domain interactions to complex cellular systems. Cell. 2004;116:191–203. [PubMed]
19. Irish JM, et al. Single cell profiling of potentiated phospho-protein networks in cancer cells. Cell. 2004;118:217–228. [PubMed]
20. Stelzl U, et al. A human protein-protein interaction network: a resource for annotating the proteome. Cell. 2005;122:957–968. [PubMed]
21. Rual JF, et al. Towards a proteome-scale map of the human protein-protein interaction network. Nature. 2005;437:1173–1178. [PubMed]
22. Albeck JG, et al. Collecting and organizing systematic sets of protein data. Nat. Rev. Mol. Cell Biol. 2006;7:803–812. [PubMed]
23. Kohn KW, Aladjem MI, Weinstein JN, Pommier Y. Molecular interaction maps of bioregulatory networks: a general rubric for systems biology. Mol. Biol. Cell. 2006;17:1–13. [PMC free article] [PubMed]
24. Cline MS, et al. Integration of biological networks and gene expression data using Cytoscape. Nat. Protoc. 2007;2:2366–2382. [PubMed]
25. Royer L, Reimann M, Andreopoulos B, Schroeder M. Unraveling protein networks with power graph analysis. PLOS Comput. Biol. 2008;4:e1000108. [PMC free article] [PubMed]
26. Hucka M, et al. The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models. Bioinformatics. 2003;19:524–531. [PubMed]
27. Xia Y, et al. Analyzing cellular biochemistry in terms of molecular networks. Annu. Rev. Biochem. 2004;73:1051–1087. [PubMed]
28. Aldridge BB, Burke JM, Lauffenburger DA, Sorger PK. Physicochemical modelling of cell signalling pathways. Nat. Cell Biol. 2006;8:1195–1203. [PubMed]
29. Zhu J, Zhang B, Schadt EE. A systems biology approach to drug discovery. Adv. Genet. 2008;60:603–635. [PubMed]
30. Ivashkiv LB. Cross-regulation of signaling by ITAM-associated receptors. Nat. Immunol. 2009;10:340–347. [PMC free article] [PubMed]
31. Wilson NS, Dixit V, Ashkenazi A. Death receptor signal transducers: nodes of coordination in immune signaling networks. Nat. Immunol. 2009;10:348–355. [PubMed]
32. Bezbradica JS, Medzhitov R. Integration of cytokine and heterologous receptor signaling pathways. Nat. Immunol. 2009;10:333–339. [PubMed]
33. Saveliev A, Tybulewicz VLJ. Lymphocyte signaling: beyond knockouts. Nat. Immunol. 2009;10:361–364. [PubMed]
34. Ghoreschi K, Laurence A, O'Shea JJ. Selectivity and therapeutic inhibition of kinases: to be or not to be? Nat. Immunol. 2009;10:356–360. [PMC free article] [PubMed]
35. Fulton RJ, McDade RL, Smith PL, Kienker LJ, Kettman JR., Jr. Advanced multiplexed analysis with the FlowMetrix system. Clin. Chem. 1997;43:1749–1756. [PubMed]
36. Berns K, et al. A large-scale RNAi screen in human cells identifies new components of the p53 pathway. Nature. 2004;428:431–437. [PubMed]
37. Paddison PJ, et al. A resource for large-scale RNA-interference-based screens in mammals. Nature. 2004;428:427–431. [PubMed]
38. Margulies M, et al. Genome sequencing in microfabricated high-density picolitre reactors. Nature. 2005;437:376–380. [PMC free article] [PubMed]
39. Shendure J, et al. Accurate multiplex polony sequencing of an evolved bacterial genome. Science. 2005;309:1728–1732. [PubMed]
40. Kocher T, Superti-Furga G. Mass spectrometry-based functional proteomics: from molecular machines to protein networks. Nat. Methods. 2007;4:807–815. [PubMed]
41. Friedman A, Perrimon N. Genetic screening for signal transduction in the era of network biology. Cell. 2007;128:225–231. [PubMed]
42. Boutros M, Ahringer J. The art and design of genetic screens: RNA interference. Nat. Rev. Genet. 2008;9:554–566. [PubMed]
43. Feinerman O, Veiga J, Dorfman JR, Germain RN, Altan-Bonnet G. Variability and robustness in T cell activation from regulated heterogeneity in protein levels. Science. 2008;321:1081–1084. [PMC free article] [PubMed]
44. Auld DS, et al. Fluorescent protein-based cellular assays analyzed by laser-scanning microplate cytometry in 1536-well plate format. Methods Enzymol. 2006;414:566–589. [PubMed]
45. Chao TC, Ros A. Microfluidic single-cell analysis of intracellular compounds. J. R. Soc. Interface. 2008;5(suppl 2):S139–S150. [PMC free article] [PubMed]
46. Germain RN, Miller MJ, Dustin ML, Nussenzweig MC. Dynamic imaging of the immune system: progress, pitfalls and promise. Nat. Rev. Immunol. 2006;6:497–507. [PubMed]
47. Shaner NC, Steinbach PA, Tsien RY. A guide to choosing fluorescent proteins. Nat. Methods. 2005;2:905–909. [PubMed]
48. Natarajan M, Lin KM, Hsueh RC, Sternweis PC, Ranganathan R. A global analysis of cross-talk in a mammalian cellular signalling network. Nat. Cell Biol. 2006;8:571–580. [PubMed]
49. Grueneberg DA, et al. Kinase requirements in human cells: I. Comparing kinase requirements across various cell types. Proc. Natl. Acad. Sci. USA. 2008;105:16472–16477. [PubMed]
50. Apsel B, et al. Targeted polypharmacology: discovery of dual inhibitors of tyrosine and phosphoinositide kinases. Nat. Chem. Biol. 2008;4:691–699. [PMC free article] [PubMed]
51. Szulc J, Wiznerowicz M, Sauvain MO, Trono D, Aebischer P. A versatile tool for conditional gene expression and knockdown. Nat. Methods. 2006;3:109–116. [PubMed]
52. Shin KJ, et al. A single lentiviral vector platform for microRNA-based conditional RNA interference and coordinated transgene expression. Proc. Natl. Acad. Sci. USA. 2006;103:13759–13764. [PubMed]
53. Stern P, et al. A system for Cre-regulated RNA interference in vivo. Proc. Natl. Acad. Sci. USA. 2008;105:13895–13900. [PubMed]
54. Meier-Schellersheim M, et al. Key role of local regulation in chemosensing revealed by a new molecular interaction-based modeling method. PLOS Comput. Biol. 2006;2:e82. [PubMed]
55. Karlebach G, Shamir R. Modelling and analysis of gene regulatory networks. Nat. Rev. Mol. Cell Biol. 2008;9:770–780. [PubMed]
56. Zak DE, Aderem A. Systems biology of innate immunity. Immunol. Rev. 2009;227:264–282. [PMC free article] [PubMed]
57. O'Shea JJ, Hunter CA, Germain RN. T cell heterogeneity: firmly fixed, predominantly plastic or merely malleable? Nat. Immunol. 2008;9:450–453. [PubMed]