- Posts: 54
Membrane protein simulations
- flaviyan
-
Topic Author
- Offline
- Senior Boarder
Less
More
9 years 9 months ago #3964
by flaviyan
Membrane protein simulations was created by flaviyan
Dear all,
I am simulating a helical trans membrane protein coarse grained using the 2.4 version of martinize.py(DSSP included) with the membrane built using insane.py.
The 2 simulations were carried out for 200ns with martini22p and elendyn22p force field with the following parameters: 310K, 0.15mM NaCl, POPC/CHOL(2:1) membrane, 1 bar with PW water model (epsilon_r 2.5) rlist cut-off 1.4nm, shifted potentials for electrostatics and VDW at 0.9 to 1.2, 20fs timesteps. Gromacs 4.5.5 was used in both simulations (after standard EM, NVT and NPT routines. Post simulation I used the backward tool to reverse map 50ns snapshots of the simulations. These are the following observations form the simulations and I am not sure if these observations are consistent with CG and I am not how reliable are my results. I need help with this :
~martini22p force field~
RMSD was stable after 2ns of the simulation. But the extra cellular loops had large deviations in the secondary structure from Beta sheets to coils in less than 15ns of the simulation. I am not sure if this is due to simulation parameters.
~elendyn22pforce feild~
in order to restrain the secondary structure from having large deviation we carried out the same simulations with elendyn22p force field. The RMSD stabilized within 2 ns and remained stable throughout the simulation. The secondary structure did not collapse during the simulation but there is huge deviations in the ramachandran plot from the initial and final structures. i.e residues shifted from initial left handed helix configuration to right handed helix and some to beta sheet configuration (though residues which were initially in the right helix and beta sheet configuration remained unchanged). These results seemed as strange as before. I did not find anything in the literature on this.
~atomistic~
I simulated the same protein in same membrane configuration, parameters with PME and TIP3 water model using CHARMM36 all atom force feild for 100ns. The RMSD deviation was stable and the secondary structure was also stable and Ramachandran plot remained unchanged except for 1 or 2 residues.
These results are confusing and I am not sure if this is due to the CG force field or backward mapping. The initial protein structure was a homology model from a close homolog crystal structure. the purpose of this simulation was to get a membrane equillbrated structure that would rectify any inaccuracies in homology model. We intend to use the equillbrated structure for further interaction and docking studies. But due to these ambiguities in the results I am not sure what to interpret from them.
Due to our limitation in high speed computer resources we are heavily reliant on CG simulations for large sampling (the 100 ns atomistic simulation took 20 days in comparison the CG simulation took only 6 days for 200ns on the same resource) so it would be great if someone provided guidance on this or share their comments on any modifications.
Thanks
Flaviyan
I am simulating a helical trans membrane protein coarse grained using the 2.4 version of martinize.py(DSSP included) with the membrane built using insane.py.
The 2 simulations were carried out for 200ns with martini22p and elendyn22p force field with the following parameters: 310K, 0.15mM NaCl, POPC/CHOL(2:1) membrane, 1 bar with PW water model (epsilon_r 2.5) rlist cut-off 1.4nm, shifted potentials for electrostatics and VDW at 0.9 to 1.2, 20fs timesteps. Gromacs 4.5.5 was used in both simulations (after standard EM, NVT and NPT routines. Post simulation I used the backward tool to reverse map 50ns snapshots of the simulations. These are the following observations form the simulations and I am not sure if these observations are consistent with CG and I am not how reliable are my results. I need help with this :
~martini22p force field~
RMSD was stable after 2ns of the simulation. But the extra cellular loops had large deviations in the secondary structure from Beta sheets to coils in less than 15ns of the simulation. I am not sure if this is due to simulation parameters.
~elendyn22pforce feild~
in order to restrain the secondary structure from having large deviation we carried out the same simulations with elendyn22p force field. The RMSD stabilized within 2 ns and remained stable throughout the simulation. The secondary structure did not collapse during the simulation but there is huge deviations in the ramachandran plot from the initial and final structures. i.e residues shifted from initial left handed helix configuration to right handed helix and some to beta sheet configuration (though residues which were initially in the right helix and beta sheet configuration remained unchanged). These results seemed as strange as before. I did not find anything in the literature on this.
