normal PME parameters

  • joshua_roc
  • joshua_roc's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
12 years 1 week ago #932 by joshua_roc
PME parameters was created by joshua_roc
The default value for fourierspacing is 0.12, but I see drastic simulation slowdowns. Does the grid need to be that fine, or do more sparse grids still work well?

Joshua

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

  • joshua_roc
  • joshua_roc's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
12 years 1 week ago #933 by joshua_roc
Replied by joshua_roc on topic PME parameters
As a followup, I've been doing some testing with bilayer systems and salt water boxes.

For large boxes of 100mM salt water (~10,000 waters), you can go as large at 0.3nm for fourierspacing without any change in the contribution of the electrostatics to the energy (for both regular water and polarizeable water).

For highly charged systems (mixed bilayers with some PG headgroups and large micelles of cationic lipopeptides), you can use a fourierspacing of 0.2 without any problems.

Has anyone else done any testing and seen if the grid can be more sparse? There are a number of papers in the literature using PME and MARTINI, but I don't know what grid they are using...

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

More
11 years 10 months ago #940 by xavier
Replied by xavier on topic PME parameters

joshua_roc wrote: As a followup, I've been doing some testing with bilayer systems and salt water boxes.

For large boxes of 100mM salt water (~10,000 waters), you can go as large at 0.3nm for fourierspacing without any change in the contribution of the electrostatics to the energy (for both regular water and polarizeable water).

For highly charged systems (mixed bilayers with some PG headgroups and large micelles of cationic lipopeptides), you can use a fourierspacing of 0.2 without any problems.

Has anyone else done any testing and seen if the grid can be more sparse? There are a number of papers in the literature using PME and MARTINI, but I don't know what grid they are using...


Thanks for reporting this information. It is always very useful.

I am not aware of any systematic study on the effect of the grid size on MARTINI+PME simulations. However we have recently noted very intriguing different behaviors of systems when the electrostatic is treated using PME or simple cutoff scheme as it is generally used.

It is therefore strongly recommended to be extremely cautious when interpreting results of simulations using PME for electrostatics, in particular for charged bilayers.

XAvier.

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

  • joshua_roc
  • joshua_roc's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
11 years 10 months ago #958 by joshua_roc
Replied by joshua_roc on topic PME parameters
Yes. You must be careful with PME and MARTINI.

We decided to start looking at PME effects when we saw salt crystals forming in the water around our bilayer systems. This was because we mistakenly set our PME grid to 1 nm spacing... which has a huge effect on the electrostatics of the system. Things work just fine without PME, but with large systems with lots of charges, we were worried that long-range effects may be important and they aren't accounted for with short range cutoffs... so we decided to test PME.

Still haven't found any papers that mention their PME grid-spacing... I feel like that should be reported if you use PME and MARTINI for a study.

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

More
11 years 10 months ago #959 by xavier
Replied by xavier on topic PME parameters

joshua_roc wrote: Yes. You must be careful with PME and MARTINI.

We decided to start looking at PME effects when we saw salt crystals forming in the water around our bilayer systems. This was because we mistakenly set our PME grid to 1 nm spacing... which has a huge effect on the electrostatics of the system. Things work just fine without PME, but with large systems with lots of charges, we were worried that long-range effects may be important and they aren't accounted for with short range cutoffs... so we decided to test PME.

Still haven't found any papers that mention their PME grid-spacing... I feel like that should be reported if you use PME and MARTINI for a study.


we use a fourierspacing = 0.12 (nm)

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

Time to create page: 0.085 seconds