normal Some questions about martini_v2.x_example.mdp

  • zhiyue
  • zhiyue's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
9 years 9 months ago - 9 years 9 months ago #3886 by zhiyue
Hi all,

I have some questions about mdrun settings in martini_v2.x_example.mdp and martini_v2.P_example.mdp on MARTINI website.

1. nstcomm
"nstcomm" is set 1 in both *.mdp. As the default value for "nstcalcenergy" is 100, "nstcomm < nstcalcenergy defeats the purpose of nstcalcenergy". How can I deal with it?

2. rlist
rlist has been set 1.2 nm, which is the same as rcoulomb. As GROMACS suggests "rlist should be 0.1 to 0.3 nm larger than rcoulomb", it's better to use 1.4 or 1.5 nm. And this is what has been suggested in *.mdp files. However, seems that 1.2 nm has been consistently used in the papers published by MARTINI people. I'm wondering whether there exists any good reason to ignore this GROMACS criterion.

3. electrostatics
I know MARTINI force field has been parameterized by using Shift. But the *.mdp files suggest PME is more realistic when polarizable waters are used. In the paper where PW was released (Yesylevskyy, PLoS Comput. Biol., 2010), it has been tested to show that both PME and Shift schemes can be used equally well for polarizable MARTINI waters. Then my understanding of "less realistic" by using Shift scheme refers to the ignorance of long-reange electrostatics. Correct me if I was wrong. Another pertinent question is has the PW+PME been ever tested with any protein/membrane system?

4. Berendsen coupling
I still don't understand the rationale of choosing Berendsen coupling for T/P controls in MARTINI. As Berendsen does not produce canonical ensemble, it is not recommended in general. As I understand, it may not be a big problem for MARTINI use as the deivation from the real system caused by coarse-grained description may be larger than what Berendsen does. But I still want to know the rationale, if there is any, of choosing Berendsen coupling when MARTINI was developed.

5. other coupling schemes
I notice in martini_v2.x_example.mdp v-rescale and parrinello-rahman are used to control T and P, respectively. Have they ever been tested? If yes, how would MARTINI people evalualte their performance compared to Berendsen?

6. Berendsen thermostat for PW
The polarizable MARTINI water model was developed with the time constant for Berendsen thermostat of 0.3 ps (Yesylevskyy, PLoS Comput. Biol., 2010). Then my question is which time constant should I use, 0.3 or 1.0, if PW is used?

7. Notes & Warnings
It happens that running grompp gives "Notes" and "Warnings". As a rigorous GROMACS user, these information should never been overlooked. One never knows whether any "noted" issues will crash future MD run or whether the simulation is trustworthy. As for "Warnings", seems for MARTINI sometimes we can ignore them by setting "maxwarn" to an appropriate number. I noticed someone skipped these "Warnings" just by increasing "maxwarn" numbers, which is obviously not right. So as an inexperienced MARTINI user, I want to know when I can ignore or skip these "Notes" and/or "Warnings". Here are some examples.

Notes:
<1>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The Berendsen thermostat does not generate the correct kinetic energy
distribution. You might want to consider using the V-rescale thermostat.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
<2>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The sum of the two largest charge group radii (0.208146) is larger than
rlist (1.400000) - rcoulomb (1.200000)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Warnings:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
You are using pressure coupling with absolute position restraints, this
will give artifacts. Use the refcoord_scaling option.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

If it's better to overcome them rather than ignore them. How can I deal with them?


Thanks in advance.

All the best

Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.
Last edit: 9 years 9 months ago by zhiyue.

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

More
9 years 9 months ago #3892 by djurre

zhiyue wrote: Hi all,

I have some questions about mdrun settings in martini_v2.x_example.mdp and martini_v2.P_example.mdp on MARTINI website.


That's a lot of questions. Let me try to answer them one by one. By the way, many of the issues have already been changed in the current example .mdp file that you can find here .

1. nstcomm

nstcomm is already 10 in the new file, but in my experience get be set to 100 or even 1000 without any COM drift.

2. rlist
I think in newer publication people use 1.4 nm, as is also used in the new example-mdp.

3. electrostatics
Neglecting the long range electrostatics is less realistic, however the average Martini force field has very little charges. Using PME would come at a considerable cost, while it would make very little difference. (You're mainly calculating fourier transforms over empty space). This is of course different for highly charged molecules, like antimicrobial peptides. That also answers the other question: for a membrane/peptide system it has been tested/used: Rzepiela et al (2010) Faraday Discussions 144.

4. Berendsen coupling
I can't say what the reason was when the force field was first developed, but my guess would be that the developers were more familiar with this thermostat.

5. other coupling schemes
I've tested it and found minimal differences in the free energy. The simulations are not noticeably slower than with berendsen. I have the feeling that for marginally stable simulations, Berendsen gives slightly better results. But I'm not sure there.

6. Berendsen thermostat for PW
I don't know. Run a simulation with both and see what the difference is. Could you report back to the forum what you find?

7. Notes & Warnings
Keep in mind that notes and warning were not always there. The fact that there is a warning about something in version B doesn't make it more wrong than it was in version A.

zhiyue wrote: <1>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The Berendsen thermostat does not generate the correct kinetic energy
distribution. You might want to consider using the V-rescale thermostat.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


It is probably better to use the Berendsen thermostat.

zhiyue wrote: <2>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The sum of the two largest charge group radii (0.208146) is larger than
rlist (1.400000) - rcoulomb (1.200000)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


In principle the Martini forcefield doesn't need charge groups, since it uses a shifted potential. So you could give every particle it's own charge group, although this would (theoretically) make the neighbour searching less efficient. Notice that the differences in very small in your example, so you could also set the rlist to 1.409.

zhiyue wrote: Warnings:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
You are using pressure coupling with absolute position restraints, this
will give artifacts. Use the refcoord_scaling option.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~


You should use the refcoord_scaling option as proposed.

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

  • zhiyue
  • zhiyue's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
9 years 9 months ago #3895 by zhiyue
Hi djurre,

Thanks for your time for having answered my questions. I really appreciate it!

As for choice of Berendsen thermostat time constant, I'll do some test. I'll post my results below once done. Thanks!

Best

Shane
School of Pharmacy
University of Maryland
Baltimore, MD 21201, U.S.A.

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

More
8 years 11 months ago #4506 by Rajat Desikan
Replied by Rajat Desikan on topic Some questions about martini_v2.x_example.mdp
Hi,
Regarding rlist, one would be forced to use rlist=1.2 with PME, right? It is not possible to use rlist=1.4 with PME (rcoulomb=1.2).

Thanks,

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

Time to create page: 0.087 seconds