Le 14 nov. 09 à 19:15, Kjeldgaard Morten a écrit :
On 14/11/2009, at 18.55, Ronald E Stenkamp wrote:The rumblings here at the Univ. of Washington among the computational modelers is that some of their current models might be more representative of protein structures in solution than are the crystal structure models. It may take less than a "couple of decades" for a reduced emphasis on crystallographic studies.Molecular models are the result of numbers emerging from computer programs. The results of such computations do not reflect anything in nature. There's no experimental evidence whatsoever, making modelling a very theoretical -- in my eyes uninteresting -- exercise.For what it's worth, protein molecules in crystal structures, with typically > 50% solvent, are already "in solution" to the extent that protein molecules are ever "in solution" in their natural environment.
Hi,I think that strong statements and future foretelling are probably not very useful to a discussion that is indeed interesting and that will be put forward more and more often. Crystal structures are actually models themselves. Sure, they are directly backed by experimental data against which they can be compared in many ways, what we call validation. But as models, they have their own limitations and, ultimately, to be useful and trustworthy, they should be backed by results from other experimentalists: biochemists, biophysicists, etc. Because computational models don't have direct experimental data behind them, they need an even stronger external validation. But this does not imply that they are useless. Other fields of science progress thanks to computational models that help understanding problems and pose new, experimentally-testable questions. Biology needs computational models as well: about structures and their dynamics, biochemical pathways, cell and tissue interplay, etc.
The question of whether theoretical models reflect better the "in solution" state is probably a precipitate one (sorry). They should first prove that they consistently lead to predictions that are backed by data from experiments "in solution", "in vitro" or, better, "in vivo". Experimental structural models have proven very useful in predicting properties of the macromolecules that they represent. One day, computational models may be considered similarly useful.
But I don't think computational modelling will replace experimental structure determination. In a way it is not completely true that there is no experimental evidence behind computational structural models: these models are not made from just first principles, they incorporate a lot of empirical knowledge. Thus, modellers need experimental models to be improve their own ones and that will probably hold true for long. It can be argued that opposite is also true and we, structuralist, have something to learn from computational models. I do think so.
In the meantime, I think it is in our best interest to keep a dialogue between "computational" and "experimental" modellers and avoid the 'we- don't-need-you' wars.
Best, -- Miguel Architecture et Fonction des Macromolécules Biologiques (UMR6098) CNRS, Universités d'Aix-Marseille I & II Case 932, 163 Avenue de Luminy, 13288 Marseille cedex 9, France Tel: +33(0) 491 82 55 93 Fax: +33(0) 491 26 67 20 e-mail: miguel.ortiz-lombar...@afmb.univ-mrs.fr Web: http://www.pangea.org/mol/spip.php?rubrique2
smime.p7s
Description: S/MIME cryptographic signature
