Recent advances in ‘-omics’ technology have allowed biologists to take snapshots of cellular components in a variety of situations, which should help us to understand how biological systems operate. It has been claimed that the interpretation of these snapshots is best provided by a shift to ‘systems’ thinking and the discipline of ‘systems biology’. Lauded as ‘the 21st century science’ (http://www.systemsbiology.org/), ‘a revolution’ (Aderem, 2005) and ‘a paradigm shift in modern life science research’ (Aggarwal & Lee, 2003), systems biology ‘promises to revolutionize our understanding of complex biological regulatory systems’ (Kitano, 2002).
But revolutions in science are rare and revolutions in biology rarer still. In fact, when illustrating his seminal hypothesis that science advances in a series of such revolutions, Thomas Kuhn declined to provide a single biological example (Kuhn, 1962). Instead, he argued that biologists, like most researchers, are usually engaged in ‘normal science’. According to Kuhn, we practitioners of normal science adhere to a reasonably coherent framework of laws, attitudes and assumptions – a ‘paradigm’– into which we fit as many phenomena as possible (Kuhn, 1962). We do not seriously question whether our prevailing paradigm is correct, although we do accept that it may be incomplete. This adherence to a paradigm means that novelty in normal science manifests itself either as the addition of missing laws, or in ‘puzzle-solving’, which is to say discovering exactly which of the known laws explain certain observations.
Kuhn went on to suggest that the picture which normal science gradually pieces together may be changed, or shattered altogether, when ‘anomalies’ are discovered. Anomalies are observations that can only be explained by assuming that at least some of the tenets of the existing paradigm are not simply incomplete, but actually incorrect. This realization triggers a ‘paradigm shift’ in which scientists replace their old set of laws and assumptions with a ‘tradition-shattering’ new one. A ‘revolution’ is consequently said to have occurred in which novelty comes not only from the acquisition of new beliefs, but also from the loss of some of the older, fundamentally incompatible, ones (Kuhn, 1962).
Kuhn trained as a physicist and supported his arguments using examples such as the Ptolemaic-to-Copernican breakthrough in astronomy (Kuhn, 1962). This bias has been remarked upon by later commentators, who either have suggested biological paradigm shift candidates such as Darwinian evolution (Cohen, 1985), or have questioned whether Kuhn's model should be applied to biology at all (O'Malley & Boucher, 2005). Such considerations notwithstanding, Kuhn still provides much of the jargon with which dramatic biological discovery is promoted and should be borne in mind by those eager to know when a revolution is under way.
Is systems biology such a revolution? There can certainly be no doubt that the systems biology approach is catching on – over the past 5 years the number of articles mentioning the term has risen exponentially (Fig. 1) – and, increasingly, employers target candidates with systems biology expertise. What is open to debate, however, is the extent to which this systems approach may properly be called a paradigm shift. In the rest of this article, I suggest that those attempts that have been made to formalize systems biology reveal an approach that falls squarely within Kuhn's definition of normal science. Indeed I argue that systems biology, which remains ‘a concept waiting for a definition’ (Fox Keller, 2005a), retreads the conceptual ground covered by post-Darwinian physiology and resembles nothing more than the application of 19th century physiological reasoning to today's molecular biological data.
The reductionist paradigm and the rise of the system
It may be helpful to remind ourselves of the ‘normal science’ paradigm which systems biology claims to replace. Normal biology, broadly speaking, is held to consist of ‘simple reductionist approaches’ (Aderem, 2005) with the main claim of reductionism being that one science reduces to another; for instance that biology reduces to chemistry, that chemistry reduces to physics, that physics reduces to maths, and so on. Central to this belief is the idea that any phenomenon in the more ‘complex’ science may be generated by algorithms applied to entities of the more ‘simple’ science (Dennett, 1995).
However, when considering the problem of how to interpret the datasets of modern ‘-omic’ experiments, systems biologists argue that there ‘are two main problems with the reductionist philosophy that have impeded progress’ (Sweetlove & Fernie, 2005). The first charge is that modern mathematical and computational techniques struggle to cope with the vast amount of information which must be processed in order to model even the simplest of biological systems using algorithms applied at the molecular level. The second charge is more revealing, and holds that reductionism cannot explain biological function because ‘function results from the complex interactions between large numbers of molecules’ (Sweetlove & Fernie, 2005).
To illustrate this, let us use the analogy of trying to understand a flying airplane (after Kitano, 2002). It is clear that giving a list of airplane components is not sufficient to comprehend flight – a random collection of nuts, bolts and sheet metal will not fly. In order to explain function, therefore, systems biology argues that we should take a more holistic view of biological phenomena, one in which function emerges at a higher-than-molecular level. This is the level of ‘systems’, whose behaviour is described by topological, not physico-chemical, laws.
