Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Hi Sébastian, I think the magic word in this issue would be surface tension and the proper ensemble for the simulation NPgammaT. This is very well discussed in the paper I advised to you a couple of days ago. The issue is by no means trivial, although I'm not an expert to judge it. You can find an imho very well-founded theoretical discussion in, e.g., Lindahl, E. and Edholm, O. Spatial and energetic-entropic decomposition of surface tension in lipid bilayers from molecular dynamics simulations. J. Chem. Phys.(2000)113, 3882. Good luck! Felipe On 08/15/2012 08:23 PM, Sebastien Cote wrote: Thanks for the advices Chris. My peptide is known to be more favorably to PE than PC membrane that is why I am using POPE. Experimentally, the liquid phase transition is at 298K for POPE (if I am not mistaken). Is your 323K refer to some simulations? At first I wanted to use the new CHARMM36 lipids parameters because they are supposed to solve the previous CHARMM27 issue with the area per lipid. However, I am consistently obtained smaller APL then experiment and I am not able to reproduce the published APL obtained for POPE, even if I am starting from their equilibrated 80-POPE membrane and use same simulation conditions. That was the reason for starting this thread on the mailing list. Unfortunately, my peptide conformational space in solution is only well-represented by CHARMM27 (equivalently in CHARMM36), so I can not use Berger's lipid parameters with OPLS or GROMOS even if it would be preferable as they do not have APL inconsistency and are united-atom. I will made some tests in the NPAT ensemble. Perhaps the NPAT effects can be made neglegible by using bigger membrane compared to my peptide's size (?). Sebastien From: chris.ne...@mail.utoronto.ca To: gmx-users@gromacs.org Date: Wed, 15 Aug 2012 17:29:29 + Subject: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why? The area per lipid (APL) will certainly affect the free energy of peptide/protein binding to a lipid bilayer. I have not used charmm lipids extensively, but from what I understand they older charmm lipids required NPAT to get the correct APL. The newer charmm lipids were supposed to solve that problem, but I have heard it said that, though the problem has been alleviated to some extend, it still remains. If I were you, I'd use POPC in place of POPE. POPE is notorious for giving too-small APL's in simulations and I think it even requires temperatures of 323 K to enter the liquid phase. That said, I don't have a specific answer to your question of whether there are other affects of NPAT vs. NPT. It is plausible that NPAT-based fluctuations could affect the pathway or the kinetics. PS: I was not referring to lipid rafts, but the separate diffusion of the upper and lower leaflets. Once the peptide is fully inserted, if it spans both leaflets, this will tend to reduce this leaflet-specific diffusion and would represent an entropic penalty for binding (not sure how large). Chris. Dear Peter, I also used h-bonds and I also switch LJ interaction from 0.8 nm to 1.2 nm (as in Klauda's paper). I will retry with a more solvated membrane. Would you have any thought on how the NPAT ensemble might affect peptide-membrane interactions like I am studying i.e. peptide is totally solvated, then adsorb, and finally may insert? The paper on peptide-membrane interaction like this usually use united-atom lipid in the NPT ensemble. Most of the work I have seen on Charmm membrane in the NPAT ensemble were for embedded membrane protein. Sorry, but I only have experience with large pre-embedded membrane proteins, and those are governed both by signal sequences and post-translational modification. Chris's last email on the subject might lead to the hypothesis that lipid raft translation as the leaflets slide past one another could be a contributing factor to adsorbption of your species. Thanks, Sebastien -- gmx-users mailing list gmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- +---+ | Luis Felipe Pineda De Castro, PhD
[gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Dear David: If the two leaflets are moving with respect to each other (along the bilayer plane), then why would this be artificial? I have also seen this (diffusive) motion, and in addition to wondering why you would call it artificial, it seems to me that the motion would have to be many orders of magnitude larger than is observed in simulations for it to affect the temperature. Subtracting the COM of each leaflet separately is not only impossible, but inadvisable. Chris. -- original message -- 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 -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
