[Molecularmechanics] first steps

[Konrad Hinsen]

On Tuesday 21 October 2003 18:16, Peter Murray-Rust wrote:

> I completely agree. One of the biggest problems to interoperability is the
> implicit semantics in many programs. By externalising the semantics you get
> a much cleaner design as well as a better understanding of what the program
> is actually doing. I'm fairly confident that for most systems the XML can
> abstract the force field. As a general rule, when something proves
> difficult in XML it tells you there is a fundamental design feature that
> must be addressed.

I am not sure that this general rule applies here. I think it is important to 
realize that force fields are conceptually algorithms. In the framework of 
functional programming, a forcefield would be a function mapping molecular 
structures to a function that calculates the energy for a given conformation 
(the "energy subroutine" in a typical simulation code). A fully general 
description of forcefields (which would allow the implementation of something 
like DFT, which is in principle a force field as well) therefore requires a 
Turing-complete language. While such a language could of course be coded in 
XML, I think that any "real" programming language would be a more natural 
choice.

We can of course restrict ourselves to "standard" molecular mechanics force 
fields. Then we need to agree on a common force field algorithm that is 
flexible enough to cover all force fields under consideration, and convert 
the currently available force field parameters to an equivalent parameter set 
for the new unified algorithm. That parameter set could then be represented 
in XML.

Such a project would be quite ambitious because it requires significant coding 
efforts in all programs concerned in addition to the common representation 
development. At the same time, it risks being of limited use because whatever 
common algorithm we may come up with risks being insufficient for the next 
generation of force fields. 

Konrad.
-- 
-------------------------------------------------------------------------------
Konrad Hinsen                            | E-Mail: hinsen@cnrs-orleans.fr
Centre de Biophysique Moleculaire (CNRS) | Tel.: +33-2.38.25.56.24
Rue Charles Sadron                       | Fax:  +33-2.38.63.15.17
45071 Orleans Cedex 2                    | Deutsch/Esperanto/English/
France                                   | Nederlands/Francais
-------------------------------------------------------------------------------