[gmx-users] OPLS-AA Unknown Atomtype

Justin A. Lemkul jalemkul at vt.edu
Tue Aug 23 23:07:28 CEST 2011

Yao Yao wrote:
> Hi Justin,
> Thanks for your reply. Here is the exact error message,
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Back Off! I just backed up mdout.mdp to ./#mdout.mdp.1#
> checking input for internal consistency...
> processing topology...
> Opening library file /share/apps/gromacs407/share/gromacs/top/ffoplsaa.itp
> Opening library file /share/apps/gromacs407/share/gromacs/top/ffoplsaanb.itp
> Opening library file 
> /share/apps/gromacs407/share/gromacs/top/ffoplsaabon.itp
> -------------------------------------------------------
> Program grompp, VERSION 4.0.7
> Source code file: toppush.c, line: 620
> Fatal error:
> Unknown bond_atomtype CA1
> -------------------------------------------------------
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> I understand that I have not inserted CA1 into "atomtypes.atp". because 
> if I add CA1 as a new atom, I have to mention all bonds, angles, 
> dihedrals of it.

The atomtypes.atp file is irrelevant here; it is only used by pdb2gmx.  What you 
are trying to avoid is what you have to do - if you introduce a new atom type 
and intend to use it in any bonded interactions, you must introduce relevant 
parameters for all the interactions in which it will participate.

> So I denoted the type of it as a known carbon atom in my topology file 
> (in the attachment). I thought in OPLS-AA, CA1 can "cite" this known 
> atom since CA1 is just a name.

Names and types are different.  You can name an atom anything you like, but atom 
types must be judiciously assigned.

> Meanwhile, in the ffbonded.itp and ffnonbonded.itp, I still use CA1 to 
> give bonds, angles, and dihedrals of CA1.

If you have inserted all of the correct parameters in these files, you would not 
receive the error above.  Perhaps herein lies the problem - if you have modified 
ffnonbonded.itp (which belongs to the force field organization of the 4.5.x 
series), it will have no effect on a 4.0.7 executable, which is being called 
above.  So (at least) one of two possibilities is true:

1. You're not using the Gromacs version you intend to.
2. You haven't introduced all of the parameters you need to.

The bond_atomtype values are assigned in ffnonbonded.itp (counterintuitive, I 
know, but that's how it works).  So for the problematic atom (type opls_137) the 
bond_atomtype is CT (second column of ffnonbonded.itp in [atomtypes]).

> So generally speaking, I am wondering if an atom with a different name 
> gets introduced into OPLS, it has to be denoted again (bonds, angles, 
> ...) and added into the atomtypes.atp, even it is a carbon atom?
> And do you happen to know any tutorial about how to introduce new 
> molecules into OPLS?

Again, names are irrelevant but types are critical.  There is no tutorial, per 
se, but the instructions are on the wiki in a general form:


Most of the instructions are designed around residues for which pdb2gmx will 
act.  In your case, any new atom types need to be defined in ffnonbonded.itp and 
then bonded interactions making use of those types in ffbonded.itp.



Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list