Systems biology versus molecular biology

Charles F. Stevens
2004 Current Biology  
Systems biology and molecular biology can answer the same question in quite different ways, and frequently the answers given by one sub-discipline seem no answer at all to the other. For many molecular biologists, a systemslevel explanation leaves them feeling that, until the underlying molecular mechanisms are known, the approach is hopelessly superficial. For systems biologists, on the other hand, an account of the genes and gene interactions responsible for a phenomenon is just a list, and
more » ... ey hunger for the underlying principles that make sense out of the list. I believe that these two styles of answering questions are complementary; at best, a more complete understanding is reached when the two approaches are unified. I shall illustrate how these two approaches differ by considering the different answers they give to two questions. Why do separate neocortical areas exist? And why are the cortical areas arranged in the way they are? The example I give is flawed, however, and the nature of the flaw is revealing about the challenges facing systems and molecular biology. Since Brodmann [1], we have recognized the existence of many anatomically and molecularly distinct neocortical areas, each with a different function. About 100 areas have now been identified in the human cortex, ranging from the primary visual cortex (V1) through temporal and frontal lobe language areas. The first question, then, is: why are the neurons responsible for a particular calculation grouped together in one region of cortex? A given area, V1 for example, varies considerably in size and in its precise location from one individual to the next (and from one species to the next), but its relations to other areas never change. For example, V1 is always next to V2 and never next to V3 (V2 and V3 are other areas devoted to vision). The second question is: why are these neighbor relations the same from one individual to the next? Molecular biology answers Patterning in developing cortex [2,3], like that elsewhere in the embryo, is believed to be governed by a coordinate system established by concentration gradients of signaling molecules. Different concentrations of these signaling molecules are thought to activate different combinations of transcription factors, giving rise to the various distinct areas of the mature cortex. The molecular nature of at least some of the anterior-posterior and mediallateral signals has been tentatively identified, and a large number of genes appear to be involved in specifying cortical areas and their pattern of arrangement. To explain the existence of cortical areas and their arrangement, then, one must identify: first, the genes that instruct newborn neurons to adopt a cortical fate; second, the genes that define the coordinate system; and third, the transcription factors that specify the positions of areal boundaries, and the combinations of genes and gene expression levels that define the specific areas. Once we know the genes responsible for cortical patterning and all interactions between them, the molecular biologist has answered the questions of why we have areas and why the areas are arranged as they are. Systems biology answers Each neocortical area sends information to about ten other areas and receives information from about ten areas, generally not just the same areas to which information is sent. The axons carrying this information usually run in the white matter, and most of the white matter volume is made up of these intercortical axons. In the human brain, about 40% of the entire volume consists of white matter, so it would seem that arranging cortical areas in a way that minimizes the volume of the interconnecting axons would be advantageous; because the cranial volume is limited, minimizing the volume taken by the 'data buses' leaves more space for the brain's computational elements. The systems biologist, then, would say that the cortical areas are arranged by evolution as they are to optimize the use of space in the head. This idea has been tested for 11 areas in monkey prefrontal cortex for which all or most of the intercortical connections are known [4] . Just under 40 million arrangements of the 11 areas, all of the possible arrangements, were examined by computing the volume of interconnecting axons required for each arrangement. Every alternative was found to be worse than the actual area arrangement, in the sense that the actual arrangement required the least volume of intercortical connecting axons. This systems-level explanation, then, holds that areas have evolved an arrangement that minimizes the volume of white matter required for intercortical communications, conferring a selective advantage over evolutionary alternatives with less efficient use of cranial space, and the fact that the actual arrangement is better than any of the approximately 40 million possible alternative patterns gives strong support to the white matter minimization hypothesis. This discussion gives an answer to the second question, and suggests one to the first: grouping the most highly interconnected cells (those with the same function) together to form a cortical area saves space by minimizing the volume of axons needed to construct the computational circuit for a particular function. This notion is, however, hard to test and I consider an alternative, somewhat deeper, explanation for why we have separate cortical areas. The cortical neuropil can be divided conceptually into conducting ('wire') and nonconducting ('non-wire') components: axons and dendrites conduct information over relatively long distances, whereas glia, extracellular space and synapses can be thought of as
doi:10.1016/j.cub.2003.12.040 pmid:14738745 fatcat:vicfpvpaxzgg5mmwafpn7gmcty