[Molecularmechanics] first steps

[Pengyu Ren]

======= 2003-10-21 14:51:00 original messages===

>On Tuesday 21 October 2003 14:25, Peter Murray-Rust wrote:
>
>> >The Amber force field, for example, requires as input an atom type and a
>> >partical charge per atom
>>
>> Where do these come from? They could be:
>> - assigned from calculation (in which case no problem - they are carried
>> over) - assigned from local atom or fragment dictionaries ("an amide N has
>> a charge of ...")
>> - guessed by the human driver
>
>They come from fragment dictionaries. Of course the fragments are not exactly 
>the same for each force field.
>
For proteins and peptides, the charge is already defined for each atom in each residue. AMBER, CHARMM and TINKER and other "bio" package can handle protein without problem, only structure/topology info is needed. All the bond, angle, dihedral... are assigned according to the connectivity by the program. TINKER can assign various charmm and amber parameters to a given protein because we have the full residue library.

For any other molecules that are not covered by the amber/charmm/tinker force fields (a libarary of parameters), you have to come up with charges by yourself. It's important to do it in a way that is consistent with the particular force field you are using. Mixing force fields parameters from different force field is dangerous in general.
There are programs/algorithm that can make guesses of charges based on bonding enviroment, uisng bond increment scheme. Most commercial software has this capability, even though personally I donot suggest such usage unless all you are interested in quick and crude estimations.

In conclusion, the way of assigning parameters is force field specific. The minimum requirement for input should be coordinates and connectivity. If it is desired to include force field parameters and function forms in the input ( i imagine a separate file from the coordinates), we should leave room for future addition of new functions.



>> In the first case CMLComp can hold a fractional charge. In the second it
>> can hold a dictionary of fragments with charges
>
>I think the difficulty is in describing the relation between the fragment 
>dictionaries and the molecules.
>
>To make life even more complicated, many force fields use a united-atom 
>representation in which not all hydrogens are modelled explicitly. A force 
>field description must therefore also contain a list of conditions for 
>applicability to a given model.
>
>> The crystallographers use "dictionaries" for their force field refinement
>> and these seem to require a certain amount of human effort and anguish.
>> There are standard libraries of fragments which include most of the common
>> ligands or prosthetic groups in PDB. I assume the same holds here. So I
>
>Yes. Even the force fields are essentially the same.
>
>> suspect that CMLComp can hold such fragments and their associated
>> parameters if required. Actually calculating them for a given problem may
>> well be tricky, XML is not directly involved in that.
>
>No, but a unified representation loses much of its value when it still 
>requires an adaption of the code to each individual force field. Therefore I 
>am not very enthusiastic about an effort to define force field descriptions. 
>Time is better spent on defining structure and trajectory representations, 
>which allow real exchange of data immediately.
>
>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
>-------------------------------------------------------------------------------
>
>_______________________________________________
>Molecularmechanics mailing list
>Molecularmechanics@tddft.org
>http://www.tddft.org/mailman/listinfo/molecularmechanics

= = = = = = = = = = = = = = = = = = = =