I've reproduced your problem, and re-run with "torsions 1" which will
activate fitting for all torsions. There are no torsions involving the (H1
type) HB2 and HB3 atoms, either. In fact, the report shows that the only
torsion terms in the system involve the backbone. Likely what's happened
is that the original parm.dat / frcmod files do not have any torsion terms
for the CYM side chain. The original parm10.dat has only an X -CX-CT-X
parameter that might be of general applicability, but it seems that it does
not get applied in your file. The frcmod.ff14SB file has these dihedrals,
which are what I think you want to target for parameterization:
C -CX-2C-SH 1 0.075 0.0 -4. C
C -CX-2C-SH 1 0.251 0.0 -3.
C -CX-2C-SH 1 0.337 180.0 -2.
C -CX-2C-SH 1 0.269 180.0 1.
N -CX-2C-SH 1 0.033 0.0 -4.
N -CX-2C-SH 1 0.251 0.0 -3.
N -CX-2C-SH 1 0.486 180.0 -2.
N -CX-2C-SH 1 0.154 0.0 1.
CX-2C-SH-HS 1 0.030 0.0 -4.
CX-2C-SH-HS 1 0.252 0.0 -3.
CX-2C-SH-HS 1 0.612 0.0 -2.
CX-2C-SH-HS 1 0.092 0.0 1.
Those hinge (literally) on the 2C atom type, which replaced CT as the beta
carbon type for some amino acids in ff14SB (see also type 3C). So my
thinking is that you need to look back at the way this prmtop was
constructed, because mdgx has given you a clue that leads to the fact that
the molecule you have made has literally no side chain torsions at all.
This is a feature (detecting missing dihedrals that should be present to
describe some twist) that would be much easier to code in my new C++
framework, and is meticulous enough that unit testing with numerous hazard
cases would have to be implemented. But that's the best answer I can give
you for now: something prior to your mdgx step has created a molecule with
no torsions at all. Re-rolling with a 2C beta carbon atom type will
probably fix your problem by helping tleap to implement the torsions in
frcmod.ff14SB. You will then have a substrate to optimize.
Dave
On Tue, Apr 5, 2022 at 7:15 PM David Cerutti <dscerutti.gmail.com> wrote:
> I am re-reading the email from Yigitcan and I'm not sure why the program
> would not detect the SH CT CX C torsion. It may be that, while all of
> these atom types are present in the topology, the parm.dat and frcmod files
> contain SH-CT bonds and SH-CT-CX angles but no examples of the SH-CT-CX-C
> torsion itself. One thing to try might be to look at the HB2 / HB3
> hydrogen atom types and see if the torsions between C CX CT and [ HB2 / HB3
> type ] can be parameterized. This is not a general solution, but it would
> shed light on the problem.
>
> Parameter fitting is a tough problem, and if I could retire mdgx in favor
> of a different implementation, I would. The plain C framework mdgx is
> written in doesn't permit the ease of error checking that I would like, and
> there is so much that can go wrong in the parameter files, which are
> probably the least bulletproofed of the Amber file formats. There are only
> a handful of programs that read the things and most of them are used by
> only a few researchers in one work group. In time the new code base I am
> developing will be on its feet, and I may be able to take my "google 20%"
> to swing back to this parameter fitting problem in a new and more powerful
> way. If I could wave a magic wand, mdgx would have a few hundred unit
> tests for just the parameter fitting module, but this is what I've learned
> in the ten years since I started down that path.
>
> Dave
>
>
> On Sat, Apr 2, 2022 at 12:20 AM Erdem Yeler <erdemyeler.gmail.com> wrote:
>
>> >
>> >
>> > Personally, I couldn't use mdgx because of the same problem. I changed
>> > amber atom masks and tried new ways again and again, but I couldn't.
>> > Then I wrote my own script to generate dihedrals. Perhaps you should
>> > contact the mdgx developer (D.S. Cerutti) directly. I am sure he will
>> > answer.
>> >
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Apr 05 2022 - 17:30:02 PDT
