Re: [AMBER] Quasi-harmonic Approximation Using a Trajectory Stripped of Hydrogens

From: Laura Tociu <ltociu.princeton.edu>
Date: Wed, 29 Apr 2015 00:44:57 -0400

Dear Jason,

I am using the multiple trajectory approach. I did a successful trial run
today with thee trajectories: one of arginine (all atoms), one of the
protein (just alpha carbons) and one of the complex (alpha carbons of
protein + all atoms of arginine). Is there a reason why this should not be
ok? You were proposing two slightly different approaches. I am assuming the
quasi harmonic approximation is quite bad for a small molecule such as
arginine, but isn't it better than nothing (assuming it will underestimate
the real entropy, which I think it does) ?

Another thing that I've been meaning to ask for a long time but is a little
unrelated is how important is it to do MMPBSA on an equilibrated segment of
a trajectory, and is RMS enough to establish if this equilibration has been
achieved? I am a little confused because theoretically in order for an
MMPBSA+ entropy calculation to give ensemble-average type quantities the
trajectory should be equilibrated and long I believe. This should be
especially true for a quasi-harmonic approximation since the assumption is
that the system lies in what can be approximated as a broad energy well
(i.e. equilibration I assume). Nonetheless in many articles I see people
running simulations for a certain time and then performing mmpbsa on the
last few ns of the trajectories, without any reference to whether
equilibration was achieved or not. I'd be very thankful if someone could
shine light on this matter for me.

Thanks,

Laura


On Tue, Apr 28, 2015 at 9:40 PM, Jason Swails <jason.swails.gmail.com>
wrote:

> On Tue, Apr 28, 2015 at 3:40 PM, Laura Tociu <ltociu.princeton.edu> wrote:
>
> > Thank you both!
> >
> > Yes, I was saving frames every 20 femtoseconds in order to obtain that
> many
> > frames (I had 5 ns of equilibrated trajectory I wanted to use). It seems
> > like that is bad because the frames are not decorrelated.
>
>
> ​Exactly. Two correlated snapshots contain significantly less information
> than two independent ones.
> ​
>
>
> > ​​
> > What exactly is a
> > ​​
> > good time step for entropy calculation? Around 1 ps?
> >
>
> ​It's a compromise. The longer the better as far as ensuring that whatever
> property you're measuring will have a correlation time shorter than the
> sampling interval. But if you wait *too* long, and your simulation is too
> short, then you lose potentially helpful information.
>
> So the answer likely depends on how long you plan on running (or how long
> you can afford to run). I typically sample around 1 to 2 ps. Other
> colleagues sample more like every 10 ps (or maybe longer?) -- but they run
> longer simulations than I do.​
>
> One more question: If I use only the alpha carbons, what do I do with the
> > ligand, which is free arginine? If I represent it just by one atom, its
> > alpha atom, it seems a little weird. Should I include all of the
> arginine's
> > atoms both in the compex trajectory and the arginine trajectory? Or does
> it
> > not make a big difference what I do with the arginine since its entropy
> > will be small compared to that of the complex and receptor?
> >
>
> ​Are you using a single- or multiple-trajectory approach here? If you use
> the single-trajectory approach, ignoring the ligand will be pointless --
> the complex and receptor will then be identical and your entropy
> contributions will be exactly zero.
>
> You could clip your system to the entire ligand + receptor alpha carbons
> and see how that fares. You could also simulate the free receptor and do a
> multiple-trajectory approach to compute the binding entropy, and ignore the
> ligand (you could always add a correction assuming the ligand was rigid and
> compute just the rotational and translational entropy)...
>
> I'm not sure what the best approach would be. The paper Josh mentioned is
> probably worth a read to see what they suggest.
>
> All the best,
> Jason
>
> --
> Jason M. Swails
> BioMaPS,
> Rutgers University
> Postdoctoral Researcher
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Apr 28 2015 - 22:00:03 PDT
Custom Search