VMD-L Mailing List

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Fri Jan 15 2010 - 22:36:14 CST

On Sat, 2010-01-16 at 04:54 +0100, MyLinkka wrote:
> Hi, Alex,

> Indeed we have different understanding of improper.
> To me it's nothing very different from the dihedral.

it is not very different from how it is being implemented
or how you would define it. but they are _very_ different
in how they are used.

> I think that's also explains why most books and tutorials
> about MD draw pictures like I referred to. Or vice versa.
> However, from quantum mechanics point of views, the sticks
> connecting atoms (so-called bond in force field) does not
> really exist, let alone the angle, dihedral and improper.

this is totally irrelevant for our discussion.

> Then, indeed, we have to guess everything in order to use
> force fields, because most force fields contain those four
> terms. On the other hand, molecular structure file usually

the force field indeed contains parameters for all, but
that again is irrelevant, since they get looked up or
assigned based on the topology information.

> contains only positions or position+bonding, like car/mdf,bgf,
> because these two things really define the topology of a
> molecule. (or we can say the topology people want. )
> However, this is not enough to meet the requirement of a
> force field, we still need to provide angle, dihedral,
> improper or maybe more depending on the force fields.

> by this common definitions, improper does not need to be guessed
> it can be derived or searched base on the bonding information.

in all the topology generating tools that i know of, improper
definitions are always looked up based on residue or template
data base. and in several cases impropers were defined between
atoms that were _not_ bound together in the way you have
described it. what you described, is only the most common
form of an improper.

in any case, just create a .psf file with psfgen for example
and look up the number of dihedrals and impropers, and you'll
see that there are by far fewer impropers than dihedrals.
however, given your reasoning, that should not be the case.
furthermore impropers are explicitly given in charmm force
field topology files, angles and dihedrals not.

> We can define more distinct structures and names, but those are
> not needed by force fields normally. The essential point here
> is we do 'guess' for these geometric information for the force
> fields, or they are meaningless. Then, why not do it following
> the most common definition of force fields. In this sense, tools

topotools _is_ following the most common way force fields are
defined. what you are suggesting is uncommon. what you suggest
is technically the same as to defining additional bond/angle/dihedral
interactions and then setting all their coefficients to zero.

> like msi2lmp do not just do translation, they build angle, dihedral
> and improper geometric information then map the force field
> parameters to them. Please have a look of some force field files

but this is exactly what i am saying. you cannot guess impropers
without _additional_ information. in your case you get it from
the force field database, where distinct atom types contain the
important information about what kind of structure you have and
then you can derive whether you need to add an improper interaction.

this is effectively the same as setting them explicitly.
topotools does not have that information available.

> you can easily understand this. The definitions of improper angle in
> different force fields are essentially the same. This way people don't
> really have to specify the parameters manually.

this is not about parameters, and different force fields
do have

> Back to topotools, if some tools can export bond, angle, dihedral
> and improper explicitly, it is very convenient for users to assign
> force field parameters, although it can be not necessary. see the
> lammps command improper_style you will understand this. Or this
> automatically typing and force field parameter mapping will automatically
> done by some sophisticated software like Materials Studio.

for that kind of functionality VMD already has psfgen.
this is not what i am aiming at.

you have the source code for topotools, feel free to implement
a guessimpropers functionality still believe you need it. i am
certain, that you will realize in the process that it is a
pointless effort. i didn't omit it, because it didn't think
about it, but rather because it doesn't make sense in the
context of the other guessXXX functions.

cheers,
    axel.

> Thank to the discussion. It makes me get better understanding of
> the relationship of topology and force field.
>
> Ting 

-- 
Dr. Axel Kohlmeyer  akohlmey_at_gmail.com 
Institute for Computational Molecular Science
College of Science and Technology
Temple University, Philadelphia PA, USA.