Having added systems to our biological ontology, systems biology goes on to give a ‘framework for defining the hierarchy of essential questions’ (Somerville et al., 2004) which need answering if we are to understand emergent phenomena. In other words, it defines an epistemology with which to investigate and describe systems. The best description of this epistemology may be found in the most-cited systems biology article, and states that a ‘system-level understanding’ (Kitano, 2002) requires the elucidation of four key elements. Using our airplane example once again, we need to know how components are fitted together into structures (the plane flies because nuts and bolts are assembled into wings and an engine), we need to know how these structures dynamically interact (the plane flies because the wings provide lift at the same time as the engine provides thrust), we need to know how this dynamism is monitored and controlled (the plane flies because the pilots are in the cockpit correcting observed deviations from an ideal flight path), and finally we need to know the desired outcome which this control can give us (the plane flies because we want to get from London to Plymouth cheaply and safely). These four epistemological elements have been named, respectively, the ‘system structures’, ‘system dynamics’, ‘control method’, and ‘design method’ (Kitano, 2002).
The final piece of the systems biology jigsaw adds methodology to ontology and epistemology. Data acquisition methods span the ‘-omics’ spectrum and data analysis concentrates on describing information flow and regulatory feedback using mathematical techniques which focus on the mid-20th century ‘cybernetics’ of Norbert Weiner (Kitano, 2002; Westerhoff & Palsson, 2004) and the exciting new field of network analysis (Barabási & Oltvai, 2004).
Two millennia of explaining function
To recapitulate, proponents of a systems approach to biological phenomena set themselves in opposition to ‘simple reductionist approaches’ and argue that biological function is generated at a higher-than-molecular level. Accordingly, biological systems require ‘two distinct levels of completion’–‘parts list completion’ and ‘systems biology’ (Selinger et al., 2003). The latter provides an epistemology with which to understand the emergence and regulation of function by drawing conceptual and mathematical analogies between biological and designed, end-directed, systems. Is this enough for systems biology to constitute a paradigm shift?
I argue that it is not, for two reasons. First, we have seen that systems biology explicitly sets itself against the paradigm of ‘reductionist philosophy’ (Sweetlove & Fernie, 2005). This is a disingenuous stance because ‘the reductionist philosophy’ argues that biology may be reduced – it does not specify which level it should be reduced to. While ‘Reductionism is a dirty word, and a kind of “holistier than thou” self-righteousness has become fashionable’ (Dawkins, 1982), arguments that seek to reduce biology to engineering are, fundamentally, as reductionist as those which seek to reduce biology to chemistry. Systems biology may add systems to the current molecular ontology of ‘parts list’ biology, but it does not seek to falsify the laws of molecular behaviour. Instead it suggests that systems made up of many components can display emergent properties that must be considered in addition to, and not instead of, the properties of the components themselves.
Second, systems biologists are not the only ones who have noticed the importance of connectivity. To say that function is an emergent property is to say that the connections between the components of a system help to determine whether the systems can perform a particular task. This idea, and the corresponding end-directed epistemological toolkit of systems biology, has been around for over 2300 years (Box 1), although a convincing explanation for exactly how specific connections become established had to wait until Darwin's theory of natural selection (Darwin, 1859). Even so, the idea that the components of biological systems have been assembled in such a way, albeit by chance, that they may be explored using the teleological language of end-directed design has been one of the key assumptions of the modern, reductionist, evolutionary synthesis for well over a century (Dennett, 1995).
In fact, the implications of Darwinian ‘teleology’ (Gray, 1874; Lennox, 1993) were grasped more immediately than we might remember today and plant science was a particular beneficiary. Darwin's keen interest in plant function (Darwin, 1880) and constant correspondence with professional botanists helped inspire the first ‘systems’ synthesis of functional ontology, teleological epistemology, and novel experimental and analytical methodology that characterized European plant sciences at the end of the 19th century (Box 2). The plant physiologists of the 1880s may have studied how cells co-operate to give plant functions, rather than how molecules co-operate to give cell functions, but their comprehension of the questions that needed to be asked of end-directed systems was every bit as good as ours, and deserves to stand as an example to today's generation of botanists.
Systems biology is neither a revolution nor a paradigm shift
I hope it is now clear why systems biology should neither be called a paradigm shift nor a revolution. Systems biology seeks to investigate cellular events by using novel experimental and mathematical methodologies, but ontological and epistemological principles that have formed the basis of physiology for over a century. The idea that function is an emergent property of a system, requiring a description of parameters beyond those of the system's components, is a well-established tenet of reductionism (Dennett, 1995) and systems biology therefore extends, but does not replace, today's reductionist paradigm. As such, systems biology is simply a step in the natural progression of normal biology, albeit a potentially important one.
