From: Gianluca Interlandi (gianluca_at_u.washington.edu)
Date: Wed Dec 16 2015 - 12:49:26 CST
Do you select BAR in the VMD ParseFEP window? Have you looked at the
histograms for each window and do you see overlap between forward and
backward?
I don't know if it helps, I make the equilibration window as large as the
data collection window.
An important consideration is that the ideal situation for FEP is when the
target state is a subset of the initial state. Since this is often not the
case, the transformation is broken down into multiple windows but the
windows need to overlap. One way to find this out is by looking at the
histograms, as I mentioned above.
Also, I am not familiar with your system. Is the mutated side chain buried
in the protein or out in the solvent? Mutation of a side chain that
samples a vast conformational space might be difficult to assess with FEP.
You might want to first run a 50-100 ns simulation after mutating the side
chain (with no FEP) to get an idea what the side chain does in the mutated
state.
The following papers might be helpful to you:
http://www.ncbi.nlm.nih.gov/pubmed/22978683
(in particular supporting information)
http://www.ncbi.nlm.nih.gov/pubmed/14634991
Gianluca
On Wed, 16 Dec 2015, Brian Radak wrote:
> My guess is that your first thought is correct - the simulations are not long enough.
>
> I would recommend:
>
> 1) assessing the statistical error (I'm not really familiar with ParseFEP, so I don't
> know if it does this already)
>
> and/or
>
> 2) trying another analysis method (such as a multistate technique)
>
> HTH,
> Brian
>
> On 12/15/2015 09:59 AM, Haleh a wrote:
> Dear NAMD users,
> I have been trying to calculate free energy changes upon a mutation in a
> protein via FEP method. The transformations were subdivided into 32 windows.
> Each stratum consisted of 100,000 data-collection steps preceded by 25,000
> equilibration steps.
> For a few of my mutations the free energy change upon forward and backward
> transformations were not the same. An example of a Parsefep file for dG column
> is given below:
>
> fw bw
> 0.0000 19.2024
> 7.8597 27.3887
> 15.8585 35.4349
> 23.8350 42.7734
> 30.9869 49.8046
> 38.1433 56.4785
> 45.1326 63.2339
> 51.5564 69.8128
> 57.9795 75.7305
> 64.2681 81.6443
> 70.2732 86.9015
> 76.0547 91.6371
> 81.0941 95.7292
> 85.2021 99.6043
> 88.3743 103.2919
> 91.4066 106.3001
> 93.3668 108.3961
> 91.4244 106.6218
> 89.3347 104.5966
> 87.0627 101.5024
> 84.6330 97.8175
> 81.7766 93.0725
> 78.8473 88.2228
> 75.5645 83.1926
> 70.4596 77.7998
> 65.0760 71.9186
> 59.6825 65.5346
> 54.1291 59.0037
> 48.5446 52.6676
> 42.7863 45.8779
> 36.9261 38.9123
> 31.0231 31.7249
> 25.1445 25.1445
>
>
> I tried running the simulation with longer data collection and equilibration
> time but it did not help. Does anyone have any idea why such a problem exist
> and how to solve it?
>
> Many thanks,
> Haleh Abdizadeh
>
>
> --
> Brian Radak
> Postdoctoral Appointee
> Leadership Computing Facility
> Argonne National Laboratory
>
> 9700 South Cass Avenue, Bldg. 240
> Argonne, IL 60439-4854
> (630) 252-8643
> brian.radak_at_anl.gov
>
>
-----------------------------------------------------
Gianluca Interlandi, PhD gianluca_at_u.washington.edu
+1 (206) 685 4435
http://artemide.bioeng.washington.edu/
Research Assistant Professor at the Department of Bioengineering
at the University of Washington, Seattle WA U.S.A.
-----------------------------------------------------
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:22:20 CST