Hello,
On Thu, Jan 28, 2010 at 12:14 AM, Ed Pate <pate.math.wsu.edu> wrote:
> Dear Amber Community:
>
> I am studying a protein with an exposed surface loop. I am interested in
> whether the loop could deform and interact with another part of the protein.
> As a "quick and dirty" method to investigate the possibility, I have tried
> to do an energy minimization in a GB simulation with a distance restraint
> involving an increasing force constant as a function of minimization step
> number to bring the loop and potential interaction site close together. The
> approach is not working. I have increased the force constant by orders of
> magnitude and the loop never appears to be moving toward the potential
> interaction site.
If I understand what you're doing exactly (and I'm guessing this based
also on your restraint file shown below), is you're restraining the
loop to a rather small distance, and slowly increasing the force
constant in an attempt to drive the loop closer. Moreover, it seems
as though you're adding the biassing potential to two atoms that are
25 angstroms apart, attempting to force them to be somewhere between
2.5 and 3.5 angstroms in a *single* step (not one minimization step,
but a single minimization run).
I'm not surprised that this is not moving the atoms, since the barrier
to inducing such a rapid, major conformational change is probably well
greater than value of the restraint potential you're adding. That is
to say, even if you did get something to happen with this approach,
I'm skeptical that it would reveal anything useful. I think a more
effective approach would be to move the position of the distance
restraint steadily from 25 angstroms (or whatever the starting
position is) to your desired position, and minimize along the way.
(i.e. minimize at 1-angstrom intervals from 25 to 2 with a restraint
potential large enough to ensure it moves). The slower and more
carefully you induce a change, the more it will tell you about the
equilibrium/thermodynamic properties of a system (since reversible
processes, which take an infinite amount of time to occur, are
described by equilibrium thermodynamics, whereas irreversible
processes do not, per se, though for an example of an exception, see
Jarzynski's equation).
That being said, if you are looking for free energies of the change,
you would need to use some kind of molecular dynamics methods, like
steered MD or umbrella sampling that can will generate a PMF along
some well-defined coordinate.
>
> I have several questions.
>
> 1. Is this approach even feasible in Amber8? What am I missing?
>
> 2. If it is, could someone help me understand what I am doing wrong?
>
> I am using Amber8. The min.in file follows:
>
> minimize structure
> &cntrl
> imin=1,
> ntmin=1,
> ncyc=200,
all steps are steepest descent, which is a slowly converging method
> maxcyc=200,
I can't imagine too much can happen with so few minimization steps
> cut=300.0,
> igb=2,
> saltcon=0.2,
> gbsa=1,
> ntpr=10,
> ntx=1,
> ntb=0,
> ntr=1,
> nmropt=1,
> lastist=30000000,
> lastrst=30000000,
> &end
> &wt
> type='END',
> &end
> LISTOUT=/home/pate/david_eg5.d/listout
> DISANG=/home/pate/david_eg5.d/RST
> restrain the following
> 5.0
> RES 1 117
> RES 133 500
> END
> END
>
> The loop in question is residues 118-132. The distance restraint on the
> rest of the protein was an effort to try and keep the protein from being
> pulled apart during the restraint minimization. Elimination of this group
> definition does not affect the end result.
I'm not sure exactly what you're trying to learn here... It appears as
though you've already decided what change is going to take place, and
you're simply applying restraints to make it happen. At the end of
the day, you'll get an energy profile that will almost certainly not
be followed, and by not allowing the structure to relax you won't get
a very good idea of the barrier height either.
>
> The RST file is
>
> &rst
> ixpk=0, nxpk =0, iat(1)=1971, iat(2)=5834, iat(3)=0, iat(4)=1,
> r1=1.5, r2=2.5, r3=3.5, r4=45., r1a=1.5, r2a=2.5, r3a=3.5, r4a=45.,
This describes a flat-well potential, that is linear at distances
smaller than r1 and greater than r4, parabolic according to your force
constants between (r1 and r2) and (r3 and r4) and zero between r2 and
r3.
> rk2=0.1, rk3=0.1, rk2a=10000000., rk3a=10000000.,
> ir6=0, ialtd=0, ifvari=1, ninc=0, iresid=0, imult=0,
> nstep1=0, nstep2=201,
> &end
>
> Atom 1971 is in the loop. Atom 5834 is in a ligand bound to the protein. It
> is in residue 387. The .top and .crd files for the protein were generated
> using the outline in the tutorial at
> http://ambermd.org/tutorial/gb/gb_tutorial.html. tleap reported no
> problems. Atoms/residues were identified using ptraj to generate a pdb
> file. Atoms 1971 and 5834 were initially approx. 25A apart. After the
> iteration, they remained approx. the same distance apart. Smaller, more
> reasonable values of rk2a and rk3a yield the same result.
>
> If someone could help me understand whether the approach is feasible, and if
> so, where I am going wrong, I would appreciate it.
I would probably perform multiple minimizations at many distances
along the path from 25 to 2.5, and increase your maxcyc considerably
for each step (especially if it's a large protein). This *might* give
some qualitative description to what you're trying to observe, and I
think it has a better chance at inducing the change you're attempting
to induce. Others may, of course, add their views/suggestions as well
or point out something I may have missed, but this is my 2 cents.
Good luck!
Jason
--
---------------------------------------
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jan 27 2010 - 22:00:02 PST
