[gmx-users] constrained minimization

[David Mobley]

Rob,

I'll give you what I've got in terms of documentation, if I can find it
again (Here is the thread, I think:
http://www.gromacs.org/pipermail/gmx-users/2002-September/002760.html).
There are some other remarks related to this elsewhere on the mailing list
but they're a bit hard to find. As you noticed, it's not actually documented
in the manual. (And, when I started using dihedral restraints in 3.2, that
section of code hadn't been actually written yet, there was just a
placeholder which printed out "Dihedral restraints on," and set the energies
to zero). (Another e-mail I got info from was this one:
http://www.gromacs.org/pipermail/gmx-users/2004-February/009536.html)

I should also mention that I found a bug in this section of code in 3.3, so
you will need the dihres.c file from CVS, or alternatively if you e-mail me
directly I can e-mail you the fixed version. (The bug was that the energies
were right, but the forces were the negative of what they should have been,
meaning that it would make the dihedrals exactly the opposite of what they
should have been. See the bugzilla here:
http://bugzilla.gromacs.org/show_bug.cgi?id=32)

More below.



> >[ dihedral_restraints ]
> >; i j k l            type     label  phi  dphi  kfac          power
> >1410 1393 1391 2610  1        0      -140  0  fc_dihed     2
> >
> >mdp file:
> ><snip>
> >define = -Dfc_dihed=41.84
> ><snip>
> >dihre=simple
> >dihre-fc=1
> >
> >David
>
> Thanks David. Yes constraint = restraint in my last message. Where do you
> find the documentation for the [dihedral_restraints], because i didn't
> find
> it in the manual.
> Just so that i understand it correctly, the restraint is of the form:
> fc_dihed *(phi-dphi) ^power


Similar to that. There's a factor of 1/2 in front and it's better written by
taking phi'=phi-phi0, then V=fc_dihed/2*(phi-dphi)^power when phi'>dphi or
phi'<-dphi and zero elsewhere (that is, setting dphi to be nonzero makes it
a flat-bottom well of width 2*dphi).

so what then is type (1 for proper 2 for improper?!) and what is label?..
> and can i define my fc_dihed in my .top file directly? what is the
> advantage
> of defining it in the .mdp file
> what do dihre=simple, and dihre-fc=1 mean? are there any other "keywords"
> out there?


I really have no idea about the types, sorry. Maybe we can get the
developers to provide some documentation on this. Or you can look at the
source (dihres.c isn't particularly complicated) to see what it's doing. I
did at one point but I don't remember checking to see what the types do.
Maybe if you do so you can send an e-mail to the list for documentation
purposes. :)

"label", however, I can explain. Like with atoms, you number your restraints
in each section. So label 0 is the first restraint, label 1 the second, etc.


fc_dihed isn't necessary. You can replace it with numbers in your topology
file, or with a define fc_dihed section in your topology file if you want. I
just had it there for convenience (I was running with a bunch of different
dihedral restraint values for different jobs which would run at the same
time, and didn't want to have to make N different topology files, so I put
the fc_dihed value in my topology file and then used a define statement in
my individual mdp files to set it to different values for different runs.

dihre-fc is a multiplier for the force constant value in the topology file,
supposedly, if I remember correctly. I am not sure that it works, but it
might.

much appreciated if you can give me a pointer or 2 to the above questions.
> Thank you very much in advance.
>
> rob


Sorry I can't be  of more help.  Again, if you dig into these things more,
be sure to send an e-mail back to the list for documentation purposes. And
let me know if you need the dihres.c source rather than getting the fixed
version from CVS.

David


_______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20060111/8f65d027/attachment.html>