Stamp-Collecting in the Post-“Omic” Playground

“My own scientific career was a descent from higher to lower dimension, led by the desire to understand life. I went from animals to cells, from cells to bacteria, from bacteria to molecules, from molecules to electrons. The story had its irony, for molecules and electrons have no life at all. On my way life ran out between my fingers.” -Albert Szent-Gyorgi, The Living State, Academic Press, 1972

As we work out intricate details regarding protein-protein interactions and discover how genes are regulated in specific cellular environments, are we tending towards a better understanding of the cell? Or are we merely stamp collecting, mindlessly cataloguing useless details of cell function? Like Szent-Gyorgi in the quote above, most molecular biologists work according to a reductionist paradigm. Reductionism is any theory or method that seeks to explain a complex system by the behaviour of limited numbers of simpler components. This approach has been fruitful for centuries and is the foundation on which the scientific method is built. Its usefulness is especially evident in biology, where scientists working according to the reductionist paradigm have made tremendous advances in understanding the molecular basis of cellular life and disease. Nevertheless, the limitations of the reductionist paradigm are becoming increasingly obvious in the present “omics” era. Cells are complex non-linear systems, and their behaviour is difficult to predict, even with detailed knowledge about their components. For example, even though the complete genome sequences of many organisms are available, it is clear that knowledge of cells at the DNA level alone is insufficient for even a basic understanding of the processes collectively known as “life.”

With the “omics” era now upon us, many scientists hope that one day their efforts will lead to a “wiring diagram” of the cell. They optimistically envision that all the interactions between proteins and other cellular components will be known and that the strengths of these interactions will have been tested empirically and documented. But even if such dreams are fulfilled, how much closer will we be to truly understanding cells and being able to predict their behaviour in a given environment?

A study by Guet et al. (1,2) offers an interesting perspective to this question. The authors constructed a series of plasmids where each plasmid could express three well-studied transcriptional regulators in bacteria: LacI, TetR and Lambda CI. Each of these three transcriptional regulators was driven by one of five promoters (Fig. 1A and B). The promoters were chosen so that each one was either repressed or activated by one of the three transcription regulators (Fig.1A). Guet et al. constructed 125 plasmids, each containing the three transcriptional regulators under the control of the different promoters in various combinations (Fig.1B). These plasmids were introduced into E.coli, which were grown in different conditions (in the presence and absence of repressors such as IPTG ) and the regulatory genes on each plasmid activated or repressed one another in various ways, generating networks with diverse connectivities (Fig.1C and D). To detect how the genes were being regulated in the cell, the plasmids were designed to express or suppress a reporter Green Fluorescent Protein (GFP).

Figure 1: Experimental design of Guet et al. A) The five promoters used in this study (PT, Pλ+, Pλ-, PL2, and PL1) are either repressed or activated by LacI, TetR or λcI. The two small molecule repressors used in the study, aTc and IPTG inhibit the gene products LacI and TetR respectively. B) The five promoters were cloned combinatorially in front of LacI, TetR and λcI to generate 53=125 plasmids. In each of the plasmids a Green Fluorescent Proteins (GFP) gene was cloned under the control of a Pλ- promoter. The product of this “reporter” gene produces detectable fluorescence. C) Each plasmid was introduced into bacterial cells and the cells were grown in four conditions (in the presence or absence of the small molecule repressors IPTG and aTc). By introducing these plasmids into cells, Guet et al. created mini-regulatory networks whose end-point was the suppression or expression of the detectable GFP gene. D) If the LacI, λcI and TetR genes are under the control of the promoters PL1, PT, and PL1 respectively, a genetic network is created in which the presence of IPTG should result in the expression of GFP.

In principle, knowing the design of the plasmids should have allowed the authors to predict how the engineered networks regulate the expression of the GFP gene. For example, introduction of the plasmid diagrammed in Figure 1D into the bacteria should create a regulatory network that results in the expression of the GFP gene in the presence of IPTG. The authors found that this is not always the case, as they observed many surprising host behaviours (that is, the expression of the GFP reporter). For example, they showed that circuits of the same design but built of different components sometimes show very different activity. The study points to the idea that very simple cellular networks, built out of a few well characterized components, are not always predictable. The authors point out that genetic networks are “nonlinear, stochastic systems in which the unknown details of interactions between components may be of crucial importance”.

The study raises difficult questions about the nature of genetic networks. If we are unable to understand and predict the behaviour of such simple systems, how will we ever understand more complex networks like entire cellular pathways that lead to disease? The masses of data unloaded by microarray experiments and proteomic efforts allow researchers to clump proteins and genes into pathways and functional groups. Will these efforts translate into useful information about cells?

