- Posts: 89
Simulation of CG GPCR-membrane system
- James Starlight
-
Topic Author
- Visitor
8 years 1 month ago #5441
by James Starlight
Simulation of CG GPCR-membrane system was created by James Starlight
Dear MARTINI users!
I would like to discuss the proposed within the tutorial work-flow for the equilibration of membrane protein system using Martini force field. In particular regarding the equilibration 1- I have not found in the mdp file
2- Regarding restrains in the system during the equilibration
which is usually good for all-atom simulation providing the stability of the system during the equilibration. Is it required for the coarse'graining md?
3- Is it possible using insane.py script to use full atomic water e.g from the martini2.2p force field7
Thanks in advance!
Gleb
I would like to discuss the proposed within the tutorial work-flow for the equilibration of membrane protein system using Martini force field. In particular regarding the equilibration 1- I have not found in the mdp file
which is always required for the equilibration of the system after minimization. In my case I turn that flag on and grompp preposed not to use parrinello barostat which usually is not stable during the equilibration. However in my case there was no problems (I also restrained protein during the equilibration to freeze it conformation)gen_vel = yes
2- Regarding restrains in the system during the equilibration
; Centre of mass (COM) motion removal relative to the specified groups
nstcomm = 1 ; COM removal frequency (steps)
comm_mode = Linear ; Remove COM translation (linear / angular / no)
comm_grps = Protein_Lipids Solvent ; COM removal relative to the specified groups
which is usually good for all-atom simulation providing the stability of the system during the equilibration. Is it required for the coarse'graining md?
3- Is it possible using insane.py script to use full atomic water e.g from the martini2.2p force field7
Thanks in advance!
Gleb
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
8 years 3 weeks ago #5455
by mnmelo
Replied by mnmelo on topic Simulation of CG GPCR-membrane system
Hi again,
1- gen_vel will cause your particles to be assigned random velocities according to the Maxwell-Boltzmann distribution for the chosen temperature. Leaving it out will cause particles to retain their current velocities.
If you're proceeding from a minimization there won't be any velocities, which means that leaving out gen_vel causes particles to start with zero velocity (or an 'instantaneous' temperature of 0K). In this case your thermostat will still slowly bring your degrees of freedom up to the desired temperature (and even without thermostat a part of the potential energy will still be converted to kinetic). As far as stability goes, I'd expect that starting out from v=0 will be more stable than assigning random velocities, but I know no citation to give you. I'd also expect this to be really of relevance only if you're close to some sort of singularity, which a proper minimization should take care of anyway.
When you use gen_vel GROMACS uses it to infer that you're coming from a minimization, and therefore warn you against potentially diverging thermostats such as Parrinello-Rahman. If it works for you, then no problem, but more sensitive systems will need a couple hundred ps thermal equilibration with something robust like the Berendsen thermostat before you can use Parrinello-Rahman. Starting out without velocities will also have the consequence that thermal equilibrium will take longer to reach, and also make it more likely that Parinello-Rahman diverges if used from the beginning.
2- Of those lines only nstcomm = 1 is overkill. Doing it every 100 or even 1000 steps is more than enough (I think GROMACS overrides it for you anyway because it's more frequent that nstcalcenergy).
The other two settings are important when having phase-separated systems with potentially weak inter-transfer of heat (as is the case of a membrane-water system). Having an embedded protein that sticks out might mitigate this problem. See my reply to a similar question .
Good luck,
Manel
1- gen_vel will cause your particles to be assigned random velocities according to the Maxwell-Boltzmann distribution for the chosen temperature. Leaving it out will cause particles to retain their current velocities.
If you're proceeding from a minimization there won't be any velocities, which means that leaving out gen_vel causes particles to start with zero velocity (or an 'instantaneous' temperature of 0K). In this case your thermostat will still slowly bring your degrees of freedom up to the desired temperature (and even without thermostat a part of the potential energy will still be converted to kinetic). As far as stability goes, I'd expect that starting out from v=0 will be more stable than assigning random velocities, but I know no citation to give you. I'd also expect this to be really of relevance only if you're close to some sort of singularity, which a proper minimization should take care of anyway.
When you use gen_vel GROMACS uses it to infer that you're coming from a minimization, and therefore warn you against potentially diverging thermostats such as Parrinello-Rahman. If it works for you, then no problem, but more sensitive systems will need a couple hundred ps thermal equilibration with something robust like the Berendsen thermostat before you can use Parrinello-Rahman. Starting out without velocities will also have the consequence that thermal equilibrium will take longer to reach, and also make it more likely that Parinello-Rahman diverges if used from the beginning.
2- Of those lines only nstcomm = 1 is overkill. Doing it every 100 or even 1000 steps is more than enough (I think GROMACS overrides it for you anyway because it's more frequent that nstcalcenergy).
The other two settings are important when having phase-separated systems with potentially weak inter-transfer of heat (as is the case of a membrane-water system). Having an embedded protein that sticks out might mitigate this problem. See my reply to a similar question .
Good luck,
Manel
Please Log in or Create an account to join the conversation.
- James Starlight
-
Topic Author
- Visitor
8 years 4 days ago #5491
by James Starlight
Replied by James Starlight on topic Simulation of CG GPCR-membrane system
Thanks so much for the detailed comments!
What I actually have done with my MARTINI membrane-protein simulations by now is
1 - minimize system
2- start NPT equilibration (protein is restrained fully) with gen_vel = YES
3- start production run from previous step with gen_vel = No providing cpt file for this run with velocities.
I used Berendsen thermostat (gromppt send me NOTE that simulation might be not stable and propose to use V-rescale) for all runs and default martini params only adding -rdd 1.5 flag to mdrun and generally it was no problems with the MD.
James
What I actually have done with my MARTINI membrane-protein simulations by now is
1 - minimize system
2- start NPT equilibration (protein is restrained fully) with gen_vel = YES
3- start production run from previous step with gen_vel = No providing cpt file for this run with velocities.
I used Berendsen thermostat (gromppt send me NOTE that simulation might be not stable and propose to use V-rescale) for all runs and default martini params only adding -rdd 1.5 flag to mdrun and generally it was no problems with the MD.
James
Please Log in or Create an account to join the conversation.
Time to create page: 0.102 seconds