[gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?

Felipe Pineda, PhD luis.pinedadecastro at lnu.se
Tue Aug 14 09:38:47 CEST 2012

Hi Sébastien,

I found the following paper very instructive about this issue (simulated 
areas per lipid in bilayers):

Jensen, M. et al. Simulations of a membrane anchored peptide: structure, 
dynamics, and influence on bilayer properties. Biophys. J. (2004)86, 3556-75

Take maybe a look at it, if you haven't done it already.



On 08/13/2012 11:12 PM, Peter C. Lai wrote:
> Oh something I didn't mention: for bond constraints I used h-bonds instead
> of all-bonds. This may or may not make a difference (although I switched to
> h-bonds based on the suggestion of some charmm/lipid thread on here from
> a couple of years ago).
> On 2012-08-09 12:34:19PM -0300, Sebastien Cote wrote:
>> Dear Peter,
>> Did you use any different simulation conditions for your POPC membrane? I tried many different ones for POPE, without never reproducing Klauda's results. I may try yours on my POPE membrane.
>> In my simulations, I want to study peptide-membrane interactions. The peptide is not embedded in the membrane. It is initially completely solvated without any interactions with the membrane. Then, I want to look at its adsorption and degree of insertion in the membrane. For that system, I can not remove the CoM motion of the protein alone, otherwise it will not adsorb and insert in the membrane.
>> I may try (as you suggested) to remove CoM of the bottom leaflet on one hand, and the peptide-upperleaflet on the other hand. My peptide is not very long (17 to 35 amino acids), so I believe that remove the CoM of the peptide-upperleaflet/bottomleaflet will not have any pernicious effect. What do you think?
>> Thanks for the suggestion,
>> Sébastien
>> ----------------------------------------
>>> Date: Wed, 8 Aug 2012 20:19:56 -0500
>>> From: pcl at uab.edu
>>> To: gmx-users at gromacs.org
>>> Subject: Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
>>> Personally, I could remove the COM of each leaflet when equilibrating the
>>> bilayer by itself (and as a side note I am not experiencing a similar problem
>>> with POPC that you're having with POPE...). However, after the protein is
>>> embedded, I have gotten good results for my protein, which extends from the
>>> water through the entire membrane into more water, by using a whole System
>>> COM removal. The introduction of my particular embedded protein acts as a
>>> physical coupling between the water layers with the lipids (not to mention if
>>> I choose to model the lipid raft localization crosslink, it will have to
>>> happen anyway). If your protein doesn't extend fully past both layers of the
>>> membrane you may want to stick with just coupling a Membrane+Protein+1 layer
>>> of water or Membrane+Protein and Water separately (like in Justin's KALP15
>>> tutorial). You will have to decide what you think is physically realistic
>>> based on the interaction between the water, membrane, and protein when the
>>> protein is embedded. (if your protein is assymetrically embedded you may even
>>> use the following COM groups: protein+involved leaflet, second leaflet,
>>> water).
>>> On 2012-08-09 09:38:01AM +1000, Mark Abraham wrote:
>>>> On 9/08/2012 3:28 AM, Sebastien Cote wrote:
>>>>> Thanks for the suggestion. I tried it, but for my system the gain is not significant.
>>>>> I was aware that it is preferable to remove the centre-of-mass for each leaflet separately. However, in my tests, I removed the center-of-mass of the membrane because I intent to simulate peptide-membrane interactions. In such case, the center-of-mass of the protein-membrane system is usually removed. Is their any way to remove the CoM motion of each leaflet separately on one hand, and peptide-membrane system CoM motion on the other?
>>>> See 7.3.3 of manual.
>>>> Mark
>>>>> Thanks,
>>>>> Sebastien
>>>>> ----------------------------------------
>>>>>> Date: Fri, 3 Aug 2012 11:10:22 -0400
>>>>>> Subject: Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
>>>>>> From: da294 at cornell.edu
>>>>>> To: gmx-users at gromacs.org
>>>>>> Hello,
>>>>>> I ran into similar issues for a DPPC bilayer. It might be possible
>>>>>> that the two leaflets of the bilayer are moving with respect to
>>>>>> eachother. If this is not taken into account, these artificial
