in gromos force fields, PRO-NH n-terminus type is missing a dihedral term between CD-N-CA-C atoms.
#1 Updated by Roland Schulz almost 5 years ago
Could you give some more detail? Please explain what the correct dihedral term is (ideally with citation) and a pointer to the line and file where it is missing. Also please say in what Gromacs version it is wrong. Ideally you could upload a fixed version to gerrit.gromacs.org.
#3 Updated by Drazen Petrov almost 5 years ago
This is more of a gromos force field problem than a gromacs problem, meaning, the error is in the n-terminus database file of different gromos force fields (*n.tdb files) and the error is in all versions.
The problem is that pdb2gmx produces topology file with a missing dihedral when PRO-NH type of n-terminus is chosen. Simply by comparing PRO-NH to PRO-NH2 type (in the n-terminus database file "*n.tdb"), one can see that the latter has H1-N-CA-C dihedral term, which is missing in the former.
I have written that CD-N-CA-C dihedral term is missing because a heavier atom is usually used if possible, which is the case for proline. However, to be consistent with the PRO-NH2 type, it is probably better to add H-N-CA-C dihedral. The correct term depends on the force field in question (e.g. correct term for gromos45a3 is gd_14).
I could not find where to upload fixed files on gerrit.gromacs.org. In principle, I could send it via email.
Also, in the meanwhile, I have found another problem with pdb2gmx related to gromos force fields. CA atom of the fist residue in a produced topology file is not in the same charge group as the n-terminal nitrogen and hydrogen atoms. Therefore, the sum of partial charges in the first and the second change group is a non-integer number.
#4 Updated by Justin Lemkul almost 5 years ago
I looked into this a little bit myself and I'd like to add a bit more information. In fact, pdb2gmx does write a dihedral term for
N-CA for both termini; the problem is that for PRO-NH, there are no associated parameters (should be gd_39 in 53A6). PRO-NH2+ writes a dihedral term correctly, but it appears as if this was added later as a workaround. If one comments the dihedral line for PRO-NH2+ in the .n.tdb file out, it produces the same as PRO-NH (writing a dihedral for CD-N-CA-CB). The H-N-CA-C, CD-N-CA-C, and CD-N-CA-CB dihedrals would all be equivalent, so only one needs to be assigned.
There is a compounded problem, in that the .n.tdb file assigns gd_29 instead of gd_39. I think a universal solution is to assign the dihedral as CD-N-CA-C in the .rtp file for [PRO] and remove the added [dihedral] in the .n.tdb file. The .rtp entry was built from the original GROMOS force field files, which specify it as (
C)-N-CA-C, but that obviously is causing problems. Any -N-CA dihedral should have the same parameters, irrespective of atoms i and l.
As for the charge group issue, I can't say what is right, but the GROMOS force fields appear to be the only ones that add an extra field in the [replace] directive that seems to indicate the CA atom is placed in a new charge group. This behavior is not described in the manual. Roland, any ideas? I can make the appropriate fixes in the git repo and submit to gerrit, would just prefer to do one fix rather than two tiny ones. The dihedral issue seems to be easy to fix, I'm just not so sure about the charge group thing.
#7 Updated by Drazen Petrov almost 5 years ago
I have completely missed that the dihedral term in question is actually incorrect for the PRO-NH2+ type. So, it should be gd_19 for 45a3, not gd_14 as I suggested.
I think that assigning CD-N-CA-C dihedral for PRO in the rtp file is not a very good solution because it is redundant, since every amino acid, including PRO, has -C N CA C dihedral already assigned, which is the same dihedral. Probably the easiest would be to assign CD-N-CA-C dihedral for both PRO-NH and PRO-NH2+ types in the n.tdb file.