lundi 8 février 2010

Symplectic Biology

Je reçois ce courrier. Captivant...

"Dear friends,

At a time when we are about to create a new journal (Symplectic Biology), it is important to bear in mind a deep reflection of the situation of publications at the present time.

Please read this very significant paper:

http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal.pmed.0050201

Our aim is *not* advertisement but first, *creation of knowledge* and second, dissemination of knowledge. This is why we wish to create several new tracks for publication in our journal. The deepest one is the creation of an automatic way to extract knowledge present in the articles published in the journal. I will keep you posted about our efforts in the domain. One important aspect is that the journal will be progressively (and I hope fast) taken in the hands of young scientists, with the expectation that they might avoid the traps which caught their elders.

It seems essential, at present, that we begin to shift from trends which dominated for a century, at create some new way of recognizing creativity. This is the most important as the increase in noise has become enormous. The internet is extremely useful, but so noisy that it becomes sometimes useless or counterproductive.

Typically an effort such as wikipedia, which appeared initially to be an excellent idea, is leading to nowhere, or worse, to the loss of knowledge. The reason is well understood by those who studied the primitive model of neuronal networks, the perceptron. Learning begins, from either one of two extremes, a random situation or a uniform situation. And, indeed, learning proceeds. So, this looks good. However learning reaches a plateau, and then, unfortunately, it begins to be contaminated by noise, so much that, after some time, the noisy situation prevails. This is exactly what is seen with wikipedia, which reached a more or less average good level (thought to be similar to that of some encyclopedia) but which now is completely unreliable (except in very specific domains such as mathematics, where the community is small enough to exert selective control).

The reason is that there is no selection, and no limitation in length of the papers. One cannot *vote* for knowledge. And there is nothing such as "democratic knowledge".

We can see that even in specialized situations such as genome annotation: there must be some sort of control over the quality of knowledge which is put in the annotations. Otherwise percolation of errors prevail, and misinformation is much worse than lack of information.

Amusingly, all this reflection relates to our discussion (still ongoing) about information in biology.

A central question is to have an agreement on what life is. And, to this aim we need *first* to make a copernician revolution, i.e. to abandon the idea to treat multicellular organisms (and man in particular) as the centre of the living universe. Multicellular organisms introduce a lot of further complications that are not interesting for our purpose. Indeed they are made of individual cells, and we need to understand first what happen to them, before considering the whole. Note that this goes very much against fashion, in particular because of the dominance of medical research, and of the enormous market of pharmaceutical industry. However, the present time is fascinanting, because the current model no longer works. We need to see drugs in a completely different way. In particular we need to consider communities of organisms, and in particular microbes.

Now we are left with single cell organisms. To progress conceptually (and experimentally) we need also to make them simpler, and forget about nuclei and organelles. This leaves the majority of bacteria. This is already a universe in itself, but there we can find rules, and, of course, all the "nanomachines" that permit the cell to reproduce (ie make a *similar* copy of itself), while its genome
replicates (ie makes an *identical* copy of itself). Btw, even replication, in this case, needs to be qualified, but at least for the beginning, one may do asif it was real replication (without errors). Finally, a central tenet is the existence of a coding process, which permits separation between the world of actions and
the world of memory.

And life has to be seen as composed of two different moments: survival of the individual cell, and reproduction of that cell. Survival is quite limited in time, while reproduction (also limited, on the long term, of course) can be considered unlimited for our purpose. The purpose of reproduction is to "rejuvenate" the parent
cells which can only keep ageing. The most common way is that aged structures are put on one side of the cell, so that the ageing side will act as a garbage bin, and die out at some point (after ca 80 divisions).

Now, the role of Maxwell's demons genes we discussed in several of our recent exchanges, is to measure what is young, using energy to prevent degradation, or accumulation at the ageing pole... In this first model, there is no real creation of information, but simply use of the memory (the genome) to recover information. Note that in any case one needs to make room, ie destroy some molecules to replace
them by new ones. The place where *measurement* comes in is in the *choice* of the things which should *not* be destroyed.

Reading Landauer and Bennett (and recently Zurek), I think that the idea of making room is there, but the idea that energy is used to *preserve* information is novel. This is why it will take time to be acknowledgded as a significant element of our reflection on information. This is something also discussed, in a completely
different context, by John Hopfield in 1974 (kinetic proofreading).
The importance of this conjecture is that it is amenable to experiments! And this is of course absolutely essential. If nothing changed, the system, with its fixed environment, would be somehow immortal.

Now, with this view, time is only a parameter, monitoring the reproduction time course. I think that we need to introduce, at the end, time as a dimension. This is when information has to be *created*. And this creation couples the environment with the genome (which needs, therefore, to vary: hence it also fails to replicate, but reproduces, however at a scale of time totally different, much slower, than that considered above). The way it does that is more or less described in my "toy model". The idea is that, when the environment changes, it makes it possible for the genome to "memorize" these changes, thus creating novel information (i.e. a
coupling between the evolving organism and changing environment).
Remarkably, to this aim, we do not need further Maxwell's demons than those described above, as used to tell whether an entity is young or old.

If I summarized here the discussion we have had for some time it is because I think that Synthetic (Symplectic) Biology is exactly what we need to explore these ideas further. And this we need to do not only by reconstructing what we know, but by constructing novel cells, with novel metabolic pathways. Yet, in the absence of any economic incentive nobody would be interested (see the article mentioned above about our present policy for publication of scientific articles). Fortunately, there is a gigantic underlying economic interest: the need to invent novel energy sources. And to this aim, implementation of a novel chemistry, with novel metabolic pathways in bacteria or yeasts will be essential. Yet, the amount needed is so large that we cannot see how these synthetic organisms would be able to multiply at the required scale if we do not take into account the problem of their ageing, and... the way we can enslave their Maxwell's demon's genes to our human targets: producing an energy source.

I will try, in the next series of discussions to explore this situation with concrete examples. We hope that Symplectic Biology
(beta-test at http://www.symplecticbiology.org) will be the place where the peer-reviewed corresponding work will appear.

Antoine Danchin"

Aucun commentaire:

 
Site Meter