On 2012-08-14 08:52:26PM -0300, Sebastien Cote wrote: Dear Peter, I also used h-bonds and I also switch LJ interaction from 0.8 nm to 1.2 nm (as in Klauda's paper). I will retry with a more solvated membrane. Would you have any thought on how the NPAT ensemble might affect peptide-membrane interactions like I am studying i.e. peptide is totally solvated, then adsorb, and finally may insert? The paper on peptide-membrane interaction like this usually use united-atom lipid in the NPT ensemble. Most of the work I have seen on Charmm membrane in the NPAT ensemble were for embedded membrane protein. Sorry, but I only have experience with large pre-embedded membrane proteins, and those are governed both by signal sequences and post-translational modification. Chris's last email on the subject might lead to the hypothesis that lipid raft translation as the leaflets slide past one another could be a contributing factor to adsorbption of your species. Thanks, Sebastien Date: Mon, 13 Aug 2012 16:12:29 -0500 From: p...@uab.edu To: gmx-users@gromacs.org Subject: Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why? 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: p...@uab.edu To: gmx-users@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
[gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
The area per lipid (APL) will certainly affect the free energy of peptide/protein binding to a lipid bilayer. I have not used charmm lipids extensively, but from what I understand they older charmm lipids required NPAT to get the correct APL. The newer charmm lipids were supposed to solve that problem, but I have heard it said that, though the problem has been alleviated to some extend, it still remains. If I were you, I'd use POPC in place of POPE. POPE is notorious for giving too-small APL's in simulations and I think it even requires temperatures of 323 K to enter the liquid phase. That said, I don't have a specific answer to your question of whether there are other affects of NPAT vs. NPT. It is plausible that NPAT-based fluctuations could affect the pathway or the kinetics. PS: I was not referring to lipid rafts, but the separate diffusion of the upper and lower leaflets. Once the peptide is fully inserted, if it spans both leaflets, this will tend to reduce this leaflet-specific diffusion and would represent an entropic penalty for binding (not sure how large). Chris. Dear Peter, I also used h-bonds and I also switch LJ interaction from 0.8 nm to 1.2 nm (as in Klauda's paper). I will retry with a more solvated membrane. Would you have any thought on how the NPAT ensemble might affect peptide-membrane interactions like I am studying i.e. peptide is totally solvated, then adsorb, and finally may insert? The paper on peptide-membrane interaction like this usually use united-atom lipid in the NPT ensemble. Most of the work I have seen on Charmm membrane in the NPAT ensemble were for embedded membrane protein. Sorry, but I only have experience with large pre-embedded membrane proteins, and those are governed both by signal sequences and post-translational modification. Chris's last email on the subject might lead to the hypothesis that lipid raft translation as the leaflets slide past one another could be a contributing factor to adsorbption of your species. Thanks, Sebastien -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
RE: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Thanks for the advices Chris. My peptide is known to be more favorably to PE than PC membrane that is why I am using POPE. Experimentally, the liquid phase transition is at 298K for POPE (if I am not mistaken). Is your 323K refer to some simulations? At first I wanted to use the new CHARMM36 lipids parameters because they are supposed to solve the previous CHARMM27 issue with the area per lipid. However, I am consistently obtained smaller APL then experiment and I am not able to reproduce the published APL obtained for POPE, even if I am starting from their equilibrated 80-POPE membrane and use same simulation conditions. That was the reason for starting this thread on the mailing list. Unfortunately, my peptide conformational space in solution is only well-represented by CHARMM27 (equivalently in CHARMM36), so I can not use Berger's lipid parameters with OPLS or GROMOS even if it would be preferable as they do not have APL inconsistency and are united-atom. I will made some tests in the NPAT ensemble. Perhaps the NPAT effects can be made neglegible by using bigger membrane compared to my peptide's size (?). Sebastien From: chris.ne...@mail.utoronto.ca To: gmx-users@gromacs.org Date: Wed, 15 Aug 2012 17:29:29 + Subject: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why? The area per lipid (APL) will certainly affect the free energy of peptide/protein binding to a lipid bilayer. I have not used charmm lipids extensively, but from what I understand they older charmm lipids required NPAT to get the correct APL. The newer charmm lipids were supposed to solve that problem, but I have heard it said that, though the problem has been alleviated to some extend, it still remains. If I were you, I'd use POPC in place of POPE. POPE is notorious for giving too-small APL's in simulations and I think it even requires temperatures of 323 K to enter the liquid phase. That said, I don't have a specific answer to your question of whether there are other affects of NPAT vs. NPT. It is plausible that NPAT-based fluctuations could affect the pathway or the kinetics. PS: I was not referring to lipid rafts, but the separate diffusion of the upper and lower leaflets. Once the peptide is fully inserted, if it spans both leaflets, this will tend