gromacs topology file generated with amber99sb-ildn force field
The amber99sb-ildn force field applies several special 'define' commands to include the ILDN dihedral angle corrections over original amber99sb force field. For each residue among Ile/Leu/Asp/Asn there are several additional dihedral angle correction terms, but the topology file generated with gromacs 5.0.5 only includes the first term. This bug is also observed on gromacs 5.0.2, but not in old version of gromacs (4.5.5). The topology files generated based on 1UBQ using old and newer versions of gromacs are attached.
Fix bug removing multiple dihedrals in main rtp entries
The Gromacs-5.0 series has had a serious bug where pdb2gmx
would only consider the first entry when several explicit
bonds were listed for the same atoms in an RTP entry. Older
topologies have worked fine.
#2 Updated by Tairan Yuwen almost 2 years ago
Shun Sakuraba wrote:
The problem seems to exist in constructing hackblock, where the entries with the identical atom names are merged (i.e. erased).
Attached patch seems to fix the problem but I don't know whether this is a correct solution.
Thanks a lot. I tested that patch and now it could produce correct topology file with amber99sb-ildn force field.
#11 Updated by Berk Hess almost 2 years ago
I did some analysis to estimate what the effect is of the pdb2gmx bug on simulations with amber99sb-ildn.
With amber99sb-ildn the dihedrals that are modified with respect to amber99sb are added as a series expansion. This effect of this bug is that only the first, more important, term in the expansion is included. The result is that these modified dihedral potentials are in between amber99sb and amber99sb-ildn, most roughly half-way in between, in particular in terms of trans-gauche energy differences. Only the torsion_ASP_N_CA_CB_CG shows a lower gauche energy than both amber99sb and amber99sb-ildn. So I ran a simulation of Ala-Asp-Ala with all three force-fields to check sidechain conformations. The ensemble of amber99sb-ildn with bug ends up in between amber99sb and amber99sb-ildn. So in conclusion, this bugs caused amber99sb-ildn to generate ensembles that are (probably half-way) in between amber99sb and amber99sn-ildn. I have not seen evidence for new states that are not present in either amber99sb or amber99sb-ildn.
A fix has been submitted, but in the meantime pdb2gmx of 4.6 can be used to generate topologies for use with grompp and mdrun of 5.0.