Evolution@home/Interview en

Aus Rechenkraft
Wechseln zu: Navigation, Suche
In Mai 2008 Rechenkraft (RKN) made an interview with the author of Evolution@home, Laurence Loewe (LL).
RKN:
Hi Laurence, when did you learn about globally distributed computing for the first time?
LL:
In 1999 I read a newspaper article about GIMP. I thought the idea was fascinating, as I already had computing time problems back then.
RKN:
Did you donate computing time to other globally distributed computing projects by downloading a worker to your computer before evolution@home? If yes, which project did you support?
LL:
Yes, I once contributed to SETI@home (classic) to see how it works. Otherwise my CPUs were very busy with my own problems.
RKN:
When did you come up with the idea to start your own globally distributed computing project?
LL:
By 2000 it was clear to me that my computing time problems would not go away. I noticed back then that I will probably always be interested in some evolutionary models that I will not be able to compute with whatever computing power I have. Global computing simply allows me to extend the horizon of questions that I can address.
RKN:
Who supported you and what did the other scientists at your institute think?
LL:
I was always lucky to have supervisors that gave me a lot of freedom. This allowed me to start evolution@home and to develop it. After the initial launch, people from all over the world started to contribute computing power, Rechenkraft.net included. This showed my colleagues that the idea works in principle.
RKN:
The project is going on for a few years now. What exactly is being simulated?
LL:
Generally evolution@home deals with simulations of evolution and related questions. Some parts of these problems are often easy to parallelize. Since problems in evolutionary theory are often very different, I decided to partition them into different simulators that can then be distributed globally. The current Simulator005 investigates what happens if genetic systems have to survive an onslaught of slightly harmful DNA changes for extended periods of time without recombination. This problem is important for understanding the importance of sex, as recombination is the main function of sex from an evolutionary perspective. In this context one particular model is very interesting. It is called Mullerâ??s ratchet. Understanding this model allows predicting how fast slightly harmful DNA changes can accumulate.
RKN:
What did you find?
LL:
There are several types of results. One type of result belongs to the category of computer science. Evolution@home started as one of the simplest global computing systems that could possibly be conceived. It was not clear if such a simple system would actually work in practice. By now this can be answered with a clear Yes and in the process the limits of the system became clear. These results have been published (link)

The biology, however, is much more important. Evolution@home has become the largest database of Mullerâ??s ratchet simulation results that currently exists. Even if this database is not (yet) public, it can answer some relatively complex biological questions by reasonably simple queries. This allows testing of hypotheses about Mullerâ??s ratchet.

For example, a few years ago scientists wondered about the high mutation rates that have been observed in human mitochondria. These appeared to be too high considering that there is (apparently) no recombination in this genetic system. I could demonstrate with the help of evolution@home that this is indeed the case. If current textbook knowledge about human mitochondria is combined with the standard model of Mullerâ??s ratchet and these observed mutation rates, then one would expect that the evolutionary line that leads to humans should have become extinct within the last 20 million years. This shows that the current standard model needs refinement and in the corresponding research paper I list more than 10 potential solutions for this paradox (PDF, 1.7MB).

RKN:
Do scientists take these results serious, even though they have been computed by the PCs of ordinary Internet users from all over the world and not by bought CPU time on supercomputers? What do your colleagues say?
LL:
The results are being taken seriously. After all one has to demonstrate in both cases that the results make sense biologically and that they donâ??t go back to programming errors.
RKN:
Will more publications follow?
LL:
Yes. After investigating Mullerâ??s ratchet in human mitochondria a series of other organisms are waiting to be analysed. There are a number of asexual genetic systems, where people have speculated that Mullerâ??s ratchet should have driven them to extinction. I believe that it is important for this type of research not just to play with words, but to test clearly defined hypotheses. Future publications will be listed here.
RKN:
What does the future hold for evolution@home?
LL:
Besides additional tests of the consequences of Mullerâ??s ratchet in various asexual species, I also plan to analyse more complex models to further develop the theory of Mullerâ??s ratchet.

I also work on the next simulator that will look at a different question. That simulator will use a few specific examples to analyse the question how many advantageous and harmful DNA changes can be expected in molecular systems. The systems investigated by Simulator006 are circadian clocks. These molecular clocks are responsible for waking us up in the morning and making us tired in the evening.

After Simulator006 there will be others that target various questions in evolutionary biology. There is much work to do in this area.