The study conducted by Guet et al. exemplifies a scenario in which detailed knowledge about a system’s components does not allow us to truly understand the dynamics of a system or to make useful predictions about its behaviour. It is the reductionist scientist’s nightmare. Below is presented an opposing example, which illustrates that complex systems sometimes behave according to simple rules and that it is not necessary to understand the components of the system in detail in order to predict how the system will behave.

The Brain-Computer Interface (BCI) field is a division of neuroscience interested in monitoring neural activity and decoding it into meaningful information using mathematics and computers. In one study, Wessberg et al. (3,4) were able to teach a monkey to control a robotic arm by way of electrodes implanted in its brain. In the experiment, the monkey was taught to follow a cursor on a screen with a robotic arm. The group then implanted a set of electrodes into its motor cortex that enabled them to monitor neuronal signals as the monkey moved its arm to control the joystick. The speed, direction and force of the movements were recorded and correlated with the neural activity detected by the implanted electrodes. Wessberg et al. developed simple mathematical algorithms that reliably and accurately translated the neuronal activity or “intention” of the monkey to movements of the robotic arm. Their experiment was validated when the monkey was able to control a robotic arm and follow the cursor on the screen using thought alone.

Wessberg et al. were able to predict the behaviour of a very complex system – the brain – based on a very limited number of variables, namely the on/off firing pattern of only a subset of neurons. Had Wessberg et al. decided to apply a reductionist approach to understand how the motor cortex of a monkey is involved in moving the robotic arm they may never have been successful.

The goal of molecular biology is to understand, control, and predict certain aspects of cellular behaviour by understanding the cell at a molecular level. It is encouraging to think that an experimental paradigm like that used by Wessberg et al. may be applied to problems in other areas of research, such as cell and cancer biology. Maybe it is possible to develop models for cell growth and division without knowing the intricacies cellular mechanisms such as protein-protein interactions (Fig.2)? There are many examples of “multi-faceted” or complex diseases, such as pernicious anemia, diabetes, and syphilis, that were treatable before their complexity was known or understood.

Figure 2: Studying the cell as a "black box". One way of studying the cell is to consider only the inputs into a cell (such as environmental stimuli) and the cell’s responses to those inputs. The cell itself is treated as a black box about which we have little detailed information.

Of course a certain amount of knowledge about a given system is necessary in order to extract its critical parameters; however, in this era of complete genome sequences, mammoth proteomics projects, deluges of microarray data, and automation of almost every conceivable experiment, is it not possible that we are collecting information and processing it in a way that gives little insight about cell function? If we are not careful, one day, as we decode the mountains of data uncovered by our high-throughput experiments, we may realize that “life” – or whatever it was that we initially yearned to understand – “ran out between our fingers.”

A popular analogy compares the genome sequence of an organism to a “parts list” of an airplane. In the postgenomic era, we are trying to assemble the full wiring diagram of the airplane and an instruction manual for building it. As the experiments of Guet et al. suggest, this may not be enough to understand why the airplane flies. This analogy can be extended: even with an instruction manual and wiring diagram, we would have a difficult time understanding why an airplane flies if we did not know about aerodynamics, electricity, and jet propulsion. Perhaps we have yet to discover fundamental biological, chemical, or physical laws that underlie the very essence of cellular life, and develop a new mathematical language to describe it. Maybe a completely novel experimental paradigm and scientific method must be applied in order to reach the next level of understanding cellular dynamics. A similar idea is articulated by Paul Nurse in his discussion of the cell cycle:

“Newtonian physics operates in the three-dimensional space and time of everyday experience and is easily contemplated by a human mind, which has evolved to function in such an environment. With relativity and quantum mechanics, physics moved from this accessible common sense world into a far more abstract one, much more difficult for the human mind to imagine and conceive. Perhaps a proper understanding of the complex regulatory networks making up cellular systems like the cell cycle will require a similar shift from common sense thinking. We might need to move into a strange more abstract world, more readily analyzable in terms of mathematics than our present imaginings of cells operating as a microcosm of our everyday world.” –Sir Paul Nurse (5)


1. Guet C., et al. Science 296, 1466 (2002)

2. Wigler M, Bud M. Science 296, 1407 (2002)

3. Wessberg, J. et al.. Nature 408,361 (2000)

4. Hoag H. Nature, 423 (2003)

5. Nurse P. Cell 100, 71 (2000)

Share your thoughts

Leave a Reply

You must be logged in to post a comment.