Re: Calculation of PMF between protein domains NAMD

From: Aron Broom (broomsday_at_gmail.com)
Date: Fri Apr 17 2015 - 17:57:48 CDT

There are certainly a large number of options for something like this:
umbrella sampling (as Josh mentioned), ABF, Metadynamics, and so forth.
But, to get a reasonable result, you'll need a very long total simulation
time. This is already a difficult problem for a simple ligand binding to a
protein, but protein-protein binding involves a lot more interactions and
moving parts. Based on what you've written, it seems like the binding
would have been completely broken by about 10ns into the simulation? I
would suggest, if you want to keep trying SMD, that you slow the pulling
velocity substantially, that is 10-100 fold, such that you capture the
relevant motion over the course of at least 100ns.

If you are thinking about something like umbrella sampling, I suggest doing
a search for NAMD and ligand binding with the authors: Gumbart, Roux and
Chipot. That will give you the full framework for how to get an accurate
energy with the least simulation time. Note that such a task isn't trivial
yet, it requires a lot of human time investment and understanding.

On Fri, Apr 17, 2015 at 6:43 PM, Josh Vermaas <vermaas2_at_illinois.edu> wrote:

> Hi Chola,
>
> The way I like doing this is to use the collective variables module to do
> umbrella sampling along some reaction coordinate that describes the binding
> process (usually based on coordnum, but unfortunately its broken in the
> released binaries of 2.10). Usually this doesn't deform the protein like an
> SMD pull would, since the weakest contacts would break first, and the
> protein is free to do what it wants, so long as the coordination number
> between the two parts is restrained.
>
> -Josh
>
>
> On 04/17/2015 05:26 PM, Chola Regmi wrote:
>
> Dear all,
> I used SMD simulation to calculate the PMF between two protein domains. I
> oriented the domains
> along the z-axis and applied the force in one domains using backbone
> atoms and kept other domain fixed at its COM of backbone atoms using
> "fixedAtoms on" command. I found the PMF very high as compare to
> experimental value. When I observed the trajectories I found that not all
> system of protein uniformly moved along z but the weak interaction region
> of protein moves faster than strong keeping COM motion correct with
> structural deformation. For me this gave the information about weak and
> strong binding sites but not the true PMF. I used spring constant 15
> kcal/mol/A^2 and pulling velocity 1A/ns. Can anyone suggest the best way to
> calculate PMF between protein protein domains using NAMD.
>
> Following is the extra-parameter section I used in NVT ensemble.
>
> #####################################################
> fixedAtoms on
> fixedAtomsFile ref.pdb
> fixedAtomsCol B
> #######################################################
> SMD on
> SMDFile ref.pdb
> SMDk 15
> SMDVel 0.000002 ;#velocity/2fs
> SMDDir 0 0 -1 ;#pulling direction -z
> SMDOutputFreq 50
> ##########################################################
>
> Thank you.
>
> Chola Regmi
> Virginia Tech
>
>
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:21:48 CST