[MMTK] Hessian Matrix

Konrad Hinsen research at khinsen.fastmail.net
Wed Jan 11 11:09:50 UTC 2012


santhu kumar writes:

 > I want to compute the Hessian matrix for given protein conformation using
 > force-field. Here are the steps I have done 
 > 
 > #create universe
 > universe = InfiniteUniverse(Amber94ForceField())
 > #add protein to Universe
 > universe.protein = Protein(pdbFile)
 > energyAndHess = universe.energyAndForceConstants()
 > hess = energyAndHess[1].array
 > numAtom = hess.shape[0]
 > #reshape hess
 > hess.shape = numAtom*3,numAtom*3

You can stop there, the array is already symmetric.

 > Am i correct in obtaining hessian this way?

Yes!

 > I see that it is possible that I am using a completely different set-up than
 > that is used in MMTK's universe. If I have to do that, what is the
 > configuration file I have to use for Gromacs? If not, is there anyway I can
 > check my hessian's validity?

I can't help with Gromacs, which I don't know very well. In general,
it is very difficult to compare results between different programs
when dealing with biomolecular simulations, because there are so many
parameters that enter into the calculation, some of which may not even
be documented.

For this particular comparison, there is little hope of getting
exactly the same numbers because every program has its own
specificities in calculating electrostatic interactions. You could try
setting all charges to zero in both programs and see if the comparison
looks better.

Another way to validate a Hessian is to compute it by numerical
differentiation of the energy gradients (forces). But this only tests
the code doing the computations, not the parameters that define the
interactions in your protein.

Konrad.
-- 
---------------------------------------------------------------------
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
http://dirac.cnrs-orleans.fr/~hinsen/
---------------------------------------------------------------------



More information about the mmtk mailing list