Systems and components in biology

Abstract

A recent editorial [1Editorial In pursuit of systems.Nature. 2005; 435: 1Google Scholar] in Nature suggested that ‘systems biology’ is the key to answering that old chestnut and cause of consternation to pet owners the World over: the difference between a live cat and a dead one. The implication, perhaps unintended, might be encapsulated thus (usual apologies to Pope):Nature and Nature’s lawslay hid in night;God said, Let Systems Biology be!and all was light To what extent might this be justified? My impression is that ‘traditional’ biology has provided us with a fair understanding of what makes a ‘live cat’, though clearly there is still a lot we do not understand. Faced with the raw genome sequences of cat and dog, for example, I don’t think we could yet get very far in predicting which would encode the willing subject of Pavlov’s famous experiment, and which the usual adornment of the witch’s broom. But has systems biology yet added significantly to the rich canon of knowledge derived from the last two hundred years’ biology research? To my mind, the jury is very much out. What precisely is ‘systems biology’? Not to be confused with the well-established field of systems neurobiology — which simply refers to studies of multi-neuron systems, in contrast to studies that focus on properties of single neurons — systems biology is a child of genomics. A characteristic feature is the ‘top-down’ approach — faced with large datasets generated by genomic analyses, systems biologists aim to gain insights by considering the ‘system’ as a whole, often represented in a rather abstract way as a system of interacting components, shorn so far as possible from unattractive details. As Nature’s leader writer says [1Editorial In pursuit of systems.Nature. 2005; 435: 1Google Scholar]: “Properties such as robustness and evolvability, essential characteristics of life, then emerge from the topology of biological networks, independent of the constituents from which they are built.” There are examples of general, constituent- independent laws in biology, of course, but not many [2North G. Biophysics and the place of theory in biology.Curr. Biol. 2003; 13: R719-R720Abstract Full Text Full Text PDF PubMed Scopus (5) Google Scholar]. To my mind, one of the key lessons of biology’s history is that, for those specific systems we understand best — such as the phage lambda and its genetic switch [3Ptashne M. 2004Google Scholar] or haemoglobin and its cooperative oxygen binding — that understanding has come from a very close attention to the details of the system’s components. Much of biology can be considered in terms of components and their interactions within some system: from the atoms that make up a protein and its molecular environment to the species that make up an ecosystem. The question is: what features of the components are required to have an ‘understanding’ of the ‘system’? An assumption has to be that one can simplify — but how much? This issue came up in the 1980s in particular in computational neurobiology: a group of neural network modellers made great play of the way you could simulate certain aspects of ‘learning’ in abstract neural networks based on very simplified components — artificial neurons arranged in some pattern with modifiable connections mimicking plastic synapses of real neural circuits. The neurons gather inputs and give outputs as a function of the total input; the connection strengths are varied according to some rule — a particularly famous version is ‘back propagation’, where the output of the network is compared against some desired output to give an error signal that is propagated back through the network, updating the connections so to reduce the output error. This caused quite a lot of excitement at the time [4North G. A celebration of connectionism.Nature. 1987; 328: 107Crossref PubMed Scopus (7) Google Scholar], but looking back one wonders how much lasting contribution it made. No doubt such approaches are informative — certainly I agree that there are multiple useful levels of understanding how the brain works, not all couched in terms of the cellular nuts and bolts. It is useful to know just what a connection of simple neurons with a learning rule — and the oft-used Hebbian rule has a basis in experimental biology — can do. But obviously it is not sufficient to say we ‘understand’ a real neural circuit, let alone ‘how the brain works’. As Crick [5Crick F. The recent excitement about neural networks.Nature. 1989; 337: 129-132Crossref PubMed Scopus (319) Google Scholar] argued eloquently, for real understanding, you have to consider real neurons and their real properties. This will be a key issue for systems biologists attempting to make their mark — getting it right will make the difference between advancing our understanding of biology as it is, and simply establishing properties of abstract constructs.