Re: [AMBER] Issues with running simulation after ParmEd conversion

From: Jason Swails <>
Date: Wed, 29 Jul 2015 09:49:44 -0400

On Tue, 2015-07-28 at 16:30 -0700, Rohith Mohan wrote:
> Hello all,
> I have a fairly large protein system (approximately 394651 atoms including
> solvent) that I have minimized, heated and equilibrated in NAMD and am
> trying to perform the production run in AMBER (with CUDA) so that I can use
> a specific aMD method. I utilized the latest version of ParmEd to convert
> the PDB (taken from the last frame of the NAMD equilibration trajectory)
> and PSF using the CHARMM c36_aug14 topology and parameter files as shown in
> the commands below:
> chamber -param par_all36_prot.prm -param par_all36_lipid.prm -param
> par_all36_cgenff.prm -str toppar_water_ions.str -psf
> C3_full_solv_ionized.psf -crd C3_full_solv_ionized.pdb
> outparm C3_equil.prmtop C3_equil.inpcrd
> After several errors during production run attempts and troubleshooting, I
> finally utilized cpptraj to check the structure and found that there were
> many clashes with atoms in close contact (<.2 A) with each other. I also
> verified that these clashes did not exist before conversion in the input
> PDB. Additionally, I repeated the conversion process with a much smaller
> system (~17,000 atoms) and found that there was no issues there so I think
> the size of my system could be affecting the conversion.
> I was wondering if anyone else had any issues with converting large systems
> using ParmEd and might have any suggestions or guidance for me.

I don't think that the size of the system should really impact things
here (outside of perhaps taking a few minutes to process completely).
What I've usually had the most trouble with in terms of chamber
conversion has been the unit cell definition.

Unlike Amber, CHARMM doesn't store the unit cell vectors or dimensions
in the coordinate files or the PSF (I'm actually not sure where that box
information is stored for restarting simulations with a barostat). So I
suspect what's happening here is that the unit cell dimensions are not
being set correctly for your system. Does your PDB file define a CRYST1
record? If so, where did those values come from? If not, then you are
missing unit cell information and it doesn't seem to be getting set

If you are using NAMD, then I think it prints an XST file or something
like that that defines unit cell vectors. Have you made sure to convert
those unit cell vectors to the appropriate tuple of vector lengths and

What happens if you visualize your system in VMD (using the original PSF
file)? Does the structure looked warped?


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Wed Jul 29 2015 - 07:00:03 PDT
Custom Search