Re: vmd-l: Restraints on coarse grained model

From: Anton Arkhipov (
Date: Wed Dec 09 2009 - 10:58:56 CST

Hi Francesco,

> Now (never carried out a simulation with namd) I plan to proceed
> as I used to do for all-atoms with Amber:
The procedure you described should work fine. In fact, I don't think
all the steps are necessary, even for an all-atom simulation. A
minimization followed by a simulation right away at 310 K (or whatever
temperature you need) should be fine. Maybe use some restraints for
the first few nanoseconds of the simulation, and then release them.

> Has any adaptation to CG already been posted beyond setting a much
> larger time step? If I understand the sample configuration file for
> minimization-equilibration, why 300K and constant pressure? I expected
> minimization at constant volume and 0K, then gradually bringing the
> system to room temp.
Just like I said, you can be extra careful if you want, but usually
it's not necessary. And yes, there's not much difference between
running an rbcg simulation and an all-atom simulation other than the
larger timestep. The rbcg model also uses cosine-based terms for angle
potential energy, but that's specified in the parameter file.
Sometimes in an rbcg simulation we may use larger cutoffs and change
some other parameters a bit, but that's not essential. Just follow the
settings in your sample rbcg configuration file for NAMD, that should
be good enough.

> "Has any adaptation [of membrane tutorial] to CG
> already been posted beyond setting a much larger time step?"
As far as I know, no. But I hope the previous paragraph answers your

> Do restraints on CG models apply like for all-atoms?
Yes, that works in exactly the same way.

> set fixed [atomselect top "protein or (chain O and name CHO PHO)"]
> It worked for the lipid, not for the protein, whose beta column
> (tempFactor in PDB language) 0.00 was unchanged.
This is because VMD doesn't recognize an rbcg protein as protein. The
reason is that atom names are different from what VMD is used to - no
"name CA", for example. Change "protein" in your atomselection to,
say, "name BB", and it will work fine.

> My aim is to arrive at multimicroseconds in the hope to substantiate
> (or disprove) mechanistic guesses about how the protein works.
Unfortunately, this might be difficult to achieve. We warn everyone
that rbcg model is not quite suitable for studying proteins with well-
defined tertiary structure, because the model doesn't reproduce fine
amino acid interactions so well. So if you start from a protein with a
well-defined tertiary structure, it will quickly collapse or unfold in
your rbcg simulation. So rbcg is more suitable to studying small
peptides or proteins with little of tertiary structure interacting
with membranes, such as in the case of nanodisc. Otherwise, one has to
apply extra springs to the protein structure, connecting amino acids
that are in reality interact only through non-bonded interactions, but
that will bias your simulation, so that you get out only what you put
in in terms of the protein mechanics.

See these two papers as examples:

Four-scale description of membrane sculpting by BAR domains. Anton
Arkhipov, Ying Yin, and Klaus Schulten. Biophysical Journal,
95:2806-2821, 2008.

Assembly of lipids and proteins into lipoprotein particles. Amy Y.
Shih, Anton Arkhipov, Peter L. Freddolino, Stephen G. Sligar, and
Klaus Schulten. Journal of Physical Chemistry B, 111:11095-11104, 2007.



On Dec 9, 2009, at 1:07 AM, Francesco Pietra wrote:

> I went on with the VMD TkConsole commands
> $all set beta 0
> set fixed [atomselect top "protein or (chain O and name CHO PHO)"]
> $fixed set beta 1
> It worked for the lipid, not for the protein, whose beta column
> (tempFactor in PDB language) 0.00 was unchanged.
> I found no way out from manuals. Do cg proteins behave differently
> from all-atoms proteins to these commands or I just made some mistake?
> Should a Python script be prepared to modify the beta column for the
> protein?
> thanks
> francesco
> ---------- Forwarded message ----------
> From: Francesco Pietra <>
> Date: Tue, Dec 8, 2009 at 8:23 PM
> Subject: vmd-l: Restraints on coarse grained model
> To: vmd <>
> Hi:
> Do restraints on CG models apply like for all-atoms? For example, in
> order to restrain the polar head of POPC lipid
> ATOM 2463 CHO POPCO 1 37.490 25.415 4.653 1.00
> 0.00 O1 C
> ATOM 2464 PHO POPCO 1 39.740 24.614 0.940 1.00
> 0.00 O1 P
> ATOM 2465 ES1 POPCO 1 37.568 20.460 -0.263 1.00
> 0.00 O1
> ATOM 2466 ES2 POPCO 1 37.020 21.866 -4.322 1.00
> 0.00 O1
> ATOM 2467 ME1 POPCO 1 36.057 16.858 1.933 1.00
> 0.00 O1
> ATOM 2468 ME2 POPCO 1 35.653 13.060 0.274 1.00
> 0.00 O1
> ATOM 2469 ME3 POPCO 1 37.948 9.854 -0.816 1.00
> 0.00 O1
> ATOM 2470 MT1 POPCO 1 38.047 4.914 0.293 1.00
> 0.00 O1
> ATOM 2471 ME4 POPCO 1 35.557 18.929 -6.037 1.00
> 0.00 O1
> ATOM 2472 ME5 POPCO 1 36.308 15.668 -9.314 1.00
> 0.00 O1
> ATOM 2473 ME6 POPCO 1 37.484 11.986 -12.138 1.00
> 0.00 O1
> ATOM 2474 MT2 POPCO 1 37.279 8.543 -12.520 1.00
> 0.00 O1
> are restraints to be applied to beads CHO and PHO? If so, by making
> the .fix file in VMD as if it were an all-atoms model are the
> restraint forces applied correctly?
> And if I want to restraint the whole protein and the whole bilayer (to
> work at constant pressure on the solvent water around), is any way to
> simply specify the range of residue numbers?
> That is, I assume to have to prepare the system for production as
> carefully as if it were all-atoms.
> thanks
> francesco pietra

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:53:34 CST