Reporting from Lyon

Still in Lyon after attending the cancer modeling workshop mentioned in my previous post.

From a couple of very brief escapades, Lyon seems to be quite a pleasant town but the workshop has been interesting enough that I didn't had a lot of time for tourism. Nice talks from the likes of Philip Maini from Oxford and Vito Quaranta from Nashville and chats with Benjamin Ribba from Lyon have kept me entertained. The word from modelers: multiscale modeling. Lots of researchers producing models studying cancer at all sorts of scales from molecular to tissue and from seconds to years and we still have not got the way to integrate them.

Tomorrow back to Dresden

I am off!!

Only until next Wednesday. The Marie Curie Training Network that sponsors my research here in Dresden is organising a meeting of all the scientists involved in the different projects it manages. The meeting is this monday in Paris!!

It will not be my first time in Paris but I am still looking forward spending the weekend there and meeting some friends.

Review: "From artificial evolution to computational evolution: a research agenda"

A group of people whose work I respect (Dr. Miller was examiner at my PhD viva and I met Prof. Banzhaf in EA conferences) have recently written a paper entitled “From artificial evolution to computational evolution: a research agenda” that was published in the latest Nature Review Genetics (vol 7, page 729):

Despite the journal in which they decided to publish it, the paper is addressed mainly to computer scientists working in the field of evolutionary computing. Researchers in this field use algorithms inspired by evolution in order to solve problems of optimisation in all sorts of field of engineering. Say you have to find the parameters that optimise a set of equations. If you encode these parameters into a string of number and create a bunch of these strings initialising them with random values you can use selection and crossover to find values that optimise the equation. Since not all the strings will produce the same results in the equation, we can discard the worst performing ones and fill the space they left with variations of the best performing ones. If we iterate this algorithm a number of times, thus producing successive generations of the initial population of strings, we are likely to obtain sets of parameters that, if not optimal, will likely to be fairly close to it.

This neat idea of using evolution in engineering (or even in art! I know a few examples of people that have used evolution inspired algorithms to produce music or paintings) has produced some interesting results but several people have already found that the very simplistic interpretation of evolution that computer scientists and engineers use in their algorithms is no match for the real thing. Real natural evolution (as opposed to artificial one) is both creative and open-ended.

Banzhaf et al identifies some of the shortcomings of traditional evolution-inspired algorithms and proposes a number of improvements framed in the new context of Computational Evolution (CE). This seems to consist, mainly, on adding extra bits of reality in the abstraction of evolution used by engineers and computer scientists in order to provide evolution with some complexity to play with. By doing this, for instance embedding it into analogue electronic circuits, they hope that artificial evolution will be successful were it was not before: solving ill-defined open-ended problems.

While I entirely agree with the idea that the original evolution-inspired algorithms could be significantly improved by enriching the stuff on which evolution works more complex and life-like, I am not sure that what they suggest is so ground breaking as to call the field a different name. Besides, many of the suggestions mentioned have been in use by computer scientists (especially the authors of the article) for some time (for example: more biologically plausible genotype-phenotype mappings, on which people like Peter Bentley or Julian Miller have done very useful work).

In any case I would not like to sound as if I did not like the paper. I did and I think that despite some further objections (such as: why don’t they explain why the bits of nature that they decided to pick are the really necessary ones?). Actually I would like to join my voice to theirs and suggest a further use of computational evolution – A more realistic model of evolution can be used not only to do engineering but also to study evolution per se. I would definitely be interested (and in a way that is what I do for a living) in using Computational Evolution to study evolution from a theoretical perspective.

Genes and cancer

I start the week with a post about something really exciting that I read in Science. Unfortunately my institution does not have access to articles in Science published online before they have been printed on paper so I had to be satisfied for the time being with the reports pusblished by conventional media like the Washington Post.

It seems that researchers have screened for and found 189 genes that are altered in colon and breast cancers. Although we are talking about only two types of cancer, breast and colon cancer are two of the most diagnosed cancers in the western hemisphere. It is remarkable that both types of cancer share very few cancer-related genes and that most of the genes discovered to have a role in these cancers have not been known to be so before.

Tumour supressor gene and aging

Read at the NYT: Researchers at the universities of North Carolina, Michigan and Harvard have found that p16 gradually inhibits the proliferation capabilities of stem cells when they reach certain age. The mechanism is useful to prevent the proliferation of cells that, due to their age, have a significantly increased probability of creating tumours.

The paper reporting the research will be published in Nature. One interesting comment by one of the authors is that in his opinion aging is not random but an anticancer mechanism. I find this observation plausible but having an interest in evolution I cannot help thinking that the reason for aging could also be that once an organism has fulfilled its replication duties, its evolutionary-shaped genetic program does not care much for the long term survival of the individual. In other words, evolution does not favour individuals who are good at surviving for ever but that are good at surviving for long enough as to have lots of equally successful offspring.