to reduce this leaflet-specific diffusion and would represent an entropic penalty for binding (not sure how large). Chris. Dear Peter, I also used h-bonds and I also switch LJ interaction from 0.8 nm to 1.2 nm (as in Klauda's paper). I will retry with a more solvated membrane. Would you have any thought on how the NPAT ensemble might affect peptide-membrane interactions like I am studying i.e. peptide is totally solvated, then adsorb, and finally may insert? The paper on peptide-membrane interaction like this usually use united-atom lipid in the NPT ensemble. Most of the work I have seen on Charmm membrane in the NPAT ensemble were for embedded membrane protein. Sorry, but I only have experience with large pre-embedded membrane proteins, and those are governed both by signal sequences and post-translational modification. Chris's last email on the subject might lead to the hypothesis that lipid raft translation as the leaflets slide past one another could be a contributing factor to adsorbption of your species. Thanks, Sebastien -- gmx-users mailing list gmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
[gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Write the authors of the simulation paper that has a correct APL for POPE and ask them for an input file. That is really the only way to be sure that you are not doing something different than they did. In my experience, people are quite willing to provide you with their input file(s). If you still get a different APL than they reported, then see if your simulation times are similar and repeat your run a few times to see if it's just statistical noise. Regarding 323 K, I don't recall... it's just a number that sticks in my head. Perhaps it is for DPPE or DPPC. I'd still suggest that you at least try POPC. So your peptide binds more favourably to POPE than to POPC... that alone does not limit you to POPE. Then again, I don;t know exactly what you are trying to do. Chris. -- original message -- My peptide is known to be more favorably to PE than PC membrane that is why I am using POPE. Experimentally, the liquid phase transition is at 298K for POPE (if I am not mistaken). Is your 323K refer to some simulations? At first I wanted to use the new CHARMM36 lipids parameters because they are supposed to solve the previous CHARMM27 issue with the area per lipid. However, I am consistently obtained smaller APL then experiment and I am not able to reproduce the published APL obtained for POPE, even if I am starting from their equilibrated 80-POPE membrane and use same simulation conditions. That was the reason for starting this thread on the mailing list. Unfortunately, my peptide conformational space in solution is only well-represented by CHARMM27 (equivalently in CHARMM36), so I can not use Berger's lipid parameters with OPLS or GROMOS even if it would be preferable as they do not have APL inconsistency and are united-atom. I will made some tests in the NPAT ensemble. Perhaps the NPAT effects can be made neglegible by using bigger membrane compared to my peptide's size (?). Sebastien -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
On 2012-08-15 06:55:59PM +, Christopher Neale wrote: Write the authors of the simulation paper that has a correct APL for POPE and ask them for an input file. That is really the only way to be sure that you are not doing something different than they did. In my experience, people are quite willing to provide you with their input file(s). If you still get a different APL than they reported, then see if your simulation times are similar and repeat your run a few times to see if it's just statistical noise. The fundamental problem Sebastian will have is that Klauda obtained their APLs using CHARMM software, and he is trying to reproduce this using the forcefield in Gromacs software. So even if the CHARMM input files were provided, it maybe difficult to exactly reproduce the conditions in Gromacs (if certain parameters were implemented differently) Regarding 323 K, I don't recall... it's just a number that sticks in my head. Perhaps it is for DPPE or DPPC. I'd still suggest that you at least try POPC. So your peptide binds more favourably to POPE than to POPC... that alone does not limit you to POPE. Then again, I don;t know exactly what you are trying to do. Chris. It is generally a good idea to use a higher temp than the phase transition temperature, since during equilibration close to the phase transition temp there is a risk of inducing some ordering due to uneven heating. People run DPPC at 323 because its phase transition temp is 315K. If POPC's is 271 and people typically run POPC at 300, then it may be wise to bump up the running temp of a POPE system. Of course, your APL will inflate at higher temperatures... -- original message -- My peptide is known to be more favorably to PE than PC membrane that is why I am using POPE. Experimentally, the liquid phase transition is at 298K for POPE (if I am not mistaken). Is your 323K refer to some simulations? At first I wanted to use the new CHARMM36 lipids parameters because they are supposed to solve the previous CHARMM27 issue with the area per lipid. However, I am consistently obtained smaller