~atomistic~
I simulated the same protein in same membrane configuration, parameters with PME and TIP3 water model using CHARMM36 all atom force feild for 100ns. The RMSD deviation was stable and the secondary structure was also stable and Ramachandran plot remained unchanged except for 1 or 2 residues.
These results are confusing and I am not sure if this is due to the CG force field or backward mapping. The initial protein structure was a homology model from a close homolog crystal structure. the purpose of this simulation was to get a membrane equillbrated structure that would rectify any inaccuracies in homology model. We intend to use the equillbrated structure for further interaction and docking studies. But due to these ambiguities in the results I am not sure what to interpret from them.
Due to our limitation in high speed computer resources we are heavily reliant on CG simulations for large sampling (the 100 ns atomistic simulation took 20 days in comparison the CG simulation took only 6 days for 200ns on the same resource) so it would be great if someone provided guidance on this or share their comments on any modifications.
Thanks
Flaviyan
Please Log in or Create an account to join the conversation.
- djurre
-
- Offline
- Admin
Less
More
- Posts: 272
9 years 9 months ago #3966
by djurre
Using the Martini forcefield secondary structure should be stable (in contrast, possibly, to ternary structure). That is, unless there was no secondary structure type recognized/given when you generated the topology (with the martinize.py script). The secondary structure type found/used is given as a string at the top of your .itp file (and it should not be all 'C'=coil)
With the elendyn FF the secondary and ternary structure should be stable, as you find. That the ramachandran plot changes extremely (!). If the handedness of the helix changes, you should be able to seet his when looking at the structure and the RMSD should also not stay stable.
So I would say these changes have something to do with the backmapping. (You are measuring the RMSD on the backmapped/atomistic structure right?)
Since 2nd/3th structure can't change in the Martini/Elnedyn forcefields they are not well suited to 'rectify inaccuracies'. For this you could use the AA simulations. For the second part you could than use the CG model.
Did the systems contain the same number of proteins/lipids? Because three times faster simulations (~20/6) is not so much. You should be able to get much faster results with the CG model.
Replied by djurre on topic Membrane protein simulations
flaviyan wrote: ~martini22p force field~
RMSD was stable after 2ns of the simulation. But the extra cellular loops had large deviations in the secondary structure from Beta sheets to coils in less than 15ns of the simulation. I am not sure if this is due to simulation parameters.
Using the Martini forcefield secondary structure should be stable (in contrast, possibly, to ternary structure). That is, unless there was no secondary structure type recognized/given when you generated the topology (with the martinize.py script). The secondary structure type found/used is given as a string at the top of your .itp file (and it should not be all 'C'=coil)
flaviyan wrote: ~elendyn22pforce feild~
in order to restrain the secondary structure from having large deviation we carried out the same simulations with elendyn22p force field. The RMSD stabilized within 2 ns and remained stable throughout the simulation. The secondary structure did not collapse during the simulation but there is huge deviations in the ramachandran plot from the initial and final structures. i.e residues shifted from initial left handed helix configuration to right handed helix and some to beta sheet configuration (though residues which were initially in the right helix and beta sheet configuration remained unchanged). These results seemed as strange as before. I did not find anything in the literature on this.
With the elendyn FF the secondary and ternary structure should be stable, as you find. That the ramachandran plot changes extremely (!). If the handedness of the helix changes, you should be able to seet his when looking at the structure and the RMSD should also not stay stable.
So I would say these changes have something to do with the backmapping. (You are measuring the RMSD on the backmapped/atomistic structure right?)
flaviyan wrote: The purpose of this simulation was to get a membrane equillbrated structure that would rectify any inaccuracies in homology model. We intend to use the equillbrated structure for further interaction and docking studies. But due to these ambiguities in the results I am not sure what to interpret from them.
Since 2nd/3th structure can't change in the Martini/Elnedyn forcefields they are not well suited to 'rectify inaccuracies'. For this you could use the AA simulations. For the second part you could than use the CG model.
flaviyan wrote: ... took 20 days in comparison the CG simulation took only 6 days for 200ns
Did the systems contain the same number of proteins/lipids? Because three times faster simulations (~20/6) is not so much. You should be able to get much faster results with the CG model.
Please Log in or Create an account to join the conversation.
Time to create page: 0.087 seconds