Re: [AMBER] Regarding impropers from and guessing from bond topology

From: Jason Swails <>
Date: Sun, 19 Oct 2014 00:52:51 -0400

On Oct 18, 2014, at 11:20 PM, Arun Srikanth <> wrote:

> Thanks Jason and Daniel. I am using topotools (A plugin in vmd). You can
> load the mol2 file generated by antechamber in vmd.
> Then use topotools guessimproper in vmd to get the impropers for the
> system. The improper definition it employs is
> "Guess improper angles. Are defined for atoms that are bonded to *exactly*
> 3 atoms when they form a near flat structure. The tolerance flag can be
> used as a cutoff for what is to be considered as "flat" and what not. “

This is not the same criteria as that used for tleap, which my last email described. Now I’m not *absolutely* positive that what I said there is the actual algorithm implemented in tleap for determining potential candidates for improper torsions, but I believe it is close enough. In any case, there should not be any geometry-based decisions that are made by tleap, since then your parametrization will depend on the starting conformation of the molecule (not good).

Also, not every *potential* improper site needs to be given an improper term to keep it planar (and, as Dave said before, “planarity” isn’t even a requirement -- it’s just the typical use for improper torsions).

So the algorithms in VMD guessimproper and tleap are quite different. While I would expect them to find many of the same 4 atoms as “improper centers”, do not expect full agreement between them.

Also, the name of the VMD tool is telling -- _guess_impropers; that’s all it is, a guess based on simple geometric measurements.

> You can now compare the two impropers. They are not the same.
> I checked this for dihedrals. Some how the dihedrals from topotools and
> unique dihedrals from leap are matching.

This I would expect. Proper torsions are defined between 4 atoms that are connected by 3 sequential bonds. Every proper torsion is required to have a parameter by tleap, so there is no “guesswork” in this case.

> So My question is
> 1. WIll the connectivity of the dihedrals, impropers guessed from bond
> topology and the connectivity of the dihedrals, impropers assigned by leap
> (based on parametrization, leave out the multi terms or the duplicates)
> match always ?

You can’t guess impropers simply from bond connectivities. You can guess proper torsions based on bond connectivities. But any two conformations of the same molecule (regardless of how ridiculous those conformations may be) will be parametrized identically. [1]


[1] This does not include charge derivation, which depends sometimes strongly on the conformation; so choose a good one(s).

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Sat Oct 18 2014 - 22:00:02 PDT
Custom Search