The Panda's Thumb Legacy Archive — pandasthumb.net

Evolvable by design

Posted 30 September 2007 by PvM

The work by Panos Oikonomou, a graduate student at the Physics Department, University of Chicago, recently caught my eye. His website explains that he is

... interested in studying the relationship between network topology, dynamics and evolution. I explore possible evolutionary advantages of such features, like the scale-free distribution.

He recently wrote a paper "Effects of topology on network evolution", Panos Oikonomou and Philippe Cluzel, Nature Physics, August 2006. In the paper, the authors compare the characteristics of a random network versus a scale free network. A random network is one in which each node has on the average the same number of connections to other nodes. For scale free networks, the connectivity follows a power law distribution. They tested how the two different networks 'responded' to evolutionary processes

Our simulations show that populations containing these scale-free networks can easily produce a number of functional variations which allow each population to evolve rapidly and smoothly towards some target function. By contrast, equivalent random networks evolve slowly, through a succession of rare fortuitous random mutations.

The work is particularly relevant because 1) scale free networks can be found at all levels in nature 2) scale free networks can be explained by processes as simple as gene duplication and preferential attachment. What is even more interesting is that

For systems randomly connected it is necessary to invoke a specific tuning of their connectivity in order to access the target faster, however such fine-tuning is not required for scale-free networks.

In other words, scale free networks seem to have many features which make them very suitable for evolutionary processes. So next time you hear ID proponents argue that the networks of interactions of genes inhibits evolution, you may ask them about scale free networks. In addition, one may ask them how ID explains these findings? "The problem of biology is not to stand aghast at the complexity but to conquer it”, Sydney Brenner, Discover Dialogue, April 2004 As to the origin of complexity? These findings are but one piece of the puzzle being slowly unraveled by science. What has ID done for science lately?

17 Comments

James · 1 October 2007

I wonder what random selection of irrelevant canards Mats is going to trot out this time. yawn...

hoary puccoon · 1 October 2007

I'm not quite understanding this. Is Oikonomou saying 'systems' ARE connected through scale-free networks, or that they COULD BE connected through scale-free networks? And by systems, I'm assuming he's talking about something to do with the ways genes are expressed. But it's really not clear to me.

If everbody else got this immediately, feel free to jump in and tell me I'm an imbecile. But if other people are struggling, too, maybe a little additional clarification?

Braxton Thomason · 1 October 2007

Hoary, I believe the point of the paper is not to discuss 'real-world' systems, but to look at artificial systems and see what has the best potential to "evovle" -- so, for the purposes of this research, there are both randomly connected networks and scale-free networks.

Pim goes on to say that for a lot of natural systems, scale-free networks describe them very well.

PvM · 1 October 2007

Braxton is right, the authors study two different kind of networks, one is what is called a random network, the other one a scale free network. In random networks, every node has on average the same connections to other nodes, for scale free systems, the distribution of connections follows a power law. This means that there are a few nodes with many connections and many nodes with few connections.

The relevance of the research to evolution is that at many levels, networks in biology are found to be scale free. Protein networks, regulatory networks etc all follow a scale free distribution.

see http://www.alexeikurakin.org/img/s2l4.jpg at http://www.alexeikurakin.org/main/lecture4Ext.html for example.

PvM · 1 October 2007

For a quick intro to scale free networks Network topographies and evolution are quite an interesting topic, particularly given the next paper I will discuss Samarth Swarup Les Gasser Unifying evolutionary and network dynamics Phys. Rev. E 75, 066114 (2007)

Many important real-world networks manifest small-world properties such as scale-free degree distributions, small diameters, and clustering. The most common model of growth for these networks is preferential attachment, where nodes acquire new links with probability proportional to the number of links they already have. We show that preferential attachment is a special case of the process of molecular evolution. We present a single-parameter model of network growth that unifies varieties of preferential attachment with the quasispecies equation (which models molecular evolution), and also with the Erdo-double_acute s-Rényi random graph model. We suggest some properties of evolutionary models that might be applied to the study of networks. We also derive the form of the degree distribution resulting from our algorithm, and we show through simulations that the process also models aspects of network growth. The unification allows mathematical machinery developed for evolutionary dynamics to be applied in the study of network dynamics, and vice versa.

Ask yourself: What has ID done for science lately?

Stanton · 1 October 2007

Ask yourself: What has ID done for science lately?
Lately? Nothing much, besides DI's typical load of ignorant nonsense and that farting video.

harold · 1 October 2007

The diagrams of the scale free and random networks also remind me of what neuronal connections look like in some invertebrates, with some showing a neural net similar to the random network, and others, usually perceived as being later lineages, showing a more ganglia-based anatomy that resembles the diagram of the scale free network.

David B. Benson · 1 October 2007

To what extent are mammalian nervous systems scale-free?

sparc · 1 October 2007

Ask yourself: What has ID done for science lately?
In the moment they are quote mining this paper:
Ioannidis JPA (2007) Why Most Published Research Findings Are False PLoS Medicine Vol. 2, No. 8, e124 doi:10.1371/journal.pmed.0020124
It's not quite clear why ID-creationists then want their to be accepted as science. Maybe, they think they are on the safe side because ID-creationism doesn't generate any research findings.

hoary puccoon · 1 October 2007

Braxton, PvM,
Thanks. What I'm getting from this is that Oikonomou basically came up with an 'existence proof' as Francis Crick called them, of how scale free networks COULD work in the expression of genes. (And very likely do work, although we don't know for sure.)
In any case, the ID argument that "the networks of interactions of genes inhibit evolution" is not correct.
Am I getting warmer?

harold · 1 October 2007

David B. Benson -

Good question, although a complicated one.

