Nature Biotechnology
22, 1215 - 1217 (2004)
doi:10.1038/nbt1004-1215
The impact of systems approaches on biological problems in drug discoveryLeroy Hood1
& Roger M Perlmutter21 Leroy Hood is at the Institute for Systems Biology, 1441 North 34th Street, Seattle, Washington 98103-8904, USA lhood@systemsbiology.org 2 Roger M. Perlmutter is at Amgen, Inc., One Amgen Center Drive, Thousand Oaks, CA 91320-1799, USA. The production of new molecular entities (NMEs) endowed with salutary medicinal properties provides a key metric by which the effectiveness of the pharmaceutical industry may be judged. Nearly a decade ago, Jürgen Drews1 noted that the combined output of NMEs by all major pharmaceutical companies at the time was insufficient to support the healthy growth of the group. Indeed, a recent analysis documents a 30-year decline in pharmaceutical R&D productivity2. This decline in productivity has occurred despite accelerating investment in biomedical research on the part of both industry and governments, with compound annual growth rates in research expenditures approaching 13% (ref. 2). Approvals aside, the number of late-stage clinical trials supported by the pharmaceutical industry has also declined during the past decade3, auguring poorly for future triumphs in disease interdiction. Despite several notable discoveries, the emergence of major biotech companies has done little to improve the yield of new disease-ameliorating molecules.
How is it that extraordinary advances in biomedical research over the past decade have exerted so little positive effect on drug discovery? We argue here that an inability to visualize the complexity of biological systems has impeded the identification of novel therapies. In the future, application of systems approaches to drug discovery promises to have a profound impact on medical practice, permitting a comprehensive evaluation of underlying predisposition to disease, disease diagnosis and disease progression. In the near term, systems biology will provide powerful means for validating new drug targets, improving the success with which pharmaceuticals are identified. Farther into the future, the same approaches will drive the development of early diagnostics, enabling disease stratification, individualized therapy and ultimately preventive drugs, based on both genetic and environmental considerations. Although systems biology as currently envisioned does not have a direct impact on the chemistry of identifying drugs or pharmacological challenges of drug metabolism, it may provide rapid and useful assays for these in the future.
What is systems biology?
Systems biology represents an analytical approach to the relationships among elements of a system, with the goal of understanding its emergent properties4. Broadly speaking, a system may include just a few protein molecules that together serve a defined task (e.g., fatty acid synthesis), a more complex molecular machine (e.g., a transcription complex) or a cell or group of cells executing a particular function, such as an immune response. Thus, systems analysis can be applied to molecules, cells, organs, individuals or even ecosystems. In each case, one seeks to describe all of the elements of the system, define the biological networks that interrelate the elements of a system and characterize the flow of information that links these elements, and their networks, to an emergent biological process. The key idea is that disease reflects the operation of perturbed networks. Through a comparison of normal and diseased networks, critical nodal points (proteins) can be identified which, if modulated, are likely to reconfigure the perturbed network structure back toward its normal state or specifically kill the diseased cell. These nodal proteins represent likely drug targets.
As applied to biomedical research, systems biology provides a mechanism for resolving some of the most formidable challenges that confound successful drug discovery: rational target identification, recognition and avoidance of adverse properties of therapeutics before commitment to clinical trials, the monitoring of clinical efficacy using surrogate markers and individualized approaches to disease treatment (see Fig. 1).
Target identification
Completion of the human genome sequence provided access, in principle, to all protein primary sequences, and thus to all potential drug targets (and the drugs themselves, in cases where secreted proteins or soluble receptors can be used to modify disease manifestations). However, preclinical models have relatively little predictive power for human disease. By this, we mean that the complex set of underlying genetic traits and environmental factors that yield, for example, human asthma, have no strict parallel among various animal models used to define the response to therapy. Furthermore, studying one or a few proteins at a time, the strategy employed during the past 30 years, has generally failed to define leverage points that affect disease progression. Thus, the development of new therapeutics suffers from the lack of identification of relevant targets, among the products of 30,000 known genes, that unambiguously contribute to disease pathogenesis.
