These three different approaches to obesity each tell a partial and poorly connected story of obesogenic mechanisms. The first approach looks for genetic causes, employing genome-wide association studies, whole-genome linkage scans and molecular techniques to identify the molecular pathways involved. The second approach uses epidemiological studies, animal experiments and molecular approaches to address gene-environment interactions that manifest themselves epigenetically early in life. The third line of inquiry investigates variations in the composition of gut organisms and their relationship to obesity via epidemiological association studies, metagenomic techniques and biochemical analyses. All these approaches may have hoped initially for linear, one-dimensional and unicausal explanations but these were obviously ill-founded hopes. Assuming simple linear pathways as the causal explanation of disease has produced neither promising preventive strategies nor effective drug treatments for obesity, as the leptin story shows. While there may be diseases and biological processes for which multi-level systems explanations are unnecessary, we focus on obesity as the exemplar of a disease for which a narrow causal story and limited range of interventions will not suffice. Obesity, along with many other common conditions, has to be addressed at multiple levels, including the social [133, 134], and ultimately all three of the approaches outlined in this paper have had to widen their focus and begin to construct much more multifactorial cause-and-effect scenarios. Philosophers of science and medicine have long argued for the necessity of multilevel explanation (e.g., [135–137], and the era of systems biology and systems medicine is seeing the gradual implementation of complex causal accounts of disease.
Even though each angle of inquiry into obesogenesis has generated a range of more sophisticated understandings, the connections between these three major perspectives have barely been investigated. We have pointed to a few links made between, for example, the genetics and epigenetics of obesity, or developmental and microbiological interactions, but this is just the beginning of developing a genuinely organism-wide and dynamic life-history understanding of obesity. Such connections need to be made not simply for theoretical satisfaction, but for the practical benefits that arise out of greater capacities for the prediction, prevention and management of obesity. Such aims are part of the translational agenda of much biomedical research today, and we suggest that this agenda rests squarely on the shoulders of research capacities for intervention and integration. We suggest that what will bring these three programmes of obesity research together is a systems-oriented translational undertaking that frames the future of obesity research as it might be approached by systems medicine.
Intervention is the ground on which explanations of biological processes are built. Interventionist science involves the search for causal-mechanical explanations of phenomena . Causality, from an interventionist perspective, cannot be inferred merely from observation of a biological process but requires experimental manipulations that take control of one or more variables . Conceiving of intervention as the core of scientific practice provides a useful way in which to comprehend the life sciences in general, and obesity research in particular. First, an intervention that is successful in one research setting has to be transferred successfully to others in order to be maintained as a robust causal inference. And second, understanding translation as the successful transfer of that intervention aids the development of a broader understanding of translation, in which knowledge is understood as a practical accomplishment. We are not denying that sophisticated understandings of biological processes can be built on initially restricted combinations of techniques, questions and bodies of data. Very specific questions may continue to be fruitfully addressed by particular techniques and specific mechanistic accounts. However, the recent history of molecular biology, and obesity research in particular, shows that even as limited approaches succeed, they generate a requirement for the integration of more data and approaches in order to answer associated or wider-ranging questions .
This ramification of inquiry is what each of our three accounts of obesity research demonstrates. For example, an experimental manipulation of the leptin-producing pathway, so that it suppresses the glucose production stimulant, glucogen, and leads to the normalization of glycemia in insulin-deficient rodents, suggests a potent therapeutic intervention point [141, 142]. But this experimental intervention would need to be reproduced in a variety of experimental systems (e.g., different animal models, different conditions of calorie intake) and integrated into a wider body of intervention-oriented knowledge (e.g., insulin regulation). If such an intervention proves robust, despite its artificial isolation from a wider system of processes, it may be transferrable to a context of therapy development. However, as our outline of the history of leptin research made clear, focusing on a single point of intervention is not likely to bring about an effective and general therapeutic strategy for a broad condition. As bodies of molecular physiological knowledge have been generated, and multilevel insights into complex processes such as obesity have expanded, integration has become the catchcry of a forward-looking research agenda. This agenda is aimed at the greater transferability of research findings from one research domain to another, and the subsequent expansion of prediction and control this may enable.
Integration has emerged as a major desideratum in an era of biological practice that is characterized by the high-throughput production of vast bodies of data and extensive powers of computational analysis. An increasingly prominent response to this embarrassment of riches is systems biology, which mandates integration at all levels of practice: methods, materials, data, causal inferences and disciplinary approaches [143–145]. This new approach is characterised by the effective combination of experimental 'wet' biology with computer-based analyses and mathematical models . Through the repeated integration of biological knowledge, predictions made from mathematical models can be tested experimentally, the models modified accordingly, more data integrated, and at each step, systems-level insight improved .
Integration cannot stop in the laboratory, however, and one of the justifications of systems biology is its mooted ability to connect more closely and immediately with 'translated' benefits. The 'systems medicine' manifesto is a prime example of such an integration, where efforts are made throughout the whole research process to integrate systems-biological insights with clinically applicable results [147–149]. From this perspective, conditions such as obesity have to be understood as the consequences of complex interactions between networks of molecular activity and environmental factors. These understandings are deemed to require iterative research strategies taking multiple levels of data into account [150–152]. The main goal for these systems-oriented practitioners is not exclusively understanding or explanation but the prediction and control of systems so that basic and applied scientific practices are not separated. 'Knowledge' in such a research context means being able to intervene in a particular process and affect its outcome in a variety of contexts. Even where a systems approach does not immediately lead to increased control, it produces a better understanding of the system.