A big difference is that many simple invertebrates, the types I was thinking of, don't have what we would term a "central nervous system". (This is obviously not true of all invertebrates, octopi being an extreme example where a CNS is present.)

Mammalian nervous systems are characterized by a central nervous system and peripheral nervous system. The peripheral nervous system can be subdivided into the somatic and autonomic nervous system. The latter can be further subdivided into the sympathetic and parasympathetic nervous system, and both somatic and autonomic systems can be said to have motor and sensory components.

The mammalian nervous system is highly centralized, hierarchical, and anatomically specialized. It is several unique systems running in parallel, actually, and it has some redundancy (quite a bit during development), but less redundancy than the neural net type of NS.

The mammalian nervous system bears absolutely no resemblance to the random network, whereas a Hydra nervous system is quite similar to it.

On the other hand, some invertebrates, such as the famous Aplysia, have nervous systems which, while lacking a definite CNS per se, are organized into ganglia, and which have a striking resemblance to a diagram of a relatively simple scale free network.

Since preferential attachment is known to give rise to scale free networks, it is tempting to think that something analagous to preferential attachment may have played a role in the emergence of ganglial organization of nervous systems.

It might also be useful, in some circumstances, to conceive of the mammalian nervous system of having properties of a huge and complex scale free network. Certainly ganglia and connecting tracts are essential elements of the mammalian CNS, the brain being an especially large ganglion. Although this would be a simplifying analogy or model, it might help explain some things.

Braxton Thomason · 1 October 2007

hoary puccoon: Braxton, PvM, Thanks. What I'm getting from this is that Oikonomou basically came up with an 'existence proof' as Francis Crick called them, of how scale free networks COULD work in the expression of genes. (And very likely do work, although we don't know for sure.) In any case, the ID argument that "the networks of interactions of genes inhibit evolution" is not correct. Am I getting warmer?
I'm not sure we could definitively say that genes exactly match the scale-free network model or what the implications are for "evolvability"; there's probably a lot more research to be done in that area. But yes, I think that is the general thrust. As with everything else, I'm sure that natural systems will turn out to be vastly more complex than we initially think, and there are probably different scales on which scale-free networks can be descriptive. And I'm sure there will be levels of detail where a random network is a descriptive model. As others have pointed out though, it's not the IDiots that are doing the work/advancing the science/understanding.

PvM · 1 October 2007

hoary puccoon: Braxton, PvM, Thanks. What I'm getting from this is that Oikonomou basically came up with an 'existence proof' as Francis Crick called them, of how scale free networks COULD work in the expression of genes. (And very likely do work, although we don't know for sure.) In any case, the ID argument that "the networks of interactions of genes inhibit evolution" is not correct. Am I getting warmer?
More than luke warm :-) Yes, that is where I am going here.

Torbjörn Larsson, OM · 1 October 2007

I'm afraid that the papers conclusion that scale-free networks results from selection isn't persuasive. Scale-free networks are very common and are often a result of small-world networks, ie with next-neighbour connections. With evolutionary mechanisms such as gene duplications and variation, it is a given observation IMHO. But they explain why IDiots are so wrong about randomness in evolution, as it wouldn't work. The more fun question is why flies brains are hard-wired with scale-free “Lévy-flight” characteristics. (Lévy-flight trajectories is the efficient foraging behavior we display when we look for our keys - search the pockets, look on a desk, search a drawer, et cetera. Straight paths punctuated by saccades.) That is probably an early solution for foraging.
To what extent are mammalian nervous systems scale-free?
I don't know, but models of small-world neuronal networks display a balance in spike patterns between synchronized and desynchronized behavior and a complex behavior. This is a situation where mutual information is a relevant measure of complexity, as completely random or completely regular patterns are "non-functional". (Flat "EEG" vs periodic "EEG".)

Torbjörn Larsson, OM · 1 October 2007

That is probably an early solution for foraging.
Foraging for sparse resources, that is. And the hypotheses that it is early draws weak support from that it is so ubiquitous and observed even with small brains. I think.

Frank J · 2 October 2007

As others have pointed out though, it’s not the IDiots that are doing the work/advancing the science/understanding.

— Braxton Thomason
Exactly. And those that do advance the science have no problem challenging each other in public if their conclusions differ. Whereas anti-evolution activists increasingly cover up their differences for the sake of a political "big tent." BTW, I need to find some time to catch up on the disagreements between the authors and Kauffman, some of whose ideas I find convincing. I suspect that my agreement is on a more general level, though. I hope I can understand enough of the detail (& jargon) to figure who has the better case. On that note...

Ask yourself: What has ID done for science lately?

— PvM
I think that they are trying to show that one can have everything both ways, which is music to the ears of the millions - not all fundamentalists - who find science too "conservative." Kauffman provides a neat example; at various times he was portrayed as a "Darwinist" and a "fellow dissenter."