> As for the instability of my system, I guess I was a bit inaccurate.
> So, to clear this up: the complex contains several subunits. According
> to the literature these subunits should stick together in a certain
> way. However, we observe that some domains of the subunits drift apart
> during the simulation. On the other hand, other domains of the
...ah, this brings up another little subtlety of the restraints...
> 2) Heating with NPT to 300K
> 3) Equilibration with NPT
If you are seeing the movement of the subunits after these steps, this
likely relates to an issue with the pressure scaling and restraints.
Effectively, as the box is scaled (to reach the target pressure), the
restraint coordinates are shifted as well. By default, this shifting is
by molecule, and this will move the relative positions of the subunits
(assuming they are in separate molecules).
This has been discussed on the reflector repeatedly, but it is worth
mentioning again. What I do is to modify the prmtop (i.e. create a copy
and edit it) to put all of the subunit molecules into a single molecule
and then use this for the equilibration. For production I shift back to
the original prmtop.
Looking at a prmtop and its molecule information, we see something like:
%FLAG SOLVENT_POINTERS
%FORMAT(3I8)
177 8526 20
%FLAG ATOMS_PER_MOLECULE
%FORMAT(10I8)
2491 74 51 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 3
3 3 3 3 3 3 3 3 3 3
...
3 3 3 3 3 3
%FLAG BOX_DIMENSIONS
Under the "FLAG ATOMS_PER_MOLECULE" are how many atoms are in each
molecule. In the above case, I have protein (the first molecule), and two
ligands, followed by ions (1 atom) and then waters (3 atoms).
To put the ligands into the protein, I would increase the size of the
first molecule by 74 + 51, then shift the "1" entries, shift the "3"
entries... Additionally, we need to reduce the number of molecules and
first molecule that is solvent (above). This looks like:
%FLAG SOLVENT_POINTERS
%FORMAT(3I8)
177 8524 18
%FLAG ATOMS_PER_MOLECULE
%FORMAT(10I8)
2616 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 3 3 3
3 3 3 3 3 3 3 3 3 3
...
3 3 3 3
%FLAG BOX_DIMENSIONS
> 4) Production with NVT and ntt=3
>
> Would it be wiser to use ntt=1 instead? I believe I read in the
> archive that ntt=1 would make more sense than ntt=3. However, the
> amber manual writes: "Unless you are sure you know what you are doing,
> please don't use ntt=1"..
This gets into religious wars. My feeling is that NTT=1 is fine if the
coupling constant is long (TAUTP=5.0 or 10.0); similarly, Langevin is fine
as long as the coupling is also weak (i.e. you are not putting in too much
random force).
--tec3
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Dec 23 2007 - 06:07:35 PST
