Books and Arts

Nature 439, 19-20 (5 January 2006) | doi:10.1038/439019a; Published online 4 January 2006

A robust approach

Eörs Szathmáry1

The functional overlap between different components protects biological systems.

BOOK REVIEWED Robustness and Evolvability in Living Systems

by Andreas Wagner

Princeton University Press: 2005. 408 pp. $49.50, £32.50

When sitting on an aeroplane, we obviously hope that it won't crash. A tacit assumption behind this wish is that our biological system isn't about to crash either. It so happens that these systems share several features. Both have specific parts that serve certain functions. The plane was designed by engineers, who were in turn designed by evolution through natural selection. Both systems seem robust and yet fragile, but how can we reconcile these two seemingly opposing features? One answer is that they are robust and fragile to different perturbations, being particularly robust to perturbations that are common in their 'niche'. Another answer is that robustness can be in a trade-off with other features, such as price and reproduction rate.

In his book Robustness and Evolvability in Living Systems, Andreas Wagner deals with some hot issues of current biology. As well as the terms in the title, the main keywords are 'neutral spaces', 'redundancy' and 'networks' — all of them highly fashionable, no doubt. But fashions fade, and usually a fraction of the original claims remain robust.

The robustness in biological systems is a consequence of their complexity. Many biologists are suspicious of ideas and models in the field of complexity, but I would encourage them to set their suspicions aside and read this book. Wagner's treatise is more than good biology; it is also very interesting biology. The picture is painted by talented hands. Wagner surveys many relevant examples, from the genetic code to organismal design, taking in properties of ribonucleic acids and proteins, the fascinating robustness of metabolic pathways and networks, the inelegant but robust genetic networks in development, and the many developmental pathways that can lead to essentially the same adult form. The level of detail is adequate in most cases and the questions and explanations are lucid. The mathematical treatments are relatively easy to follow and deliver real insight.

One of the book's chief merits is that the author knows a lot about many levels of biological organization, something that can't often be said for the 'complexologists'. Many erroneous claims about, and faulty models of, various biological networks have sprung from a lack of knowledge about the complexity of biological phenomena.

If I have a favourite aspect of the book, it is the meticulous yet insightful analysis of neutral spaces and their relevance for the main themes of the book. Structures in biology can be envisaged as being embedded in a suitable 'space'. Protein-sequence space, for example, is multidimensional and discrete, as polypeptides are made up of a whole number of amino acids. There are 20 amino acids, so if the length of a polypeptide is, say, 100, then any given polypeptide chain has 1,900 (that is, 100times19) nearest 'neighbours', all of which differ by only one amino acid from the reference sequence. Many of these neighbours have the same biological function, because altering one amino acid has little effect. But some proteins much farther away in this space (that is, with many different amino acids), may also have the same function. If you imagine that the space at all these 'neutral variants' is shaded, then the shaded areas make up the original protein's 'neutral space' and indicate the organism's robustness.

My only grumble is that I would have used the word 'domain', as the neutral region is a subregion of the whole protein space, but I hope no great confusion will arise. The author shows clearly that although neutral-space analysis supports old principles of population genetics in several cases, sometimes it does not. Consider, for example, the limited validity of the Haldane–Muller principle, which states that the mutational load depends only on the mutation rate. Mean fitness is found to depend on the robustness of the population, which is actually a function of the structure of the neutral space.

I would like to know more about how the nervous system fits into the framework of Wagner's book. We know from studies ranging from the synapse to cell shape and functional neuroanatomy, for example, that at different scales a great many parts of the brain can have essentially the same function, in line with Wagner's analysis of 'distributed robustness'. But I realize that even the best story must end somewhere and leave space for future ones.

  1. Eörs Szathmáry is at the Collegium Budapest, 2 Szentháromság utca, 1014 Budapest, Hungary.


These links to content published by NPG are automatically generated.


Biomathematics Merging lines and emerging levels

Nature News and Views (02 Apr 1998)

Extra navigation




Robustness and Evolvability in Living Systems

Amazon: US | UK | Japan