RKN:
What did you gain from the integration of evolution@home into BOINC and what do you expect from this in the future?
LL:
In one word: computing power (= English translation of Rechenkraft). In the first 5 months of the integration over 120 years CPU time were contributed and up to several hundred PCs were running at the same time. Currently Iâ??m working on adjusting my analysis algorithms and my workflow so that the infrastructure can cope with this. I expect the BOINC connection to speed up many interesting research projects.
RKN:
What server hardware do you use for the project?
LL:
EvoHo Plone Server 500.JPG
EvoHo Data Storage 500.JPG
Nothing special at the moment. Besides a Plone webserver for the website I need a few hard disks to store and backup all corresponding results.
RKN:
How much storage capacity does the whole project require at the moment? How many MB are generated each week?
LL:
That depends on how this is calculated. At the moment results are stored in a very inefficient way. That is one of the main problems that I need to solve in the near future. Currently evolution@home data take up almost 100 GB on my hard disks. With an appropriate infrastructure many redundancies could be eliminated and this amount could shrink considerably. Since BOINC started to contribute about 50 MB compressed data arrive each week, but to do anything with it I currently need to unpack this.
RKN:
Evolution@home is analysing human mitochondrial DNA and the question whether the alleged lack of recombination and repair in this system could have led to the extinction of the human evolutionary line â?? and the time frame this could have happened in. You claim to have found that on the basis of the assumptions used by evolution@home humanity should have gone extinct long ago, because its mitochondrial DNA should have been degraded by the accumulation of slightly harmful DNA changes. Can you explain this a bit better: How exactly does evolution@home compute this and how reliable are the assumptions behind the program, how does the algorithm work?
LL:
These simulations are implemented in Simulator005 of evolution@home, which uses a number of assumptions. Essentially, these are based on standard textbook knowledge and on the most exact observations that exist so far. For example, so far no one has seen mitochondrial DNA recombine in the human germ line. What has been seen is that mitochondrial DNA mutates rather quickly. Simulator005 includes these observations in a simple population genetical model which is also known as the standard model of Mullerâ??s ratchet. This model is simulated in order to predict how fast slightly deleterious mutations accumulate. Taken together this is rather normal, solid modeling which makes simplifying assumptions that are typical for this field (to avoid drowning in details). One such simplification for example is that Simulator005 does not allow for backmutations, which could turn out to be critical. For a more detailed description, Iâ??d like to refer readers to the corresponding scientific paper (PDF, 1.7MB).

Surprisingly, this standard model produces all but â??standardâ? results. Based on these calculations one would expect that the human line of evolution would have gone extinct within 20 million years â?? if the model is complete and todayâ??s processes occurred at similar rates in the past. The last assumption is obviously difficult to test and allows for various possibilities to resolve the paradox. However, the most important task is not to resolve the paradox no matter how, but rather to do so based on actual evidence. In the scientific report that presents the results I discuss the extensive list of the various potential solutions that might resolve the paradox. Some of them are more credible than others. That paper contains so far the most comprehensive list of potential solutions for genomic decay paradoxes, totaling 30, and 10 of these could play a special role in saving mitochondria.

The next big challenge will be to determine which of these many potential solutions play an actual role in reality. To do this many more simulations will be needed in addition to new simulators.

RKN:
Now that evolution@home is working on the question why we are still alive even though mitochondria should have been degraded a long time ago, can you tell us how evolution@home plans to tackle this question? Is it possible that recombination and repair do exist in mitochondrial DNA and that we simply havenâ??t found the corresponding evidence so far?
LL:
Of course it is a possibility that recombination has not yet been detected in the germ line and recently there was a debate about that. However, the current evidence suggests that there is no recombination in the human germ line. Personally, I think other solutions to this paradox are more likely at the moment. For example, the current analysis ignores all back mutations, advantageous mutations and all mutations that compensate for the damage done by other harmful mutations. It might well be that including these processes in a model will solve the paradox already. They will certainly do this if enough advantageous mutations are included, but independent work will have to establish how frequently these mutations can be expected to occur. In recent years I spent much time investigating the question how mutational effects are distributed. That is a very challenging field. The next simulator (S006) will also target this question, albeit in a different model system. Modern molecular systems biology might be able to help here.
RKN:
Is it true that you were a PhD student of Prof. Siegfried Scherer at the Technical University of Munich? He happens to be well known in public as someone with affinities to creationist or intelligent design ideas. Apparently, he cannot imagine that all the complex molecular machines could have been produced by a modern version of Darwin's theory of evolution. He likes to cite the example of the bacterial flagellar motor. What do you think about all this?
LL:
During my time as a PhD student of Siegfried Scherer I experienced him as being very fair in the way he dealt with people who did not share his convictions. Although I criticised his theories considerably, he did not jeopardize my PhD work. He also changed his point of view considerably in the last few years, so I can't even tell you what he himself believes at the moment. Thus he seems to be willing to learn. However, as a microbiologist, his population genetical understanding is rather limited.

I wasn't convinced by his case for a non-evolutionary origin of the bacterial flagellar motor. At the moment, I do not know any scientific theory that could explain the bacterial flagellar motor more elegantly than the modern theory of evolution. This is true despite some questions that one might want to ask about corresponding models.

RKN:
Thank you for the interview.
LL:
You are welcome.

Eigene Werkzeuge