APL then experiment and I am not able to reproduce the published APL obtained for POPE, even if I am starting from their equilibrated 80-POPE membrane and use same simulation conditions. That was the reason for starting this thread on the mailing list. Unfortunately, my peptide conformational space in solution is only well-represented by CHARMM27 (equivalently in CHARMM36), so I can not use Berger's lipid parameters with OPLS or GROMOS even if it would be preferable as they do not have APL inconsistency and are united-atom. I will made some tests in the NPAT ensemble. Perhaps the NPAT effects can be made neglegible by using bigger membrane compared to my peptide's size (?). Sebastien -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- == Peter C. Lai| University of Alabama-Birmingham Programmer/Analyst | KAUL 752A Genetics, Div. of Research | 705 South 20th Street p...@uab.edu| Birmingham AL 35294-4461 (205) 690-0808 | == -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Hi, As I suggested earlier in this thread, I think the original poster should run test simulations of a CHARMM36 POPE membrane using either the NAMD or CHARMM softwares. It has been mentioned a couple of times in the thread thay there are differences in the implementations of the switching methods for the van der Waals interactions between softwares, and in my opinion this is one potential cause of the low area per lipid for the POPE membrane. This can be tested from these simulations in NAMD or CHARMM. Another potential explaination is that the relatively short simulation of Klauda et al. (40 ns) was not converged. As for other POPE force fields, the standard Berger parameters will not perform well. There are some force fields that have been reported to perform well. Apart from a couple of GROMOS ones I know of (GROMOS-CKP and GROMOS 43A1-S3 - I have used these in the past and both behave well at 313 K, which is well above the gel-liquid crystal phase transition temperature of 298 K), all-atom AMBER lipid parameters have been recently reported that include POPE (http://pubs.acs.org/doi/abs/10.1021/ct300342n). This may be another option that could be used for these simulations, with an AMBER force field used for the protein. Cheers Tom On 15/08/12 20:09, Peter C. Lai wrote: On 2012-08-15 06:55:59PM +, Christopher Neale wrote: Write the authors of the simulation paper that has a correct APL for POPE and ask them for an input file. That is really the only way to be sure that you are not doing something different than they did. In my experience, people are quite willing to provide you with their input file(s). If you still get a different APL than they reported, then see if your simulation times are similar and repeat your run a few times to see if it's just statistical noise. The fundamental problem Sebastian will have is that Klauda obtained their APLs using CHARMM software, and he is trying to reproduce this using the forcefield in Gromacs software. So even if the CHARMM input files were provided, it maybe difficult to exactly reproduce the conditions in Gromacs (if certain parameters were implemented differently) Regarding 323 K, I don't recall... it's just a number that sticks in my head. Perhaps it is for DPPE or DPPC. I'd still suggest that you at least try POPC. So your peptide binds more favourably to POPE than to POPC... that alone does not limit you to POPE. Then again, I don;t know exactly what you are trying to do. Chris. It is generally a good idea to use a higher temp than the phase transition temperature, since during equilibration close to the phase transition temp there is a risk of inducing some ordering due to uneven heating. People run DPPC at 323 because its phase transition temp is 315K. If POPC's is 271 and people typically run POPC at 300, then it may be wise to bump up the running temp of a POPE system. Of course, your APL will inflate at higher temperatures... -- original message -- My peptide is known to be more favorably to PE than PC membrane that is why I am using POPE. Experimentally, the liquid phase transition is at 298K for POPE (if I am not mistaken). Is your 323K refer to some simulations? At first I wanted to use the new CHARMM36 lipids parameters because they are supposed to solve the previous CHARMM27 issue with the area per lipid. However, I am consistently obtained smaller APL then experiment and I am not able to reproduce the published APL obtained for POPE, even if I am starting from their equilibrated 80-POPE membrane and use same simulation conditions. That was the reason for starting this thread on the mailing list. Unfortunately, my peptide conformational space in solution is only well-represented by CHARMM27 (equivalently in CHARMM36), so I can not use Berger's lipid parameters with OPLS or GROMOS even if it would be preferable as they do not have APL inconsistency and are united-atom. I will made some tests in the NPAT ensemble. Perhaps the NPAT effects can be made neglegible by using bigger membrane compared to my peptide's size (?). Sebastien -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- Dr Thomas Piggot University of Southampton, UK. -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post?
[gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Well, gromacs is not the only software available. I'd still ask them and then try it in gromacs after parsing. \If there is a difference, then try in NAMD and/or charmm. I know that this is the gromacs users list, but we're talking about debugging here and I think that getting the original parameter file and interpreting to gromacs or exactly rerunning in charmm is still the best way to go. If he can reproduce it in charmm, but not in gromacs, then the problem becomes more well defined. I stand by my advice. Chris. -- original message -- Peter C. Lai pcl at uab.edu Wed Aug 15 21:09:31 CEST 2012 Previous message: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why? Next message: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why? Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] On 2012-08-15 06:55:59PM +, Christopher Neale wrote: Write the authors of the simulation paper that has a correct APL for POPE and ask them for an input file. That is really the only way to be sure that you are not doing something different than they did. In my experience, people are quite willing to provide you with their input file(s). If you still get a different APL than they reported, then see if your simulation times are similar and repeat your run a few times to see if it's just statistical noise. The fundamental problem Sebastian will have is that Klauda obtained their APLs using CHARMM software, and he is trying to reproduce this using the forcefield in Gromacs software. So even if the CHARMM input files were provided, it maybe difficult to exactly reproduce the conditions in Gromacs (if certain parameters were implemented differently) Regarding 323 K, I don't recall... it's just a number that sticks in my head. Perhaps it is for DPPE or DPPC. I'd still suggest that you at least try POPC. So your peptide binds more favourably to POPE than to POPC... that alone does not limit you to POPE. Then again, I don;t know exactly what you are trying to do. Chris. It is generally a good idea to use a higher temp than the phase transition temperature, since during equilibration close to the phase transition temp there is a risk of inducing some ordering due to uneven heating. People run DPPC at 323 because its phase transition temp is 315K. If POPC's is 271 and people typically run POPC at 300, then it may be wise to bump up the running temp of a POPE system. Of course, your APL will inflate at higher temperatures... -- original message -- My peptide is known to be more favorably to PE than PC membrane that is why I am using POPE. Experimentally, the liquid phase transition is at 298K for POPE (if I am not mistaken). Is your 323K refer to some simulations? At first I wanted to use the new CHARMM36 lipids parameters because they are supposed to solve the previous CHARMM27 issue with the area per lipid. However, I am consistently obtained smaller APL then experiment and I am not able to reproduce the published APL obtained for POPE, even if I am starting from their equilibrated 80-POPE membrane and use same simulation conditions. That was the reason for starting this thread on the mailing list. Unfortunately, my peptide conformational space in solution is only well-represented by CHARMM27 (equivalently in CHARMM36), so I can not use Berger's lipid parameters with OPLS or GROMOS even if it would be preferable as they do not have APL inconsistency and are united-atom. I will made some tests in the NPAT ensemble. Perhaps the NPAT effects can be made neglegible by using bigger membrane compared to my peptide's size (?). Sebastien -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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. Regards, Felipe 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: p...@uab.edu To: gmx-users@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: da...@cornell.edu To: gmx-users@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.cot...@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
RE: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Dear Felipe, I will take a lot at this paper. Thanks! Sebastien Date: Tue, 4 Aug 012 9::8::7 +200 From: luis.pinedadecas...@lnu.se To: gmx-users@gromacs.org Subject: Re: [gmx-users] CHARMM6 - Smaller Area per lipid for POPE - Why? 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. (004))6,, 556--5 Take maybe a look at it, if you haven't done it already. Regards, Felipe On 8//3//012 1::2 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 012--8--9 2::4::9PPM -300,, 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 (7 to 5 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, Aug 012 0::9::6 -500 From: p...@uab.edu To: gmx-users@gromacs.org Subject: Re: [gmx-users] CHARMM6 - 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+ layer of water or Membrane+Protein and Water separately (like in Justin's KALP5 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 012--8--9 9::8::1AAM +000,, Mark Abraham wrote: On //8//012 ::8 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 of manual. Mark Thanks, Sebastien Date: Fri, Aug 012 1::0::2 -400 Subject: Re: [gmx-users] CHARMM6 - Smaller Area per lipid for POPE - Why? From: da94@@cornell.edu To: gmx-users@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 ,, 012 at ::2 AM, Sebastien Cote
RE: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Dear Peter, I also used h-bonds and I also switch LJ interaction from 0.8 nm to 1.2 nm (as in Klauda's paper). I will retry with a more solvated membrane. Would you have any thought on how the NPAT ensemble might affect peptide-membrane interactions like I am studying i.e. peptide is totally solvated, then adsorb, and finally may insert? The paper on peptide-membrane interaction like this usually use united-atom lipid in the NPT ensemble. Most of the work I have seen on Charmm membrane in the NPAT ensemble were for embedded membrane protein. Thanks, Sebastien Date: Mon, 13 Aug 2012 16:12:29 -0500 From: p...@uab.edu To: gmx-users@gromacs.org Subject: Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why? 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: p...@uab.edu To: gmx-users@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: da...@cornell.edu To: gmx-users@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
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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: p...@uab.edu To: gmx-users@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: da...@cornell.edu To: gmx-users@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.cot...@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