Systems biology nevertheless holds great methodological promise
I would caution against jumping onto the ‘bandwagon of “systems biology”’ (Fox Keller, 2005a), in part because of the proliferation of articles whose systems biology credentials seem, at best, tenuous. However, it remains true that the ‘complex networks of gene interactions, proteins, and signalling between the cell and other cells and the abiotic environment is probably incomprehensible without some mathematical structure perhaps yet to be invented’ (Cohen, 2004). Here, then, is scope for genuine methodological novelty and, although current efforts may not be quite as groundbreaking as their practitioners think (Bentley & Shennan, 2005; Fox Keller, 2005b), it is to their credit that this is a focus of today's systems biologists (Westerhoff & Palsson, 2004; O'Malley & Dupré, 2005). Nonetheless, given the emphasis they place on connectivity and an interdisciplinary approach, practitioners of systems biology may benefit from the realization that if they ‘have seen a little further it is by standing on the shoulders of giants’ (Newton, 1676) rather than by toppling them out of the way.
Box 1 Aristotle and the epistemology of designAristotle (384–322 bc) wanted to search for explanations of natural events that inspire wonder. His search led him to conclude that any question which might be asked about the behaviour of a complex, apparently designed, system might be answered if we knew four properties of that system. He called these the aitiai, a word which is usually rendered into English as ‘causes’, but which may be better translated as ‘explanations’ (Aristotle, APst 90a7–94b34; GA 715a1–17).The first, ‘material’, cause corresponds to ‘parts list completion’ and tells us what components the system is made from. Some questions may be answered if only this cause is known (Q. ‘Why does this table burn?’ A. ‘This table is made from wood, wood burns, so this table burns’). However, such a description can't answer many other questions, so Aristotle added three more causes which, between them, allow a ‘systems’ explanation of the apparent design seen in complex systems.The second, ‘formal’ cause defines the way in which components are arranged and maintained, in a similar way to the ‘system structures’ and ‘control method’ of systems biology. Knowing the ‘formal’ cause allows us to answer questions which knowing only the ‘material’ cause would not (Q. ‘Why is this table only stable on a flat floor?’ A. ‘This table has four equal legs, four equal legs will only be stable on a flat floor, so this table is only stable on a flat floor’). The material from which the ‘form’ is made is irrelevant and if the ‘form’ were different (three legs, say, instead of four) the behaviour of the system would also be different.Taken together, the ‘material’ and ‘formal’ causes provide a static picture of a system. But Aristotle fully understood the importance of explaining change and added two dynamic causes. The ‘efficient’ cause corresponds to the ‘system dynamics’ of systems biology and covers the states through which a system passes by describing the events which have produced or changed the system (Q. ‘Why is this table smooth?’ A. ‘This table was made by a carpenter, carpenters sand rough wood until it is smooth, so this table is smooth’). The second dynamic cause and the last of the four causes is, appropriately enough, the ‘final’ cause. This is an explanation of what something is for and so preempts the ‘design method’ of systems biology (Q. ‘Why does this table have a level surface?’ A. ‘This table is to eat off, food won't slide off a level surface, so this table has a level top’).Aristotle's appreciation that nature shows design lead him to integrate his doctrine of the four causes into his stunning biological works, particularly History of Animals and Generation of Animals (Aristotle, HA, GA).Significantly, Aristotle himself left hanging the question of how design was generated (Gotthelf, 1999). However, his reputation has unfairly suffered from the ‘argument from design’ of mediaeval Scholasticism, which reasoned that Aristotle's three ‘design’ causes reflected (and indeed proved) the existence of a designer. Thus a 12th century hymn writer, Adam of St. Victor, could write (Blume & Dreves, 1915):effectiva vel formalis The effective or the formalcausa Deus, et finalis, cause God can be, or the final,sed numquam materia But never the materialAnd neatly paraphrase the ontological distinction made over 800 years later between ‘parts list completion’ and ‘systems biology’ (Selinger et al., 2003).
Box 2 19th century plant physiologyDarwin's theory of Natural Selection was popularized by a number of his friends, with Thomas Henry Huxley foremost among them. Huxley was quick to realize that Darwinian ‘teleology’ (Gray, 1874) made the concept of function an entirely appropriate one to use in post-Darwinian biology and urged its study under the name of ‘physiology’.Huxley's insistence that ‘The final object of physiology is to deduce the facts of morphology’ (Huxley, 1870) was accepted particularly quickly in Germany, with its strong tradition of biological research, and led to a flowering of European investigations into ‘systems’ properties in plants.As an example, let us focus on the work of Wilhelm Pfeffer, one of the earliest exponents of this post-Darwinian physiology and a professor at the University of Bonn.In 1877, Pfeffer was the first to measure the osmotic potential of plant cells. This was the first step on the road to quantifying plant movement (Darwin, 1880) and transpiration (Strasburger, 1891). Jacobus van 't Hoff later modelled Pfeffer's experiments using the new mathematics of statistical mechanics, work for which van 't Hoff won the first Nobel prize in chemistry. Thus a truly ‘systems’ approach justified functional epistemology in terms of Darwinian ‘teleology’ and used novel experimental and mathematical techniques to show how ‘end-directed’ bulk directional water flow could ‘emerge’ from individual random molecular motions.
I would like to thank Alistair Hetherington, Martin McAinsh and Maureen O'Malley for reading and criticizing early drafts of this letter. This does not, of course, imply that they endorse any of the arguments which I have put forward.