normal GPU - Verlet - oscillational period

  • Hegedus
  • Hegedus's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
6 years 9 months ago #7302 by Hegedus
GPU - Verlet - oscillational period was created by Hegedus
Hi,

I run CG simulations (protein in a bilayer) on GPU with the suggested mdp options for the Verlet scheme. I run the emin and equilibration according to the charmm-gui scheme. E.g. starting equilibration with dt=0.002 and increasing in 5 steps to 0.02 (several steps for gradually closing the hole having for flip-flop in the first equilibration step).

However, when dt becomes 0.01 I get this NOTE:
"The bond in molecule-type Protein_A between atoms 1 BB and 3 BB has an
estimated oscillational period of 9.7e-02 ps, which is less than 10 times
the time step of 1.0e-02 ps.
Maybe you forgot to change the constraints mdp option."

My questions:

1. In the sample mdp file a dt between 0.02 and 0.04 is suggested. So I do not expect to have this note.

2. This happens only for my system if I run with elnedyn22, but not with martini22. I would like to understand why.

3. Can I ignore this note and worry only if it turns into a warning?
From gromacs manual: "a Verlet (leap-frog) integration scheme a minimum of 5 numerical integration steps should be performed per period of a harmonic oscillation in order to integrate it with reasonable accuracy."

Thanks for your help and suggestion,
Tamas

Please Log in or Create an account to join the conversation.

More
6 years 9 months ago #7303 by peterkroon
Replied by peterkroon on topic GPU - Verlet - oscillational period
Hi Tamas,

what is the force constant of the bond in question?

Peter

Please Log in or Create an account to join the conversation.

  • Hegedus
  • Hegedus's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
6 years 9 months ago #7304 by Hegedus
Replied by Hegedus on topic GPU - Verlet - oscillational period
[ atoms ]
1 P4 1 PRO BB 1 0.0000 ; C
2 C3 1 PRO SC1 2 0.0000 ; C
3 P5 2 LEU BB 3 0.0000 ; C
4 C1 2 LEU SC1 4 0.0000 ; C
5 P5 3 GLU BB 5 0.0000 ; C
6 Qa 3 GLU SC1 6 -1.0000 ; C

[ bonds ]
; Backbone bonds
1 3 1 0.38286 150000 ; PRO(C)-LEU(C)
3 5 1 0.38410 150000 ; LEU(C)-GLU(C)

[ angles ]
; Backbone angles
1 3 5 2 109.75582 40 ; PRO(C)-LEU(C)-GLU(C)
3 5 7 2 126.98538 40 ; LEU(C)-GLU(C)-LYS(C)

MARTINI22:
[ atoms ]
1 P4 1 PRO BB 1 0.0000 ; C
2 C3 1 PRO SC1 2 0.0000 ; C
3 P5 2 LEU BB 3 0.0000 ; C
4 AC1 2 LEU SC1 4 0.0000 ; C
5 P5 3 GLU BB 5 0.0000 ; C
6 Qa 3 GLU SC1 6 -1.0000 ; C

[ bonds ]
; Backbone bonds
1 3 1 0.35000 1250 ; PRO(C)-LEU(C)
3 5 1 0.35000 1250 ; LEU(C)-GLU(C)

[ angles ]
; Backbone angles
1 3 5 2 127 20 ; PRO(C)-LEU(C)-GLU(C)
3 5 7 2 127 20 ; LEU(C)-GLU(C)-LYS(C)

Please Log in or Create an account to join the conversation.

More
6 years 9 months ago #7305 by peterkroon
Replied by peterkroon on topic GPU - Verlet - oscillational period
So you have force constants of 150k. This is way too high, where did you get those numbers?
If you're convinced they're correct, change them to constraints.

Peter

Please Log in or Create an account to join the conversation.

  • Hegedus
  • Hegedus's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
6 years 9 months ago #7306 by Hegedus
Replied by Hegedus on topic GPU - Verlet - oscillational period
I am not convinced at all that those high numbers are correct. At least it is not intentional. I simply used martinize.py.

martinize.py -f $(NAME).pdb -o system-vaccum.top -x $(NAME)-CG.pdb -dssp /usr/bin/dssp -p backbone -ff elnedyn22 -nt

INFO MARTINIZE, script version 2.6

INFO The elnedyn forcefield has been implemented with some changes compared to the published parameters:
INFO - Backbone-Backbone bonds use high force constant bonds instead of constraints.

Thanks.

Please Log in or Create an account to join the conversation.

More
6 years 9 months ago #7307 by peterkroon
Replied by peterkroon on topic GPU - Verlet - oscillational period
Hmmm. If it explicitly says this (because this is what you see), I think you can ignore the grompp note. For such a short protein you can also safely turn it into a constraint, but that will become unstable for larger proteins.

Peter

Please Log in or Create an account to join the conversation.

  • Hegedus
  • Hegedus's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
6 years 9 months ago #7308 by Hegedus
Replied by Hegedus on topic GPU - Verlet - oscillational period
Thanks!

Please Log in or Create an account to join the conversation.

Time to create page: 0.107 seconds