>>>>>> velocities will mean the simulation thinks it is at a higher
>>>>>> temperature than it really is. If possible, you might want to try
>>>>>> subtracting the center of mass motion of each leaflet, rather than the
>>>>>> center of mass motion of the entire bilayer. This will allow the
>>>>>> system to equillibrate to the correct (higher) temperature, and should
>>>>>> increase the area per lipid of the bilayer.
>>>>>> Hope this helps.
>>>>>> -David
>>>>>> On Thu, Aug 2, 2012 at 8:22 AM, Sebastien Cote
>>>>>> <sebastien.cote.4 at umontreal.ca> wrote:
>>>>>>> Dear Gromacs users,
>>>>>>> I did new tests on the POPE membrane with CHARMM36 parameters, but I still always get area per lipid values that are smaller than experimental value by 4 to 6 Angstrom2. Here are my new tests.
>>>>>>> My initial configuration is an equilibrated POPE membrane with 80 lipids at 1 atm and 310K in NPT. It was taken from Klauda's website and it was obtained from the study in which the POPE parameters were tested (Klauda, J. B. et al. 2010 J. Phys. Chem. B, 114, 7830-7843).
>>>>>>> I use TIPS3P (Charmm's special TIP3P). My simulations parameters are similar to those used in a previous tread on the Gromacs mailing list (http://lists.gromacs.org/pipermail/gmx-users/2010-October/055161.html for DMPC, POPC and DPPC of 128 lipids each) :
>>>>>>> dt = 0.002 ps; rlist = 1.0 nm; rlistlong = 1.4 nm; coulombtype = pme; rcoulomb = 1.4 nm; vdwtype = switch or cutoff (see below); DispCorr = No; fourierspacing = 0.15 nm; pme_order = 6; tcoupl = nose-hoover; tau_t = 1.0 ps; ref_t = 310K; pcoupl = Parrinello-Rahman; pcoupltype = semiisotropic; tau_p = 5.0 ps; compressibility = 4.5e-5; ref_p = 1.0 atm; constraints = h-bonds; constraint_algorithm = LINCS. Nochargegrps was used when executing pdb2gmx.
>>>>>>> The simulation time of each simulation is 100 ns. I tried different VdW cutoff values, since it was previously mentioned that cutoff values for VdW may influence the area per lipid. The average value and standard deviation are calculated on the 20 to 100 ns time interval.
>>>>>>> 1- For VdW switch from 0.8 to 1.2 nm : The area per lipid is 54.8 +/- 1.6 A2.
>>>>>>> 2- For VdW switch from 1.1 to 1.2 nm : The area per lipid is 54.6 +/- 1.8 A2.
>>>>>>> 3- For VdW cutoff at 1.4 nm : The area per lipid is 55.9 +/- 1.6 A2.
>>>>>>> I also checked the influence of DispCorr with VdW switch from 0.8 to 1.2 nm :
>>>>>>> 1- Without DispCorr : The area per lipid is 54.8 +/- 1.6 A2.
>>>>>>> 2- With DispCorr : The area per lipid is 54.4 +/- 1.9 A2.
>>>>>>> I also checked the influence of PME cutoff with VdW switch from 0.8 to 1.2 nm :
>>>>>>> 1- For PME cutoff at 1.4 nm : The area per lipid is 54.8 +/- 1.6 A2.
>>>>>>> 2- For PME cutoff at 1.0 nm : The area per lipid is 56.4 +/- 1.5 A2.
>>>>>>> These values are smaller than 4-6 A2 when compared against the experimental value (59.75-60.75 A2) and the value obtained in Klauda's simulation (59.2 +/- 0.3 A2). DispCorr and LJ cutoff weakly impact the results. Reducing the PME cutoff seems to have the greatest effect, but the value obtained is still smaller than experimental value by 3-4 A2.
>>>>>>> I also tried other initial configurations, but the results were either very similar or worst.
>>>>>>> Larger membrane gave similar results for the mean values and smaller standard deviations.
>>>>>>> -------
>>>>>>> Have anyone else tried to simulate a CHARMM36 POPE membrane in Gromacs? Do you get similar results?
>>>>>>> Is a 3-4 A2 deviation from experiment likely to influence my membrane/peptide simulations? Would it then be preferable to go with CHARMM27 in the NPAT ensemble?
>>>>>>> At this point, I have no clue of how to reproduce correctly Klauda's results for POPE. Any suggestion is welcomed.
>>>>>>> Thanks,
>>>>>>> Sebastien
>>>>>>> ----------------------------------------
>>>>>>>> Date: Mon, 23 Jul 2012 16:06:40 -0500
>>>>>>>> From: pcl at uab.edu
>>>>>>>> To: gmx-users at gromacs.org
>>>>>>>> Subject: Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
>>>>>>>> On 2012-07-23 02:34:31PM -0300, Sebastien Cote wrote:
>>>>>>>>> There is not much difference when using DispCorr or not. At least on the same time scale as the simulation with switch cutoff from 0.8 to 1.2 nm and on the same time scale.
