AMBER Archive (2007)

Subject: Re: AMBER: Equilibration of protein complex in POPC membrane, the whole TIP3 solvated

From: M. L. Dodson (activesitedynamics_at_comcast.net)
Date: Wed Nov 28 2007 - 11:31:04 CST

Francesco Pietra wrote:

> Referring to subject, I had no answer to my question if solvate
> box (added 21847 residues, density 0.792) is preferable to
> solvateoct (added 30286 residues, density 0.867). Therefore, I
> assume this means carrying out MD with solvatebox.
>

I do not know of any scientific reason to favor a box or a
truncated octahedron. So, in situations like this, I am guided by
the "shape" of my system. Truncated octahedra are useful for
roughly spherical systems such as compact proteins. Your system
seems to best fit in a rectangular box. The initial density is of
secondary importance as you will need to do a constant pressure
run as part of your equilibration to prevent the "microvoids" from
growing into "vacuum bubbles" during production dynamics.

> My question now is how to carry out equilibration for this
> system. The pore protein entails a docked large ligand (118
> atoms) within the pore, coming from amber rescore in DOCK. It
> was immersed in a POPC membrane, where the polar heads of POPC
> are solvated TIP3P. Then the whole was TIP3PBOX solvated with
> 12.0 A buffer.
>
> Summing up all that I could learn from the literature,I guess
> that the system should be first energy minimized with
> protein-complex restrain (SHAKE on H atoms and PME). For
> membrane I found indications of ca. 30 kcal/molxA^2. No idea how
> that could be applied to my protein-complex.
>

Yes, the initial minimization is primarily to relieve VdW overlaps
created by whatever means you modeled your initial structure. 30
kcal/molxA^2 seems dramatically too big. I use 1-2 kcal/molxA^2,
although you can get away with larger values in the minimization
steps. This depends entirely on the quality of your initial
structure.

> Once that (or a more reasonable) pre-equilibration is carried
> out, I suppose to have to gradually "heat" the system to 300K at
> constant volume. Here surely the protein complex should be
> restrained as above. However, probably the POPC molecules should
> also be restrained (or only their polar head?).
>

Yes, I start out at 100K, then ramp to 300K over 500 MD steps.
You do not want to start out too high in temp, as you are dumping
in kinetic energy, leading to instantaneous velocity components
pointing in essentially random directions. If this energy is too
large (too high an initial temp), it may tear the system apart. I
use restraints of 1-2 kcal/molxA^2. Too large a restraint at this
step will lead to the dreaded "vlimit violations". I run this
component of the equilibration for 25ps (probably bordering on too
long).

I simulate DNA-protein complexes, and I restrain only the protein.
This works primarily because the DNA and protein are strongly
electrostatically coupled. You want to restrain any part of the
solute that is not strongly coupled to the other parts of the
solute. The purpose of the restraint is to preserve the
"macrostructure" while allowing the "microstructure" to relax.

> With this second step (or a more reasonable one) the system
> should be ready for pre-MD at 300K and 1 atm, initially with the
> above restrain on the protein complex (and probably the
> lipid). Gradually (after how many ps?) restraint on the protein
> (and lipid, if any was applied) should be removed.
>

No. You have to follow the temperature (constant volume)
equilibration with a constant pressure density equilibration.
This relaxes the microcavities created during the solvation step.
If you do not do this the microcavities will grow into "vacuum
bubbles". You do not want this to happen. The density at the end
of this should realistically reproduce an experimental density. I
run this part for 25ps as well.

> Now the system should be ready for production MD, under SHAKE
> for H atoms.
>

Here is the way I do it. It is admittedly conservative. After
the constant pressure run, I do a sequence of minimizations,
gradually eliminating the restraints. This does two things. It
eliminates a sudden release of the restraints, and it removes the
velocities. After relaxation with the constant volume and
constant pressure dynamics, I probably could get away with a
simple single step minimization because I use such shallow well
restraints. Things should be relaxed sufficiently well to allow
this. You want to eliminate the velocities so that you can ramp a
well relaxed structure up to your final temperature. So I do the
series of minimizations (probably not needed in my case, but
depending on the magnitude of the restraints), then follow with a
200ps constant volume run with ntt=3. This equilibrates the
temperature of a well relaxed model of the system and gets ready
for production dynamics.

I follow this with NVE free dynamics production for as long as
my hypothesis requires or until it gets too expensive.

Various people have their own ideas about how this should be done
in detail. My protocol is probably more conservative than
needed. But the objectives of the sequential steps should be
achieved in some fashion without distorting the structure too
much in the process.

Hope this helps,

Bud Dodson

> In view of the computational burden, I would appreciate any
> guidance.
>
> Thanks
> francesco pietra
> 

-- 
M. L. Dodson
Business Email: activesitedynamics-at-comcast-dot-net
Personal Email:	mldodson-at-comcast-dot-net
Phone:	eight_three_two-56_three-386_one
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu