[gmx-users] how to calculate internal energy for energygroups?

[Marc Baaden]

>>> David van der Spoel said:
 >> Maybe we should go back one step. The reason this is not implemented is 
 >> because we thought it would be unnecessary. In addition, if you would do 
 >> it per residue how would you categorize e.g. the dihedral terms with one 
 >> atom in one residue and three atoms in another? Of course you can define 
 >> something, but what does it mean?


Ok, first a general comment:

I have used this in the past and found that the sum of all bonded energy
terms for a given residue or for a subpart of a molecule is a good measure
for the internal strain of that entity. Of course one may want to normalize
this depending on what one wants to compare (per residue / per atom / per
domain ...).

I agree that the part of this entity that is connected to the rest of the
molecule is ambiguous. I usually disregard all bonded terms that are not
entirely contained within the selected atoms (the logic being that you can
always make your selection bigger if you want).

The problem that there might be with the way g_energy works at the moment
is that I think g_energy tries to split the energy in parts so that by
summing the parts you obtain the total sum again. That is you can't put
an atom in two selections at the same time.


For the specific case I want to use this now:

In the system we are simulating we observe a specific pattern of contacts vs
the residues of the macromolecule. We think that the origin for this lies in
the balance of the different force-field terms for each residue. So I would
like to make a plot of the interaction energies for each residue with the rest
of the system split into electrostatic, van der Waals and bond/angle/dihedral
contributions to see what the major contributions are and whether the pattern
varies from residue to residue.

I hope this does make sense and provides at least one reason to implement
such a procedure.

Marc
-- 
 Dr. Marc Baaden  - Institut de Biologie Physico-Chimique, Paris
 mailto:baaden at smplinux.de      -      http://www.baaden.ibpc.fr
 FAX: +33 15841 5026  -  Tel: +33 15841 5176  ou  +33 609 843217