[gmx-users] RE: diffusion constant level off

[Dr. Vitaly Chaban]

The simplest advice is to increase sampling.



> Please see plot:

> http://imageshack.us/photo/my-images/35/diffusion.png/
>
> 2013/3/28 Justin Lemkul <jalemkul at vt.edu>
>
> > On Thu, Mar 28, 2013 at 9:59 AM, Ahmet yıldırım <ahmedo047 at gmail.com>
> > wrote:
> >
> > > Dear users,
> > >
> > > Again, I have strange results (for 10,50,100,150,200 ns). I am
> wondering,
> > > is there a bug with g_msd?
> > >
> > >
> > I see no evidence for a bug, and you should avoid such speculation unless
> > you know exactly how the program should behave. Only then, after an
> > analysis of known quantities or behavior, can we discuss bugginess. Do
> you
> > know how g_msd works? Do you know what all of the flags are doing, or are
> > you just making adjustments hoping for clarity? You may find the
> following
> > post very illuminating:
> >
> > http://lists.gromacs.org/pipermail/gmx-users/2010-July/052512.html
> >
> > -Justin
> >
> >
> > > Commands for trestart to 20 ps:
> > > g_msd -f traj.xtc -s topol.top -o msd1.xvg -trestart 20 -beginfit -1
> > > -endfit -1 -b 0 -e 10000
> > > g_msd -f traj.xtc -s topol.top -o msd2.xvg -trestart 20 -beginfit -1
> > > -endfit -1 -b 10000 -e 20000
> > > g_msd -f traj.xtc -s topol.top -o msd3.xvg -trestart 20 -beginfit -1
> > > -endfit -1 -b 20000 -e 30000
> > > ...
> > > g_msd -f traj.xtc -s topol.top -o msd20.xvg -trestart 20 -beginfit -1
> > > -endfit -1 -b 190000 -e 200000
> > >
> > > Commands for trestart to 1000 ps:
> > > g_msd -f traj.xtc -s topol.top -o msd1.xvg -trestart 1000 -beginfit -1
> > > -endfit -1 -b 0 -e 10000
> > > g_msd -f traj.xtc -s topol.top -o msd2.xvg -trestart 1000 -beginfit -1
> > > -endfit -1 -b 10000 -e 20000
> > > g_msd -f traj.xtc -s topol.top -o msd3.xvg -trestart 1000 -beginfit -1
> > > -endfit -1 -b 20000 -e 30000
> > > ...
> > > g_msd -f traj.xtc -s topol.top -o msd20.xvg -trestart 1000 -beginfit -1
> > > -endfit -1 -b 190000 -e 200000
> > >
> > >
> > > D1 (cm^2/s):Diffusion for trestart to 20 ps
> > > D2 (cm^2/s):Diffusion for trestart to 1000 ps
> > > Time (ns) D1 D2
> > > 10 0.1616 0.1091
> > > 20 0.0735 0.0679
> > > 30 0.0775 0.0705
> > > 40 0.1097 0.1189
> > > 50 0.1471 0.1436
> > > 60 0.0468 0.048
> > > 70 0.0667 0.0652
> > > 80 0.0727 0.086
> > > 90 0.0664 0.0707
> > > 100 0.1336 0.114
> > > 110 0.0899 0.0841
> > > 120 0.0572 0.0598
> > > 130 0.0506 0.0482
> > > 140 0.0723 0.0767
> > > 150 0.1466 0.1439
> > > 160 0.0703 0.0601
> > > 170 0.081 0.0853
> > > 180 0.0278 0.027
> > > 190 0.1121 0.1024
> > > 200 0.3136 0.2981
> > >
> > >
> > > 2013/3/28 Dr. Vitaly Chaban <vvchaban at gmail.com>
> > > >
> > > > >
> > > > > if I do as you said, I will get only one diffusion coefficient. I
> > want
> > > to
> > > > > calculate one diffusion coefficient for each 10
> > > > > ns of the simulation time of 200 ns. That is, I want to get 20
> > > diffusion
> > > > > values.
> > > > >
> > > > > 2013/3/28 David van der Spoel <spoel at xray.bmc.uu.se>
> > > > >
> > > > >> On 2013-03-28 10:40, Ahmet yıldırım wrote:
> > > > >>
> > > > >>> Dear users,
> > > > >>>
> > > > >>> This time, I calculated the diffusion coefficients of protein for
> > > each 10
> > > > >>> ns of the simulation providing a total simulation time of 200 ns.
> > > > >>> g_msd -f traj.xtc -s topol.top -o msd1.xvg -trestart 20 -beginfit
> > -1
> > > > >>> -endfit -1 -b 0 -e 10000
> > > > >>> g_msd -f traj.xtc -s topol.top -o msd2.xvg -trestart 20 -beginfit
> > -1
> > > > >>> -endfit -1 -b 10001 -e 20000
> > > > >>> g_msd -f traj.xtc -s topol.top -o msd3.xvg -trestart 20 -beginfit
> > -1
> > > > >>> -endfit -1 -b 20001 -e 30000
> > > > >>> ....
> > > > >>> g_msd -f traj.xtc -s topol.top -o msd20.xvg -trestart 20
> -beginfit
> > -1
> > > > >>> -endfit -1 -b 190001 -e 200000
> > > > >>>
> > > > >>
> > > > >> Set trestart to 10001 (no restarts), or do one run with
> > > > >>
> > > > >> g_msd -f traj.xtc -s topol.top -o msd20.xvg -trestart 10000
> > > > >>
> > > >
> > > >
> > > > I believe the advice was to increase "trestart" and you can decide
> > > > yourself how much to increase.
> > > >
> > > > Most likely, your protein deserves better sampling than 10 ns for a
> > > > linear diffusion.
> > > >
> > > > Dr. Vitaly Chaban
>