Large data sets can be acquired for genomic (e.g., transcription profiling) and proteomic (e.g., qualitative and quantitative delineation of expressed proteins) analyses. These various types of data can be integrated into a network model of how a particular biological system operates5. Through genetic and environmental perturbations of the system elements and comprehensive analyses of gene products, one can clarify the structure of the system network and delineate key nodal points (proteins). From this process, inferences can be drawn about the importance of individual proteins in the disease process. Analyses of this type have been explored in a variety of inflammatory diseases (e.g., see ref. 6). The availability of very large data sets that after integration define nodal points in the disease process should greatly assist in winnowing the mass of putative targets to a more manageable number (the enormous challenge of large biological data sets is deciphering signal from very significant levels of noise).
Recognition and avoidance of adverse properties
With a target in hand, medicinal chemists typically succeed in generating potent, low molecular weight compounds that both bind the target and alter its biological behavior. Drugs typically fail, however, because they lack appropriate pharmaceutical properties—they manifest deficiencies in absorption or pharmacokinetics or yield metabolites that have unfavorable side effects. Under these circumstances, the derivation of new compounds with improved pharmaceutical properties proceeds chaotically because the interaction of externally derived molecules with the human ecosystem, including human genome−encoded metabolizing enzymes and other molecular machines as well as the products of bacterial commensals, cannot be easily predicted.
Characterization of the 'metabonome,' the thousands of products of metabolism derived from both internal and external sources, has already altered thinking about lead compound elaboration in drug discovery7. New spectroscopic tools permit the simultaneous enumeration of thousands of metabolic products in biological fluids. By assembling metabolic profiles using these spectra, it has already become possible, in simple cases, to associate off-target toxicities with patterns of accumulation of certain metabolites. Similarly, changes in the proteome profile can provide insights into the inadvertent effects of drugs, whether because of an unappreciated interaction between the drug target and another biologically active molecule, such as a protein, or as a result of illegitimate binding of the drug to previously undefined targets. We believe that analyses of drug-induced perturbation of normal and diseased networks will yield striking insights into the adverse effects of potential pharmaceuticals. With this in mind, the combination of protein and metabolite profiling should provide unprecedented power for the identification of safer therapeutic entities.
Development of surrogate markers
Few areas of drug development are more challenging than that of monitoring the clinical efficacy of new molecules. Successful registration of a new drug requires demonstration of a satisfactory therapeutic index—the ratio of drug dosages yielding adverse effects to those required to produce clinically meaningful results. In a few diseases, such as reduction of serum glucose levels in diabetes, increases in hematocrit in anemia or the reduction in serum cholesterol levels in hyperlipidemia, satisfactory molecular surrogates exist to predict a beneficial clinical response. In most cases, however, the clinical response itself must be measured, typically requiring years of expensive, large-scale clinical evaluation. The development of improved surrogates for therapeutic efficacy would permit rapid assessment, in small numbers of subjects, of the extent to which a putative therapeutic might yield a satisfactory clinical result.
To define improved clinical surrogates, it is necessary to examine biological samples from diseased individuals over time and to evaluate proteome expression and network behavior both as a function of disease activity and in response to conventional therapy. A key point is that identification of satisfactory surrogates will require multiparameter analyses of drug responses, which reflect the changing behaviors of the relevant protein networks. In most cases, informative surrogates will be complex, reflecting the pattern of expression of hundreds, or perhaps thousands, of proteins. Validation of these multiparameter surrogates will, in the first instance, require confirmation using real clinical measures. As understanding of these protein profiles increases, confidence in retrospective clinical concordance, rather than prospective testing, will increase. This will be aided by the definition of subsets of patients, defined by genetic polymorphisms and proteome expression patterns, with comparatively homogeneous disease characteristics.
Individualized therapy
Human subjects, selected at random, are separated from one another by six million polymorphisms, representing largely single nucleotide substitutions of no special interest, but including heritable differences in disease susceptibility and disease characteristics. Common disease manifestations inevitably reflect variations in this heritable substrate, upon which environmental influences are imposed.
Stratification by genetic characteristics already permits the identification of tumor susceptibility and the prediction of response to some therapeutic interventions8,
9. Realization of personalized, predictive and preventive medicine will require new analytical approaches to extraordinarily large, and extraordinarily noisy, data sets, including medical records4. In the near term, early clinical trials will benefit from selection of more homogeneous patient populations. In the long term, idiosyncratic responses will provide the basis for developing highly personalized therapies, with substantially improved safety and efficacy profiles.
Challenges in implementing systems approaches
The establishment of an effective systems biology group requires (at a minimum) cross-disciplinary scientists, high-throughput facilities for genomics, proteomics and metabolic profiling and a strong computational infrastructure. Systems biology is an immature discipline and, accordingly, it is necessary to have scientists and engineers on the cutting edge of technology application and development and mathematical modeling. To practice systems biology at the leading edge, these technologically adept scientists must learn to speak the language of biology (and medicine), have a deep understanding of at least some aspects of biology and work effectively in teams.
The integration of disciplines, from mathematics to molecular biology, is one of the most challenging aspects of the systems approach. At this stage in the evolution of systems biology, it is essential to attract the most outstanding scientists and engineers, for they are required to deal with immature sciences. The enormous challenge of systems biology for both academia and industry is to integrate biology, technology and computation. The creation of high-throughput, computationally advanced facilities should not be difficult for industry groups. But partnerships with academic organizations will be a key to integrate leading-edge approaches and tools. Such partnerships raise concerns about intellectual property, publication rights and the advancement of research technologies and computational tools from prototypes to robust implementation— challenges that must be solved if the pharmaceutical industry is to effectively benefit from systems biology.
The future
The pharmaceutical industry will lose nearly $80 billion in revenue by 2008 due to patent expiration, and the current drug pipeline will replace only a small fraction of this value10. Existing approaches to drug discovery are failing to keep up with the demands of an industry that experienced 11% average growth per year for more than 30 years11. Three major issues are associated with identifying effective new drugs: first, discovery of a relevant drug target; second, identification of a drug that will appropriately perturb the target; and third, assessment of the possible side effects and pharmaceutical properties of the drug before its deployment in clinical trials. Systems biology offers powerful new approaches for dealing with the first and third of these problems.
It remains to be seen how industry will exploit this novel opportunity. Will in-house efforts bolstered by academic partnerships permit implementation of systems approaches in the traditional pharmaceutical houses, or will powerful and effective new systems-biology companies emerge to capture the front end of the drug discovery market? The latter approach would represent a form of horizontal stratification, where the elaboration of new molecules could occur under the protection of different corporations from those involved in clinical development and commercialization, and would resemble, in some ways, the horizontal stratification that has occurred in more mature information-technology industries.
In our view, systems biology will inevitably change the rules that govern the selection and development of new therapeutics and will catalyze the development of personalized, predictive and preventive medicine in the next decade. The fascinating question is: who will lead this extraordinary change process?
REFERENCES
- Drews, J. Nat. Biotechnol. 14, 1516−1518 (1996). | Article | PubMed | ISI | ChemPort |
- Booth, B. & Zemmel, R. Nat. Rev. Drug Disc. 3, 451−456 (2004). | Article | ISI | ChemPort |
- Anonymous. Parexel's Pharmaceutical R&D Statistical Sourcebook 2003/2004, p. 42 (Parexel International Corporation, Waltham, MA, 2003).
- Weston, A.D. & Hood, L. J. Proteome Res. 3, 179−196 (2004). | Article | PubMed | ISI | ChemPort |
- Weston, A.D., Baliga, N.S., Bonneau, R. & Hood, L. in The Genome of Homo Sapiens, Volume LXVIII, Symposia on Quantitative Biology, 345−357 (Cold Spring Harbor Press, Cold Spring Harbor, NY, 2004).
- Kunkel, E.J. et al. FASEB J. 18, 1279−1281 (2004). | PubMed | ChemPort |
- Nicholson, J.K. & Wilson, I.D. Nat. Rev. Drug. Discov. 2, 668−676 (2003). | Article | PubMed | ISI | ChemPort |
- Paez, J.G. et al. Science 304, 1497−1500 (2004). | Article | PubMed | ISI | ChemPort |
- Lynch, T.J. et al. N. Eng. J. Med. 350, 2129−2139 (2004). | Article | ISI | ChemPort |
- Drug Week, Pharmaceutical Development, 25 June 2004. http://www.NewsRx.com/
- Pharmaceutical Industry Profile 2004 (Pharmaceutical Research and Manufacturers of America, Washington, DC, 2004). http://www.phrma.org/publications/publications//2004-03-31.937.pdf
|
