normal reverse transformation and parameters for lipids

  • mferraro
  • mferraro's Avatar Topic Author
  • Offline
  • Junior Boarder
More
10 years 9 months ago #1670 by mferraro
Dear Martini-users,

I'm working with 2 type of CG membrane-protein complexes:

1) POPC-protein
2) DSPC-POPE-CHOL-protein.

Now I need to come back to the atomistic representation of my systems, so starting with the homogeneous system, I tried to follow your tutorial on reverse transformation. I read the article and all the steps needed (gromacs3.3.1 installation). I tried to generate a file for my system similar to "fg.top" described in the tutorial; I decided to map the protein, the bilayer and the water separately and to add the respective topologies with the include statement in a topology file, like the tutorial example. For the protein I used "pdb2gmx" to obtain the mapping and it works fine. I chose the gromos 53a6 force field, can I? Or do I have to choose the same foce field of the tutorial (43a2)?
For lipids, the problems grow:
I can't do the mapping with "pdb2gmx" because residue "POPC" is not found in residue type database. Is it expected? Can I use the reverse transformation with other bilayers than DPPC?
I'm going to manually do the mapping for my POPC using DPPC as a template. In this way I can avoid using "pdb2gmx" for lipids. Anyway both atomistic DSPC.itp and POPE.itp, neither for 53a6 or 43a2 forcefields can be downloaded from web. How can I do? Is it the right way?
Any suggestion will be greatly appreciated.
Thanks in advance.
Maria.

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

More
10 years 9 months ago #1671 by djurre
What you want to do can work, but it will require some more tweaking by yourself.

-You can use the 53A6 forcefield (just make sure you pick the one with "mapping" in the name.

-You can use other lipids than DPPC, you will have to make a special itp file for them. The atomistic itp file will be the same as the original, except it has a [mapping] directive. As you already mention, it is probably easiest to use the example of DPPC as template here.

-The examples of DSPC and POPE might be you biggest problem. I don't know which atomistic parameters you used for POPC (eg Berger, Slipids)? For the stockholm lipids ( people.su.se/~jjm/Stockholm_Lipids/Downloads.html ) there are parameters for quite a lot of lipids. However they are Amber compatable, so that would give a lot more work to get amber working the reverse transformation code...

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

More
10 years 9 months ago #1672 by djurre
You might already know this, but the lipidbook ( lipidbook.bioch.ox.ac.uk/ ) is a great resource for lipid parameters.

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

  • mferraro
  • mferraro's Avatar Topic Author
  • Offline
  • Junior Boarder
More
10 years 9 months ago #1676 by mferraro
Thanks Djurre for your prompt reply,

I inserted POPC (from lipidbook web site) in rtp file for 53a6 gromos force field and pdb2gmx works fine. It gives back correct .gro and .top files starting from a 53a6 pdb file.

Now, I'm trying generate the directive [ dihedral constraints ] in topology file via g_dihfix_d tool, as suggested in the paper of Rzepiela et al.
This is the command line I used for DPPC and POPC:

g_dihfix_d -c dppc.gro -fc 10 -p dppc.top -o restr-dppc.top

the output is:

calling cpp...
processing topology...
Generated 165 of the 1596 non-bonded parameter combinations
Excluding 3 bonded neighbours for Protein 1
# G96BONDS: 49
# G96ANGLES: 57
# PDIHS: 43
# IDIHS: 3
# LJ14: 53

Number of atoms in protein 50
There may be as many as 44 dihedrals...
Found 40 dihedrals
Segmentation fault (core dumped)


This occurs in processing DPPC, which is already present in rtp file for 53a6 ff, and POPC, so it let me be think that the problem doesn't come from an error in parameters used.
Could you kindly help me to fix this problem?
It could probably be interesting that g_dihfix works fine with the protein.
Thanks again for your precious help.

Maria.

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

More
10 years 9 months ago #1679 by xavier
Hi Maria,

FYI: the protein force field GROMOS-54a6 is not so good for helices. You'd might want to consider the GROMOS-54a7 by schmid et al Euro Biophys J 2011. (For protein ...)

I forgot the purpose of these dihedrals! What is it? Do you need to define all of them, automatically?

mferraro wrote: Thanks Djurre for your prompt reply,

I inserted POPC (from lipidbook web site) in rtp file for 53a6 gromos force field and pdb2gmx works fine. It gives back correct .gro and .top files starting from a 53a6 pdb file.

