Re: AMBER: questions about smd of amber 9

From: Gustavo Seabra <>
Date: Thu, 29 Nov 2007 09:53:24 -0500

On Nov 28, 2007 4:29 PM, WANG,YING <> wrote:
> Hi, Dear all,
> I have some questions about the smd method of amber, list as
> below: Thanks very much in advance.
> 1. In amber, the SMD with Jarzynski relation can do the pulling
> simulation, but this method is to put a spring inside the pulling
> molecule so pull the two ends apart. If I want to fix one end and
> pull another end. How should I do in amber? frozen one c-alpha
> atom? The ibelly parameter? Could I use it with jar together?

You can probably restrain one end, but I'm not sure what advantage
that will give you. If you want, just for visualization purposes, you
can align all frames keeping the position of this one atom fixed, and
the effect will be the same as performing the simulation with this
atom fixed.

> 2. From the mannual of amber, "One would need to repeat this
> calculation many times, starting from different snapshots from an
> equilibrium trajectory constrained at the initial distance value.
> This could be done with a long MD or a REMD simulation, and
> postprocessing with ptraj to extract snapshots." er, what's mean
> about "an equilibrium
> trajectory"?perform an energy minimisation first and then perform
> md on the stable structure? and then extract some snapshots? How
> can I know which snapshots have the same initial distance? should
> I fix the distance between these two ends and then perform an
> equlibrium simulation and then extract some snapshots?

The point is that the Jarzynski relation assumes a large number of
realizations of the process, all starting from an equilibrium
ensemble. So, the best way is to equilibrate your system, then take
snapshots from it to use as starting configurations. One good
procedure is:

1. Equilibrate you system free of restraints, using whatever method
you like. That will give you the initial equilibrium distance for the
coordinate you are interested.

2. Restrain that coordinate to a value slightly smaller than the
equilibrium value, and equilibrate again. Then, run some MD
trajectory, still with the restraint, and extract the initial
snapshots from this trajectory. You want your initial ensemble to be
at some restrained coordinate because then you can synchronize all the
pullings (starting and ending at the same values, and having all the
same intermediate values). The reason to choose a slightly smaller
value than the equilibrium value is just aesthetic: when you plot the
work or free energies, it will give you a clear minimum in the correct

> 3. As my understanding, after I get these snapshots, that is, they
> have the same distance between the two ends but with different
> atomic positions, then I perform the same smd simulation on them.
> Is it my understanding right?

The snapshots should all have approximately the same distance. It is
easy when you are close to the minimum, because a small restraint is
enough to fix this distance, but can get a bit tricky away from this
point. But notice: all you need to be the same is this distance.
Absolute atomic positions are irrelevant here, as are the positions of
all other atoms, which will be controlled by the MD simulation as
usual. The only thing you want to control is the distance you are
pulling. Then, yes, you should just run exactly the same SMD
simulation (pulling) with each snapshot, meaning you will use exactly
the same parameters for each one, each starting from a different

> Yes, we can get energy average from
> some repeating simulation like that, but how about the
> force-extension relation? Different start structure should have
> different relation.......... Could I just pick up one of them? I
> think it would not be fine.

I'm not sure what you mean by the "force extension relation". The
equations are always the same. The work done to separate the atoms
will depend on the particular trajectory that each particular pulling
is following, but you actually *want* that, since that's what will
make them different and give you an ensemble to average in the end
according to the Jarzynski relation. There's nothing to 'pick'.

> 4. And in the smd mehtod, I know that, first, let the natural
> length of the artificial spring become longer, and then, how can
> amber to figure out the position of the ends of the molecule? that
> is, how amber give the x to me? In reality, we can wait a moment
> to let the molecule acchieve a balance position. but in amber, how
> can the software to do that? also let the atoms move and then get
> the balance position? but it doesn't look like that when I do it

I'm sorry, but what you want here is not very clear.
Do you just want to get the positions of the atoms?
--> That's in the trajectory file.

Or do you want the (real) restrained distance at every step?
--> That's in the 'jar.log' file, second column.

Or are you asking how Amber calculates this distance?
--> Amber has the coordinates of each, so it's simple to calculate the
distance between two points.

What do you mean by "balance position"? When you put a restraint, what
actually happens is that you are adding to the force field an extra
term that depends *only* on that distance, then you use a large force
constant to make sure that this term will dominate over other terms
that involve the same atoms, i.e., this term will define the relative
distance between those 2 atoms. Of course, all other terms that
involve any of those atoms will also contribute, and that's what gives
you some oscillation - the distance will *not* be exactly the same all
along. All other atoms in the system move normally, and naturally
adapts to the fact that a distance is restrained (it's just an extra
term in the force field).

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Sun Dec 02 2007 - 06:07:20 PST
Custom Search