Re: [AMBER] Density for a protein ligand complex run

From: Jason Swails <>
Date: Wed, 03 Jun 2015 09:01:23 -0400

On Wed, 2015-06-03 at 11:56 +0530, Mrinda Jones wrote:
> Hi Dac and Hector,
> The system details are:
> Force field : gaff and ff99SB
> System Length : 212 residues (protein) and one Drug
> Overall Protein is Neutral and Drug has -2 Net charge
> Thermostat and barostat used: Langevin thermostat (NTT=3),
> MC Barostat was default i.e 1, ntp=1
> My question is, is this density being less than 1gg/cc justifiable in terms
> of MD run. Is the run okay?

You are placing too much emphasis on the density here. First of all,
there is no reason that your density should be 1 g/cc. In the real
world, the density of *pure* water is (almost) 1 g/cc only at 4 C. At
30C, the density is 0.995, and 0.983 at 60 C. The density of a solution
is going to be different than that of pure water. Depending on the
solute, it can be more or less dense than pure water (as two examples, a
salt solution is more dense than water, while I believe that isopropyl
alcohol solutions are less dense). It all has to do with how tightly
the water can pack around the solute, how much space the solute takes
up, and how dense the solute is within that space compared to the amount
of water it's displacing. Since proteins come in varying sizes, shapes,
and folds, it's impossible to make sweeping generalizations about how
proteins affect density.

Secondly, as I mentioned, you are not simulating pure water -- the
effective protein concentration is actually very high compared to
anything you would have experimentally (and it's hard to make it more
dilute without adversely affecting your sampling rate).

That said, if the density is absurd (like 0.5 g/mol or lower, or maybe
1.5 g/mol or higher), while you suspect it *should* be close to the
density of your water model, then there is almost certainly a problem.
Maybe your simulation blew up, the box is too big, the parameters are
wrong, ... etc. But it would be obvious upon looking at the trajectory
in VMD or PyMol (or whatever).

Your density here is quite close to 1 g/cc, so in my opinion it tells
you very little. I usually see densities with proteins slightly higher
than 1 g/cc (~1.01 to 1.02 g/cc) using TIP3P, but a value of 0.98 would
not be cause for alarm. So when you ask "Is the run okay?", the answer
is "we have no idea." And what that means is that the density is not an
obvious problem, but there are a very large number of ways to have a bad
simulation with a good density.

> When I visualize in VMD the trajectory for the drug-protein interaction
> does not any show blow off. Moreover the protein I have used is a modeled
> one. The temperature, pressure, volume and Energies are showing stable
> graphs as expected from a normal 80 ns run.
> What can a rmsd data tell about the simulation?

It's a coarse metric that measures how "different" the overall
conformation is compared to some reference structure.

> Is it necessary for the
> rmsd to get stabilized below 3 or 4 Angstrom in case of modeled protein
> runs as well as there is an non-modeled loop region towards the
> protein-tail.

Depends on your study. Sometimes an RMSD of 1 A is too large. Other
times an RMSD of 3 A is too small.

These may not be satisfying answers, but the nature of computational
research is that there is no hard-and-fast set of rules you can apply to
every study.


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Wed Jun 03 2015 - 06:30:02 PDT
Custom Search