RE: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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: p...@uab.edu To: gmx-users@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: da...@cornell.edu To: gmx-users@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.cot...@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
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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. I used vdwswitch of 0.8 instead of the 1.0 that was in Klauda's supplemental information and I had turned off DispCorr. at 300K the 238 lipid POPC layer fluctuated between 63 and 65 APL after about 90ns and I figured that was good enough for me. Of course I really only care about the hydrophobic effect on my embedded GPCR so getting enough precision in the APL wasn't my primary goal, but after re-equlibrating the bilayer around the restrained GPCR, the average APL remained around 64. I used a variant of Justin's GridMAT-MD to measure the APLs (to compensate for the embedded protein atoms). I am also using a lot of water - the thickness of water an each of the layer is almost the thickness of one leaflet because my protein has long loops. Oh btw, I am using TIPS3P (Charmm) water model. As noted in this thread, switch is implemented differently between CHARMM and Gromacs, but the recommendation for DPPC was to leave it at 0.8 for gromacs. This thread also contains the note about using TIPS3P. http://www.mail-archive.com/gmx-users@gromacs.org/msg34621.html 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? Yeah that's what I was trying to visualize. If your protein isn't embedded and it's only interacting with, say, the intracellular leaflet, then maybe try COM groups of: Protein+inner leaflet+inner water/ions, outer leaflet, outer water/ions. I hesistate to put all waters into their own separate COM group because the protein should be coupled to its starting solvent. Because your protein is small, also check your thermostat coouplings too. 35 AA might do better with the protein temperature coupled with the water it is starting in... Thanks for the suggestion, Sébastien Date: Wed, 8 Aug 2012 20:19:56 -0500 From: p...@uab.edu To: gmx-users@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: da
RE: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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? Thanks, Sebastien Date: Fri, 3 Aug 2012 11:10:22 -0400 Subject: Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why? From: da...@cornell.edu To: gmx-users@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.cot...@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: p...@uab.edu To: gmx-users@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
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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: da...@cornell.edu To: gmx-users@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.cot...@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: p...@uab.edu To: gmx-users@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
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: da...@cornell.edu To: gmx-users@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.cot...@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
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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.cot...@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: p...@uab.edu To: gmx-users@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
RE: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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: p...@uab.edu To: gmx-users@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: p...@uab.edu To: gmx-users@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
Re: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
Hi Sebastien, I have not tested POPE, but I have tested in detail DPPC and POPC. Because of this I cannot be certain of your issue. The only differences in your setup to what I generally use is I use rcoulomb of 1.2 and rlist of 1.2 (with the 0.8/1.2 vdw switching). There are differences in the way that GROMACS switches off the van der Waals interactions, compared to how it is normally done in CHARMM and so this may be impacting upon your simulation. That said, using a very similar setup to yours I have found this impact to be generally fairly minimal (but again not with POPE). The only other reason that I can think of is that simulation in the Klauda paper just may not have converged (the POPE area per lipid in Figure 5 of this paper does look like it might potentially still be decreasing). If I were you, I would perform some simulations of the POPE membrane in either NAMD or CHARMM. This way you can be sure that the area per lipid you are seeing is not due to any differences in simulation codes. Sorry I can't be of more help. Cheers Tom Sebastien Cote 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: p...@uab.edu To: gmx-users@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
RE: [gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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. Thanks, Sebastien Date: Fri, 20 Jul 2012 12:47:44 -0500 From: p...@uab.edu To: gmx-users@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 -- gmx-users mailing list gmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- == Peter C. Lai | University of Alabama-Birmingham Programmer/Analyst | KAUL 752A Genetics, Div. of Research | 705 South 20th Street p...@uab.edu | Birmingham AL 35294-4461 (205) 690-0808 | == -- gmx-users mailing list gmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
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: p...@uab.edu To: gmx-users@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 -- gmx-users mailing list gmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- == Peter C. Lai | University of Alabama-Birmingham Programmer/Analyst | KAUL 752A
[gmx-users] CHARMM36 - Smaller Area per lipid for POPE - Why?
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 -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
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 -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists -- == Peter C. Lai| University of Alabama-Birmingham Programmer/Analyst | KAUL 752A Genetics, Div. of Research | 705 South 20th Street p...@uab.edu| Birmingham AL 35294-4461 (205) 690-0808 | == -- gmx-users mailing listgmx-users@gromacs.org http://lists.gromacs.org/mailman/listinfo/gmx-users * Only plain text messages are allowed! * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting! * Please don't post (un)subscribe requests to the list. Use the www interface or send it to gmx-users-requ...@gromacs.org. * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists