Owing to the complexity of the regulatory networks that underlie human disease onset and progression, finding effective treatment strategies for many complex diseases such as cancer and cardiovascular and neurodegenerative diseases remains a challenge. Furthermore, for many of these diseases, there are comorbidities the cause of which is unknown. For complex diseases, there are often many different mechanisms that result in the same phenotype, and it is therefore difficult to identify the exact cause of the disease and design adequate and efficient treatment strategies. This calls for strategies that are tailored to the individual, a concept generally referred to as personalized medicine (Fig. 1a).
A major obstacle to the introduction of personalized medicine is the lack of identification of the molecular cause of development of the disease in individual patients. Based on the identification of such a cause, it is possible to determine patient groups that have a highly effective response to the drug and at the same time have a low level of toxicity. In this way, it is possible to find drugs to specifically treat a certain group of patients, but which may not necessarily be useful to treat all patients with the same gross phenotype.
Considering the complexity of both the cellular networks that are perturbed in connection with disease development and progression, and the complex interactions between different cell types in the human body, it is generally difficult to dissect the molecular causes of different phenotypic developments. Even though large data sets are accumulating, e.g. omics data from different tissues and in large patient cohorts, it is not straightforward to use these data to identify the underlying causes of disease development. In recent years, there has, however, been progress in the use of computer models to integrate large data sets as well as for the simulation of biological networks with the objective to study their dynamic behaviour. This approach is referred to as systems biology or, when specifically applied to medical research, systems medicine.
In the health care sector, there are generally expectations that systems biology may be useful on several fronts. It can assist in: (i) identification of novel drug targets, (ii) improving insight into the mode of action of existing drugs, (iii) terminating drug development projects that may not be successful before they become too costly, (iv) early diagnosis and thereby shift from disease treatment to disease prevention, and (v) improving prescriptions so that a drug can be avoided by nonresponders and patients in whom it may have severe side effects (Fig. 1b).
At the end of June 2011, the 8th Key Symposium sponsored by the Journal of Internal Medicine and co-organized by the Royal Swedish Academy of Sciences was held in Stockholm. The topic of the 2011 Symposium was translational and systems medicine, and key developments in this emerging field were presented by research leaders over 3 days. This issue of the Journal of Internal Medicine includes six reviews from several of the presenters at the conference, summarizing various interesting aspects of systems medicine.
Hood et al.  discussed the principles of so-called P4 medicines. This concept, introduced earlier by Leroy Hood and described at the Symposium, predicts that society will move towards medicines that are not only personalized, but also preventive, predictive and participatory. Thus, through patient stratification and improved diagnostics, it will be possible to move towards better prediction of disease risks and identification of early onset of diseases, which can lead to preventive treatment in a personalized fashion. An important component of this will be the participation of individuals, both through more active involvement in prevention strategies, and also in terms of active participation in diagnosis. This is already seen in society through several diagnostic companies such as Decode, 23andme and Navigenics, which provide services directly to individuals. The tests these companies provide are based on the identification of single-nucleotide polymorphisms in different genes that have been shown to be associated with increased disease risks.
Tian et al. provide further discussion of how genome sequencing may lead to improved cancer diagnosis and treatment. Through sequencing of large cohorts and several family lines, it is believed that it will be possible to obtain better resolution and hence better prediction compared with earlier studies. The wider use of genome sequencing in human health introduces a problem with regard to computational demand. To overcome this problem, Grossman and White  foresee that cloud computing will become an important resource in modern health care sectors to handle large data sets. It is not only from genome sequencing that large data sets are being generated but any use of modern high-throughput experimental techniques such as transcriptomics, RNA sequencing, proteomics, metabolomics and lipidomics results in very large data sets, and currently, the rate of data generation is increasing faster than the increase in computer power. This calls for methods of rapid processing and compression of these high-throughput data to enable easy storage, retrieval and linkage to other information.
This approach to systems biology of generation of large data sets followed by integrative analysis to identify biomarkers or disease mechanisms is often referred to as top-down systems biology. In this approach, bioinformatics and data-handling play an important part, but systems biology can be distinguished from bioinformatics as structural information in the context of biological networks is often used for analysis of the data. Thereby, it is possible to identify sub-networks and/or pathways that are enriched in a certain group of subjects, and these may be linked to a certain disease; this in turn may lead to identification of a novel biomarker or a disease mechanism. The quality of the biological network is very important for this kind of analysis, and one type of network, the metabolic network, has been very well annotated through detailed biochemical studies. Metabolism comprises many different pathways, but through common sharing of metabolites, cellular metabolism forms a complex network with tight connection between the many different metabolic pathways. These networks can be used for integrative analysis of high-throughput experimental data and are normally referred to as genome-scale metabolic models (GEMs). GEMs are reconstructed based on detailed metabolic information, but data from genomics and metabolomics are also used in the reconstruction process. GEMs have been reconstructed for many different types of microorganisms, and, recently, also for human cells. Bordbar and Palsson  review the concept of genome-scale modelling of human cells, discussing how human GEMs have been used for different types of integrative analysis and how different types of data can be used for developing GEMs specific for various types of cells. Mardinoglu and Nielsen  discuss this further and describe how tissue-specific experimental data, e.g. data from the Swedish Human Protein Atlas project, can drive the reconstruction of tissue- or cell-specific GEMs. It is only with the availability of such cell-specific GEMs that it will be possible to analyse tissue-specific metabolic effects. Considering the importance of metabolism in many lifestyle-related diseases, such as type 2 diabetes and cancer, this may allow for not only better mechanistic insight into these diseases but also the identification of novel metabolite biomarkers.
Whereas top-down systems biology is clearly derived from the genomics revolution, in a second approach, termed bottom-up systems biology, mathematical models are reconstructed for smaller biological sub-systems. Following evaluation against the experimental data, bottom-up systems biology can be used to predict the dynamic behaviour of the biological system under study. This approach has been used in particular for gaining new insight into the function of signal transduction pathways, the target of many drugs. This approach has recently been used successfully for drug discovery. The use of kinetic modelling of cancer signalling pathways is reviewed by Bachmann et al. , who discuss erythropoietin receptor function as an excellent example of how dynamic modelling can be used to gain novel biological insight. This receptor can generate a linear output even though the erythropoietin concentration in the blood may vary up to 1000-fold. Through complex kinetic modelling of the receptor and its interaction with signalling pathways, it has been demonstrated how a wide dynamic range can be maintained.
Another very important application for kinetic modelling is the extraction of valuable information from complex data, as demonstrated by Boren et al. . They conducted several studies using tracers to follow the dynamics of different lipid particles in the plasma of patients. Analysing these data is complex as a result of biosynthesis and degradation of lipids in the liver as well as adsorption in the adipose tissue. However, through the use of a complex kinetic model to describe these different processes, the authors have demonstrated that it is possible to gain new insight into the kinetics of the different processes involved. This knowledge may be useful for improved patient stratification. Boren et al.  provide an excellent review of earlier studies and also examine how this approach can be used in the clinic.
In summary, these six reviews in the current issue of the Journal of Internal Medicine clearly demonstrate the many facets of systems medicine. It is clear that this approach will have much to offer in terms of new insights into the function of complex biological systems, including in the extraction of increased information from clinical data, in drug design and in future patient stratification. These insights will form the basis for moving towards personalized medicine.