From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Fri Oct 03 2014 - 13:16:37 CDT

On Fri, Oct 3, 2014 at 1:21 PM, Josh Vermaas <vermaas2_at_illinois.edu> wrote:
> Hi Dave,
>
> The issue here is actually something that is usually quite benign (I think).
> You say you are loading in pdbs, which normally don't have connectivity
> information (I dunno how packmol does it, but I don't see many with PDBs
> with conect records filled in). This means you are implicitly counting on
> VMD's bond-search algorithm to determine all of your connectivity. VMD's
> bond-search algorithm is efficient, which means it breaks down the N
> particles, and assigns them to grid patches based on their location in space
> (not the order they were loaded), searching for atoms that are close from a
> small subset of the system, vastly reducing the number of distance
> comparisons that need to be made. Now you said you let packmol make all of
> these systems independently, which means that by chance alone, some
> molecules are split across these patches differently in one structure
> relative to another. The order of the bonds stored internally will be
> different each time! You can test this by asking VMD to write out a psf
> after the bonds, angles and dihedrals have been guessed. I'm betting the
> order of the indices will not be sorted, and that the bondlists will be
> different for each system. Now topotools iterates linearly over this
> bondlist to build up angles and dihedrals, which I *think* is why it has
> this weird behavior. Why does the order of the angles and dihedrals matter
> though? Is this something LAMMPS specific? I'm also confused as to why the

josh,

this was not about the order of the angles and dihedrals, but the *types*.
a psf file doesn't even store that information, but it is crucial in
LAMMPS, because they are not referred to with a symbolic name but a
number. the symbolic names that were quoted are only comments. if you
want to use the same (bond/angle/dihedral) parameter definition for
multiple simulations of the same system, those names need to be
ordered, so the assigned numbers are the same.

in fact, for systems with more variety, it may be needed to pad the
list for interaction types with dummies that are not present in a
specific topology to enforce the reuse of parameter files. with
CHARMM/NAMD this matching and assigning is all done at a later step
during the MD run. that has some advantages (it is simpler to create
the topology), but also some drawbacks (it requires that the bonded
interaction types are uniquely defined by the atom types comprising
that interaction).

axel.

> angles and dihedrals are specified based on what look like atomnames and not
> indices, but I'm sure its just a formatting thing.
>
> -Josh Vermaas
>
>
> On 10/03/2014 11:08 AM, Dave Schall wrote:
>
> Hi vmd-users,
>
> I am converting pdb files to a lammps coord files using vmd/topotools. I
> have a system with two types of molecules, say A and B. I am using packmol
> to generate pdb files for systems with different mole fractions of A and B.
> Packmol generates lists all the A molecules first then, all the B molecules
> after.
>
> I take this pdb file and read into vmd, do some atom selections to set
> charges/types/masses, etc. I have topotools guess the angles and dihedrals.
> Then I dump to a lammps coord file via topotools. I have a tcl script set up
> to do this the same way every time.
>
> Since I am running a fairly large set of simulations, I'd like this to be
> somewhat automated, but what I find when I inspect the lammps coord files is
> that the order of angles and dihedrals is not always the same. This means I
> have to check each coord file against my lammps input file to make sure the
> order of the parameters match.
>
> For example for some files I get:
>
> # 1 C3-C4-H1
> # 2 C-C1-H
> # 3 C1-C-H
> # 4 C2-C-H
> # 5 H1-C3-H1
> # 6 C-C2-H
>
> For some others:
>
> # 1 C3-C4-H1
> # 2 C-C1-H
> # 3 C1-C-H
> # 4 C2-C-H
> # 5 C-C2-H
> # 6 H1-C3-H1
>
> (5 and 6 as switched)
>
> Dihedrals are much more mixed up than angles.
>
> Am I wrong in assuming that if the order of molecules (and atoms therein)
> are the same in every pdb file they should be read in the same order by
> vmd/topotools? Is this some function of how "topo guessangles" and "topo
> guessdihedrals" works? If so I guess I better get busy building a a
> parameter database and script that can create input decks from lammps coord
> files.
>
> Thanks,
>
> Dave
>
>
>
>
> --
> J. David Schall, Assistant Professor
> Dept. of Mechanical Engineering, Oakland University
> 328 Engineering Center, Rochester, MI 48309
> 248-370-2870
>
>

-- 
Dr. Axel Kohlmeyer  akohlmey_at_gmail.com  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.