It might be thought that intervention, with its focus on linear cause-effect relationships, is at odds with system-level integration. For a therapeutic intervention to be effective from a systems point of view, it should be based on multilevel insights that pervade the whole organism, its life-course and environment [153, 139]. Yet, any experimental or therapeutic intervention, no matter how broadly informed and carefully conducted, is going to operate as a simplified modification of self-maintaining multidimensional biological processes. But just as experimental interventions focus on the manipulation of particular variables in order to understand causal relationships, systems-biological manipulations focus on nodal points of intervention from which wider system effects can be predicted [154, 151]. Manipulation of these nodes is achieved on the basis of knowledge about the interconnections between and within systems. Identifying the multilevel effects of nodal interventions allows better prediction and control not just of that node but those connected to it. Nodal interventions must also take steps to predict and avoid side-effects and compensations, which are serious problems for single-pathway pharmaceuticals that attempt coarse interventions in dynamic systems . The aim, therefore, of systems biology and systems medicine is to increase the effectiveness of experimental and therapeutic interventions, rather than simply extend them to multiple levels and aspects of the system. By framing complex processes pragmatically, translation brings together integration and intervention and potentially resolves some of the apparent contradictions between them.
Translation is often discussed as a key issue for contemporary biosciences. It is usually conceptualized as the transformation of knowledge into useful products [156–158]. In medicine and clinical practice, where translation is a frequently discussed process, it is described as the strategies that move research 'from bench to bedside' [159, 160]. Although there is considerable dispute about how to conceptualize and implement translation, we suggest it cannot be restricted to the linear transition of research from lab to clinic or industry, because of the impoverished account of scientific practice on which this definition relies. Rather than discussing translation in this restricted sense, we urge a broader understanding of translation as the transferrable achievement of intervention. Going back to obesity research for exemplification of this point, it is scientifically desirable that a laboratory-based biochemical intervention in, for example, leptin and insulin interactions in mice, is potentially transferrable to other experimental systems and other disciplinary approaches. And ultimately, genetic and biochemical understandings of leptin and insulin interactions have to be integrated into epigenetic and metaorganismal knowledge, or the successful transfer of intervention will fail. The rationale here is not purely theoretical or philosophical (i.e., whole systems are the best objects of inquiry) but pragmatic: concerned with the transfer of a particular intervention into another context.
Systems biology is not, in this scenario, a replacement paradigm for the reductive investigative strategies of the molecular sciences but their necessary extension in certain inquiries. As we have noted, some conditions and their associated explanations and interventions may not require full systems-biological and systems-medicine approaches. Systems biology and medicine seek to integrate not only approaches and data, but also discovery, analysis and therapy in a variety of contexts. We suggest that this confluence is where the true meaning of translation lies: in the capacity to transfer interventions from context to context during the pluralistic investigation of a system. These contexts are simultaneously rather than sequentially producers of experimental as well as application-oriented interventions. A therapeutic intervention, such as leptin replacement, will not achieve much in organisms that are not leptin-deficient but leptin-resistant. But this very failure of therapy can be conceived as a process manipulation and transferred into further experimental contexts to enrich understanding of success and failure of that intervention. Translation also encompasses the integration of what might be thought of as more philosophical perspectives on scientific practice. The obesity research narratives we have told show how initially fruitful reductionist strategies are being translated into the investigation of nonlinear interactions between components in multidimensional systems. Such system-oriented insights are restricted at first, but can eventually be transformed into more dynamic, interactive representations of the system under study, as each of our three bodies of research shows. In this framing of contemporary biological practice, obesity research is a prototypical example of the requirement to understand the dynamic of science in translational terms. By 'requirement' we mean not simply something mandated from above (e.g., by funders, health policy, and profit incentives) but as a descriptive and normative account that encompasses scientific practice in today's biology.
Thus, from the perspective we advocate -- which is highly compatible with several philosophical positions on pluralism, pragmatism and intervention -- translational activities occur at every level of scientific practice, such as when predictions are translated from one disciplinary or explanatory domain to another through transferable technologies and generalizable model-based insights. These cumulative processes of translation are always concerned with intervention, whether in experimental systems, clinical trials or product development. Thinking in this way about translation clearly challenges the distinction between basic and applied research -- already argued by many commentators to be a false categorization [161–163]. A broad conception of translation also includes fundamental theoretical changes and the exploration of new avenues of research occurring in the same contexts in which new interventions are being developed and applications anticipated. Such interventions thereby become part of the iterative cycle of translation, further intervention and reintegration.
Very clearly, therefore, obesity research is a field that can contribute to deep insight in the philosophy of science and medicine. As our outline of the three streams of research shows, philosophical accounts of causality, intervention, pluralism and explanation are given new and pressing reasons to be used as tools of analysis in regard to biomedical research programmes. Philosophers will gain valuable understanding of systems perspectives on health and illness, as well as the process of translating scientific knowledge and tools from one context to another within and between areas of research and application. These philosophical insights have the potential to contribute different levels of insight to emerging areas of bioscientific and biomedical research even as philosophy develops novel concepts and frameworks to accommodate this new knowledge. By thinking of the relationship between science and philosophy of science as interactive and mutually informing (e.g., ), greater joint capacity can be developed to address research impact, critical limitations, and new modes of scientific practice. It should not be necessary to argue (although sometimes the need is felt e.g., ) that an informed philosophical analysis of the science is also required for better ethical analysis. We suggest, therefore, that understanding the epistemic and ethical complexities of obesity research will illuminate the many practical scientific issues arising from the development of therapeutic interventions and the policies they are likely to inspire. This illumination will have consequences for other fields of medical and scientific research, thus resulting in further mutually beneficial interplay between science, medicine and their philosophical studies. Obesity, in all its dimensions, is able to expand philosophical discussion while the philosophy of science adds new angles of discussion to the study of obesity.