Now, I'm trying generate the directive [ dihedral constraints ] in topology file via g_dihfix_d tool, as suggested in the paper of Rzepiela et al.
This is the command line I used for DPPC and POPC:

g_dihfix_d -c dppc.gro -fc 10 -p dppc.top -o restr-dppc.top

the output is:

calling cpp...
processing topology...
Generated 165 of the 1596 non-bonded parameter combinations
Excluding 3 bonded neighbours for Protein 1
# G96BONDS: 49
# G96ANGLES: 57
# PDIHS: 43
# IDIHS: 3
# LJ14: 53

Number of atoms in protein 50
There may be as many as 44 dihedrals...
Found 40 dihedrals
Segmentation fault (core dumped)


This occurs in processing DPPC, which is already present in rtp file for 53a6 ff, and POPC, so it let me be think that the problem doesn't come from an error in parameters used.
Could you kindly help me to fix this problem?
It could probably be interesting that g_dihfix works fine with the protein.
Thanks again for your precious help.

Maria.

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

  • mferraro
  • mferraro's Avatar Topic Author
  • Offline
  • Junior Boarder
More
10 years 9 months ago #1686 by mferraro
Hi Xavier, I read about this problem with 53a6 ff, so I'm going to use 43a2 ff as suggested in tutorial both for Protein and lipid. There wouldn't be problems, right?

I have to fix dihedrals states that are separated by high energy barriers, so that interconversion between them do not occur at high temperatures provided for by cg to fg procedure (simulated annealing). It is also reported in Rzepiela's paper.
With g_dihfix tool I can automatically do that, having fixed a threshold (force constant "-fc").
Tutorial files don't give any information about this procedure because they already provide file with the section [ dihedral restraints ] in dppc.itp file.
I used rtp files found in the tutorial and I overwrote those in gromacs/share/top directory.
The error message continue to be the same.

calling cpp...
processing topology...
Generated 483 of the 1653 non-bonded parameter combinations
Excluding 3 bonded neighbours for DPPC 1
NOTE:
System has non-zero total charge: 7.580000e+02

# G96BONDS: 48
# G96ANGLES: 53
# PDIHS: 15
# RBDIHS: 24
# IDIHS: 3
# LJ14: 49

Number of atoms in protein 50
There may be as many as 16 dihedrals...
Found 14 dihedrals
Segmentation fault (core dumped)

Do you think that it would be better to inserting dihedrals manually, on acyl groups and double bonds (in case of POPC)?
I have cholesterol in my systems: considering its stiffness, I've thought not to fix dihedrals at all. Am I wrong?

Thanks again for your concern in my problem.
Maria.

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

More
10 years 9 months ago #1687 by xavier
Regarding the force field, the GROMOS 43a2 is though to be good but 54a7 should be fine. I have not used it though.

I would indeed try the alternative to write the dihedral by hand using the DPPC as example.

I do not think you need to do anything for Cholesterol.

mferraro wrote: Hi Xavier, I read about this problem with 53a6 ff, so I'm going to use 43a2 ff as suggested in tutorial both for Protein and lipid. There wouldn't be problems, right?

I have to fix dihedrals states that are separated by high energy barriers, so that interconversion between them do not occur at high temperatures provided for by cg to fg procedure (simulated annealing). It is also reported in Rzepiela's paper.
With g_dihfix tool I can automatically do that, having fixed a threshold (force constant "-fc").
Tutorial files don't give any information about this procedure because they already provide file with the section [ dihedral restraints ] in dppc.itp file.
I used rtp files found in the tutorial and I overwrote those in gromacs/share/top directory.
The error message continue to be the same.

calling cpp...
processing topology...
Generated 483 of the 1653 non-bonded parameter combinations
Excluding 3 bonded neighbours for DPPC 1
NOTE:
System has non-zero total charge: 7.580000e+02

# G96BONDS: 48
# G96ANGLES: 53
# PDIHS: 15
# RBDIHS: 24
# IDIHS: 3
# LJ14: 49

Number of atoms in protein 50
There may be as many as 16 dihedrals...
Found 14 dihedrals
Segmentation fault (core dumped)

Do you think that it would be better to inserting dihedrals manually, on acyl groups and double bonds (in case of POPC)?
I have cholesterol in my systems: considering its stiffness, I've thought not to fix dihedrals at all. Am I wrong?

Thanks again for your concern in my problem.
Maria.

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

  • mferraro
  • mferraro's Avatar Topic Author
  • Offline
  • Junior Boarder
More
10 years 9 months ago #1688 by mferraro
Hi Xavier,

ok, I will do that by hand!

Just another question: do you think I can use different united atom force fields for protein and lipids (43a2 for protein and 53a6 for lipids)? Unfortunately 43a2 parameters for POPC are not available on web. Obviously I am going to use them just for reverse transformation.

Thanks.

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

More
10 years 9 months ago #1692 by xavier
No it is never a good idea to mix force fields unless you have tested them. There is a GROMOS 54a7 which is supposed to be better than 53a6 ...

mferraro wrote: Hi Xavier,

ok, I will do that by hand!

Just another question: do you think I can use different united atom force fields for protein and lipids (43a2 for protein and 53a6 for lipids)? Unfortunately 43a2 parameters for POPC are not available on web. Obviously I am going to use them just for reverse transformation.

Thanks.

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

Time to create page: 0.105 seconds