From: "Yong Duan"
Subject: RE: AMBER: Simulation of small protein
Date: Thu, 23 Oct 2003 21:11:02 -0400


Dear Robert,

Carlos said it mostly well (except suggestions that I somehow know this
better).

Villin is a relatively challenging "protein" of its size. Helix-one does
not have strong contacts with the rest and is actually perpendicular to
the other two helices. Although I have not tried GB or GB/SA on villin,
I expect this helix to fluctuate, although one should be able to
maintain its helicity reasonably well.

I'd encourage you to look into the cause of the large RMSD. If this was
caused by the movement of helix-one, it probably implies the dynamic
nature of the helix. In which case, although this large RMSD raises a
bit concern, I would not think this constitutes a major problem. If the
overall structure is ok, continuation of the simulation may be
justified.

Other than helix-one, there is a phe residue at the C-terminal which has
the tendency to stick with hydrophobic surface. But for some reason, it
stays in solvent in the NMR structure. There could be several
explainations. An obvious explaination is that it is disordered and
actually assumes several conformations with similar stability. So, this
would explain the NMR data. Another explaination is that the force field
over-estimates the hydrophobic force. If so, correction of this problem
is pretty difficult to develop and a challenge to the community.
However, given the consistent observation in X-ray structures that phe
residue does not like to stay in solvent, which is consistent with a
good number of simulations, I am uncertain how much faith we can put on
such an explaination.

Other than these two possible problems, if you indeed observe
considerable conformational changes on other parts of the peptide, we
would like to know.

Also keep in mind, one typically observes more stable structures in
simulations when they are started from X-ray structures. For some
reason, NMR structures tend to be somewhat less stable which is
counter-intuitive. So, it may be a good idea to be cautious when you
start the simulations. For example, one may consider a relatively long
equilibration process at low temperature. My favorite choose is 1.0ns at
100K and then raise the temperature a bit slowly. In most cases, I'd say
such an elaborate equilibration is unecessary when the simulations are
started from X-ray structures and are in solvated environment. But for
GB or GB/SA simulations, this could be justified. Besides, it never
hurts to be careful.

Hope this helps!

Good luck and let us know if you need further help.

yong
***********

> -----Original Message-----
> From: owner-amber@scripps.edu
> [mailto:owner-amber@scripps.edu] On Behalf Of Carlos Simmerling
> Sent: Thursday, October 23, 2003 7:00 PM
> To: amber@scripps.edu
> Subject: Re: AMBER: Simulation of small protein
>
>
> Hi Robert,
> I guessed it was villin since there aren't so many 36 residue
> proteins.
> It is fairly compact so I am not sure why your RMSD goes
> up so fast, is it well equilibrated? equilibration in GB can be
> hard since there isn't anything to damp out initial fluctuations.
> I'm not too familiar with villin though, maybe Yong Duan can comment?
> The SA term may do the same, just damp out any initial poor contacts
> etc.
>
> As to the best force field to use, I am not sure on that
> either. Parm99
> and even 94 (though less so) seem to be significantly
> overstabilizing alpha
> conformations bu since villin is helical maybe that's
> actually better for
> getting it to fold-again maybe Yong can comment on that. For real
> dynamics I would expect it to be too helical, although helix
> extension may
> be something to look for in your RMSD increase.
>
> Try frequency filtering in mddisplay, it can help a lot with
> reducing the
> wiggles.
> Carlos
>
> ----- Original Message -----
> From: "Endres, Robert G."
> To: <amber@scripps.edu>
> Sent: Thursday, October 23, 2003 1:27 PM
> Subject: RE: AMBER: Simulation of small protein
>
>
> >
> > Dear Carlos,
> > its the villin headpiece which was described as a thermostable, fast
> folding
> > alpha-helical subdomain. I also looked at the trajectory
> with mddisplay:
> at the beginning of the movie the structure seems a bit more
> open, but its
> hard to tell since its just a bunch of wobbling atoms.
> >
> > What's a better force field - parm94, or the one you published
> > in JACS on Trpcage folding? By the way, in this paper did
> you use GB or
> GBSA?
> > When I just use GB (without SA-term) it actually partially
> unfolds at the
> beginning of trajectory (starts to open up but closes then
> slowly again).
> >
> > Robert
> >
> > -----Original Message-----
> > From: owner-amber@scripps.edu on behalf of Carlos Simmerling
> > Sent: Thu 10/23/2003 12:13 PM
> > To: amber@scripps.edu
> > Cc:
> > Subject: Re: AMBER: Simulation of small protein
> > it depends on whether you're seeing backbone changes-
> > we've publsihed that there is significant overstabilization of alpha
> > helix in parm99+GB. on a long time scale, all of
> > our sequences increased in helical content and the
> > energy dropped. Seeing this during 200ps is a bit surprising
> > though.
> >
> > without knowing what the topology looks like it is
> > hard to say whether more compact is reasonable (is it
> > a rather open structure to start with?)
> > Carlos
> >
> > ----- Original Message -----
> > From: "Endres, Robert G."
> > To: <amber@scripps.edu>
> > Sent: Thursday, October 23, 2003 11:52 AM
> > Subject: AMBER: Simulation of small protein
> >
> >
> > >
> > > Dear AMBER users,
> > >
> > > After heating/equilibration of a NMR structure of a small
> (36 res.)
> > protein with weak constraints, I was doing a short (0.2 ns)
> MD simulation
> > (parm99.dat) at 300K with the GBSA implicit solvent model.
> > > I was a bit surprised that the total energy dropped
> further by almost
> 100
> > kcal/mol compared to the starting structure (close to NMR
> structure), and
> > the ESURF term (proportional to the total surface)
> decreased from 17 to 12
> > kcal/mol.
> > > So the protein got more compact during the MD simulation. The RMSD
> started
> > from zero and increased to 3-5 A.
> > > It seems that the protein changed quite strongly the
> structure during
> the
> > simulation. Does anyone have experience with this, i.e. is
> this "normal"
> or
> > "expected", is this a problem with the force field/solvent
> model or NMR
> > structure?
> > >
> > > Many thanks for suggestions,
> > >
> > > Robert