[gmx-users] Normal Modes Analysis

[Bert de Groot]

Fabrizio Mancinelli wrote:
> Hi colleagues,
> I would like to have some delucidation, from your own experience,
> about performing an NMA.
> My system is an usual protein in water.
> After energy minimization (full precision, CG or L-BFGS), next step should be 
> Hessian calculation. The first question is: if I perform it on the whole 
> system (protein + SOL), isn't the matrix too big? 

it becomes quite big, yes. If it becomes "too" big depends on
the amount of RAM in yoru system. Including solvent in a NMA
is usually a good idea, but you may get away with just a shell of
water molecules. Remember that due to the minimisation criteria
(NMA only works in an absolute minimum), the use of PBC and
cut-offs is usually not recommended/possible, due to boundary
instabilities cutoff and/or box edge.

> On the other hand, this is 
> not what I want: I just need NMA on the backbone; but, by filtering off the 
> SOL coordinates prior to hessian calculation, don't I miscalculate the 
> hessian for my system?

yes, AFAIK, you need to pereform the Hessian coputation + diagonalisation
on the full system. There have been some recent discussions on performing
NMA on subsystems but I don't know the status on that. Maybe Berk can comment?

> Second question is, let's suppose I have correctly calculated and subseqeuntly 
> diagonalized the hessian. How can I get the frequency spectrum (in cm^{-1})?

the eigenvalues of the Hessian are the frequencies. Check the units though
(discussed before on this list).

Bert

________________________________________________
Bert de Groot, PhD

Max Planck Institute for Biophysical Chemistry
Computational biomolecular dynamics group
Am Fassberg 11
37077 Goettingen, Germany

tel: +49-551-2012308, fax: +49-551-2012302

email: bgroot at gwdg.de
http://www.mpibpc.mpg.de/groups/de_groot