Hi Dr. Cheatham,
Thank you very much for the detailed reply and your valuable time.
Even though there are many crystal structures of this protein,
unfortunately they are of low resolution >2.8 and above). The interested
binding pocket is so voluminuous with several subsites within. We have tried to dock
chemical compounds (congeneric series with substituents of different size and nature) with no success.
These ligands are thrown all over the places in erratic binding modes.We believe
that one of the subsites (mainly polar residues) might have a conserved water network.
Lu. et. al (J. Chem. Inf. Model. 2007, 47, 668-675) have shown from comprehensive analysis, that the B-factor for the
buried and interface (protein-ligand) waters are quite low and varies with the number of polar interactions.
So we think that the calculation of B-factors would possibly shed more light on this presumed water network.
I haven't tried Grid and not sure about what will be the output details. I will explore and get back to this list
if I have any questions.
I sincerely appreciate your valuable comments and suggestions.
Have a good day !
Senthil
> I would like to calculate B-factor for only ~10-15 water molecules (closest to the ligand).
Why? Perhaps if there is a high resolution crystal structure, there
are experimental values to compare too. However, what will the B-factor
tell you that you already do not know (for example that water is often at
the site). Perhaps you are looking to show that the bound waters are less
mobile in some sites than others? (but you already know this via
occupancy/lifetime and could easily visualize it and/or quantify it via a
density/occupancy grid with the grid command).
> Hbond analysis shows that there are 4-5 water molecules making strong (high occupancy and lifetime) H-bond interactions with the
> binding site residues as well as the ligand. I am not sure how to accomplish this task.
>
> I am familiar with calculating bfactors for residues (after rms fitting with the reference (average) structure).
...
> I created a new trajectory with Na+ stripped and only with closest 50 waters (the water molecules visually show huge fluctuations).
> I am not really sure that the closest water molecules that are kept as closest remain the same through out the trajectory.
They do not retain the same identity and therefore the average structure
(of the water) and the B-factors estimated are probably inaccurate. You
can see this by coloring the waters different colors (or coloring one
water a different color) and you will see it jump all around in a movie...
Alternatively, create the average structure and visualize it, and the
water will tend to coalesce into a blob rather than show distinct
identities. Without the same water ordering (and factoring in water
exchange) you cannot calculate diffusion, B-factors, average structures,
etc.
> Now I am doing the bfactor calculation with the new trajectory.
> step 1: average structure
> step 2: rms fit with average
What average structure? Average structure of only the protein or of the
protein + water? I would think that it would only make sense to create
the average structure of the protein (since the waters exchange). Even
if the water did not exchange, fitting to mobile water will distort the
structure as the mobile water will be more noisy. To create an average
structure of protein + water would require each site having the same water
# in each frame which cannot be done at present.
> step 3: bfactor
>
> Am I on the right track ? Please suggest me if I am doing anything wrong ?
At present, if I were forced or tortured into calculating B-factors for
bound water, I would do this with reference to the average structure of
the protein (solute only) and do this separately for each bound water of
interest. Then each would be in the same frame of reference. To do this,
I would look at the water density in the region of interest via "grid" and
then find nearby atoms to each specific water and make a list of those
atoms. Alternatively, from the hbond analysis, I would already have a
good idea of what those high occupancy regions were and would therefore
already know the close atoms... Then, for each atom in my list, create a
trajectory with the 1-closest water to that atom and do the analysis. If
I were interested in 5 waters, that would be five separate ptraj "closest"
runs. Finally, as the water rotates, I would look at the difference
between the so called B-factor for water-oxygen only vs. with the
hydrogens since "non-rotating" waters would presumably have lower
B-factors (meaning the oxygen only and H20 B-factors would be similar).
However it is not clear to me what this will tell you beyond what you can
infer from the grid, hbond, and perhaps RDF (radial) analysis.
This still would not be directly comparable to experiment since the
average structure would have to include the defined waters too. To get
them in the same order, you could hack it by converting to PDBs and clever
text manipulation to create the trajectories with multiple waters. There
may be a way to do this more easily in VMD, however I am not an expert.
To do this with AMBER is an advanced endeavor and will require some real
coding/scripting and work.
Note that code is in the works by various developers that will allow you
to better keep track of important bound waters for various further
analyses, but it likely will not be available in the near term...
--tec3
_______________________________________________
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 Thu Jul 26 2012 - 08:00:04 PDT
