Re: [AMBER] Incorrect Bond Assignment in AMBERTools 12 LeAP

From: Jason Swails <>
Date: Tue, 16 Oct 2012 13:15:35 -0400

On Tue, Oct 16, 2012 at 12:59 PM, Aron Broom <> wrote:

> I recently used tleap within AMBER tools version 12, to generate some
> protein parameter files, using AMBER_FF12SB.
> The command structure was:
> > set default PBradii mbondi2
> > model = loadpdb protein.pdb
> > saveamberparm model protein.prmtop protein.inpcrd
> Everything seemed to be working fine, but in trying to troubleshoot a
> temperature problem I was having with OpenMM, I realized that some of the
> glycine (CA-HA2 and CA-HA3) bonds were not being properly constrained. It
> looking at the .prmtop file, it looks as though those bonds were not added
> to the BONDS_INC_HYDROGEN list.

This would shock me. If you create a sequence ACE-GLY-NME in leap and save
the prmtop, you can clearly see the CA-HA2 and CA-HA3 bonds in the
BONDS_INC_HYDROGEN section. In this case, I would be less likely to
suspect tleap than OpenMM (tleap is significantly older and more thoroughly

I haven't done any of the bugfixes for ambertools 12, but I don't see
> anything about that listed. Is this something that anyone else has seen?
> I haven't done a rigorous test to ensure it is reproducible, thought I
> would see if anyone else has noted this kind of problem before. As an FYI,
> the input pdb files I was using don't have hydrogens at all, so there
> shouldn't have been any problem with LeAP not recognizing the atom name.

What happens if you run it through sander or pmemd? Sander reports the DOF
in your system, properly adjusted for SHAKE constraints. Therefore, you
should be able to use sander with "noshakemask" targeting the problematic
residues and see if the DOF changes between the two simulations. If the
DOF changes, then sander knows to SHAKE them, further suggesting OpenMM as
the culprit.


Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
AMBER mailing list
Received on Tue Oct 16 2012 - 10:30:06 PDT
Custom Search