[MMTK] Using MMTK for protein structure prediction

Konrad Hinsen research at khinsen.fastmail.net
Sat Oct 9 17:52:37 UTC 2010

On 8 Oct 2010, at 21:53, Siamak Ravanbakhsh wrote:

> MMTK is the right toolbox. A simple model e.g. c-alpha, is good  
> enough but I need access to individual energy terms. I'm aware of  
> the energyTerms() method of Universe class here but I need more  
> detailed version including the value for each term in summations  
> (e.g. corresponding to every bond).
> Can I play with source code to get this, or is it easier/possible  
> for me to calculate the terms for a given configuration myself?

There is no direct support for that, but two indirect ways that may  
well be sufficient for you:

1) Evaluate the energy for a subset of two atoms. The result is just  
the energy for that pair. Note however that this is expensive if you  
do it for all pairs.

2) Get the energy evaluator paremeters for your universe and calculate  
the energy yourself. For bond terms, this is rather straightforward.

> Another question is:
> How can I produce different configurations by sampling a bounded  
> range of possible values for all phi, psi and bond lengths (as they  
> are my optimization variables and I need to calculate the energy  
> corresponding to a set of samples coming from a distribution over  
> these variables)?

You can directly set phi and psi angles for each residue, so that part  
is handled easily. For varying the bond lengths, you have to do  
something more elaborate. One way is to define a collection of all  
atoms on the same side of the bond and translate it.

Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: research at  khinsen dot fastmail dot net

More information about the mmtk mailing list