>>>>>>>>> Should DispCorr be used in all membrane simulations? I thought that we should always use this correction.
>>>>>>>> I alwasy thought it was actually forcefield dependent. I never use it with
>>>>>>>> CHARMM since the mdp files I used as the basis for mine didn't with C27, and
>>>>>>>> I get acceptable APL with POPC when using the same mdp with C36. I haven't
>>>>>>>> compared the codes for CHARMM to see if dispcorr is builtin to the gromacs
>>>>>>>> implementation or not, but the reason I brought it up is that on past
>>>>>>>> mailing list discussions about TIPS3P, there were reports of significant
>>>>>>>> density differences with and without dispcorr.
>>>>>>>>> Thanks,
>>>>>>>>> Sebastien
>>>>>>>>> ----------------------------------------
>>>>>>>>>> Date: Fri, 20 Jul 2012 12:47:44 -0500
>>>>>>>>>> From: pcl at uab.edu
>>>>>>>>>> To: gmx-users at gromacs.org
>>>>>>>>>> Subject: Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
>>>>>>>>>> Did you play with DispCorr?
>>>>>>>>>> On 2012-07-20 09:46:13AM -0300, Sebastien Cote wrote:
>>>>>>>>>>> Dear Gromacs users,
>>>>>>>>>>> My simulations on a POPE membrane using the CHARMM36 parameters are giving ''area per lipid'' values well below the experimental value (59.75-60.75 Angstroms2). Is their someone else experiencing a similar problem? If yes, how did you solved it?
>>>>>>>>>>> I did the following :
>>>>>>>>>>> I used the CHARMM36 parameters kindly provided by Thomas J. Piggot on the Users contribution section on Gromacs website.
>>>>>>>>>>> My starting configuration was taken from : http://terpconnect.umd.edu/~jbklauda/research/download.html
>>>>>>>>>>> It is a POPE membrane of 80 lipids equilibrated in NPT at T=310K and P=1atm for 40 ns. It is taken from the article Klauda, J. B. et al. 2010 J. Phys. Chem. B, 114, 7830-7843.
>>>>>>>>>>> At first, I tested normal TIP3P vs. CHARMM TIP3P and saw that normal TIP3P gives smaller Area per lipid of about 2-3 Angstroms. This was also observed by T.J. Piggot (personnal communication) and Tieleman (Sapay, N. et al. 2010 J. Comp. Chem. 32, 1400-1410). So, I will present only the simulations using CHARMM TIP3P. As in Klauda's paper, my simulations are at 310K and 1 atm. As them, I used a switch cutoff for vdw, and I used normal cutoff for PME. The simulations are 20 ns. I can send my .mdp file for more details. I varied the switch condition on vdw :
>>>>>>>>>>> 1- For a switch from 0.8 to 1.2 (as in Klauda's paper), I got Area per lipid of about 56.5 Angstroms2; whereas they got 59.2 in their paper, matching the experimental value of 59.75-60.75.
>>>>>>>>>>> 2- For a switch from 1.0 to 1.2, I got Area per lipid of about 53.5 Angstroms2, which is smaller than the previous cutoff. This is surprising since a previous thread on gromacs-users mailing lists said that increasing the lower cutoff, increased the Area per lipid or had not impact on POPC of DPPC.
>>>>>>>>>>> 3- For a switch from 1.1 to 1.2, I got Area per lipid of about 55 Angstroms2.
>>>>>>>>>>> 4- For a hard cutoff at 1.4, I got Area per lipid of about 52 Angstroms2.
>>>>>>>>>>> I also tried to re-equilibrate the membrane in the NPAT ensemble for 10 ns at 310K and 1 atm. Then, when I launched the simulation in NPT, I ended up with different results :
>>>>>>>>>>> 1- Switch from 0.8 to 1.2 gave a smaller area per lipid of 54 Angstroms2.
>>>>>>>>>>> 2- Switch from 1.0 to 1.2 gave a larger area per lipid of 55 Angstroms2.
>>>>>>>>>>> 4- Hard cutoff at 1.4 gave a similar area per lipid of 52.5 Angstroms2.
>>>>>>>>>>> I looked at the POPE paramaters for CHARMM36 in Gromacs, and they agree with the published parameters.
>>>>>>>>>>> Am I doing anything wrong? Is their someone else experiencing a similar problem for POPE? If yes, how did you solved it?
>>>>>>>>>>> Should I instead use CHARMM27 parameters in the NPAT ensemble? I want to study the interaction between a peptide and the POPE membrane. I am troubled that the NPAT ensemble might influence my results in a bad way. Also, I can not use OPLS AA nor GROMOS for the protein interactions because these force fields are not giving the correct structural ensemble for my peptide in solution.
>>>>>>>>>>> I am willing to send more information if you need.
>>>>>>>>>>> Thanks a lot,
>>>>>>>>>>> Sincerely,
>>>>>>>>>>> Sébastien --

More information about the gromacs.org_gmx-users mailing list