Systems biology is taking off.

There is a revolution occurring in the biological sciences. It took off just a couple of years ago and is now clearly visible in the literature. Some scientists in the field like to refer to the development as the birth of systems biology, whereas others prefer not to put a label on what is happening. Modern molecular biology was born with the discovery that genetics is based on nucleic acid chemistry (Watson and Crick 1953), and one way to define it is to say that molecular biology is a large box of tools to do genetics by manipulating DNA. This definition may sound disheartening, but its positive side is that the tools can be applied to all aspects of biology to solve essentially all scientific problems that may arise. One result of molecular biology is large-scale sequencing of genomes from a rapidly growing number of organisms. Genome sequencing is not possible without the use of computers with large memory and tools to handle the enormous amounts of data that are generated in the massive sequencing efforts. The need for data handling led to another box of tools, called bioinformatics, which is now an established part of molecular biology. However, when all this sequence data got into computers, it became obvious that the genetic blueprints by themselves tell us very little about the functional behavior of cells and multicellular organisms; that is, about what we really want to know about biological systems. In this way, the human genome project, which is perhaps the most spectacular success of molecular biology, also meant that a vast space of future research of a radically different kind became visible. To understand the causal connections between genotype and phenotype will require a very significant expansion of the traditional toolbox used by molecular biologists. It must include concepts and techniques from many other scientific disciplines such as physics, mathematics, numerical analysis, stochastic processes, and control theory. Many novel tools that do not exist today must be forged to understand how dynamic, adapting, and developing systems can emerge from the information buried in the genomes. The development of such an extended toolbox for quantitative reasoning about the dynamics of living systems, and the application of its contents to solve a variety of scientific problems, is one way to define systems biology, analogous to our definition of molecular biology above. It is our belief that systems biology will enrich the biological sciences and transform our thinking about biological problems, in analogy with what has been happening in molecular biology during the 50 years that have passed since the discovery of the double helix. Systems biology will always bring the functional aspects into focus, sometimes close to genomics and sometimes far out in areas not visited before. Below will follow some examples of what we consider significant developments of systems biology, which is still in its infancy but has great future promise. The selection of topics is limited by the format of this mini-review, and many important contributions could therefore not be covered. Models of Growing Bacteria One question that has been with us for a long time is whether bacteria have evolved to maximize their growth rate (Ehrenberg and Kurland 1984), and in its simplest form, it refers to logarithmically growing cells in media containing different types of carbon sources. It is now possible to use genomic information to reconstruct the metabolic pathways in whole cells (Edwards et al. 2001; Forster et al. 2003). The freedom of choices of pathways can be reduced by in silico models of metabolic flows, in which physical and other constraints are introduced to decrease the number of degrees of freedom. Interestingly, the growth rate of bacterial cells is accounted for in such in silico models and can be maximized by simple linear optimization techniques (Ibarra et al. 2002) and then compared with the growth rate of real bacteria. It turns out the Escherichia coli bacteria grow with maximal rate according to the prediction of its in silico model in media containing a number of common carbon sources. In contrast, when the carbon source is glycerol, the growth is much slower than that in the optimized in silico model. When, however, the cells are allowed to grow in the presence of glycerol for a large number of generations, they reproducibly evolve toward a maximal growth rate, as predicted by the optimized in silico model. In this case, the step from the annotated genome sequence of E. coli to predictions about how fast the organism can grow is amazingly short. Because rapid growth correlates positively with the population genetic fitness parameter, it is now possible to make in silico simulations of the evolution of bacteria that lead to predictions that are experimentally testable. Another case, in which an analysis of metabolic flows in growing bacteria leads to a number of interesting testable predictions, concerns amino acid limitation. In all organisms, many amino acids are encoded by several synonymous code words (Crick et al. 1961), and these are often read by several tRNA isoacceptors (Björk 1996). When such an amino acid becomes rate-limiting for protein synthesis in the cell, the charged levels of the members of a family of isaoaccepting tRNAs will react very differently to the deficiency. One tRNA, or a few tRNAs, will totally lose its charging with amino acid, whereas others can remain almost fully charged even when the rate of supply of the amino acid becomes negligible (Elf et al. 2003). This result, following from a very simple model, has been used to rationalize the choice of synonymous codons in control systems for transcriptional regulation (attenuation of transcription; Yanofsky 1981), as well as in genes encoding enzymes that synthesize amino acids. These results also indicate that the codon adaption index (CAI) for highly expressed genes (Sharp and Li 1987) should be complemented with a starvation CAI (sCAI) for gene expression during amino acid limitation.