normal Problems with CG2FG - moving atoms out of protein

  • DavidSBCB
  • DavidSBCB's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
13 years 1 month ago #511 by DavidSBCB
First off, I can complete the reverse transformation method in its entirety to produce an atomistic system from a coarse grained system, however, whenever I do this certain atoms within my molecule are moved around 8 angstroms away from my protein.

The step at which this problem occurs is the g_fg2cg step, after which my output has moved atoms. At first I thought this was because some of my coarse grained beads were slightly over the periodic boundary conditions and so atoms were being placed randomly in the beads across the box, however by increasing the box size dramatically and moving the protein I have been unable to solve this. I was originally simulating within a vacuum, and so proceeded to try again with a box filled with water, however the problem still persists. I have tried the method using numerous different proteins of different sizes (including the Cohesin/Dockerin complex and the Barnase/Barstar system) and still have not solved the issue. The atoms being moved each time vary in number and location within the sequence which leads me to believe it is something to do with their location within the system (ie on the edge of the protein).
I have also looked at changing the bond energy cutoff ( -fc flag in the g_dihfix option) to no avail and have also looked at using the method on individual proteins out of a complex and havent had any luck.

Has anyone had any problem like this before, not necessarily using the CG2FG method but in general, as I have absolutely no idea what is going on or where else to look to try and fix this.

Does anyone have any ideas?

Many thanks for reading.

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

More
13 years 1 month ago #512 by xavier

DavidSBCB wrote: First off, I can complete the reverse transformation method in its entirety to produce an atomistic system from a coarse grained system, however, whenever I do this certain atoms within my molecule are moved around 8 angstroms away from my protein.

The step at which this problem occurs is the g_fg2cg step, after which my output has moved atoms. At first I thought this was because some of my coarse grained beads were slightly over the periodic boundary conditions and so atoms were being placed randomly in the beads across the box, however by increasing the box size dramatically and moving the protein I have been unable to solve this. I was originally simulating within a vacuum, and so proceeded to try again with a box filled with water, however the problem still persists. I have tried the method using numerous different proteins of different sizes (including the Cohesin/Dockerin complex and the Barnase/Barstar system) and still have not solved the issue. The atoms being moved each time vary in number and location within the sequence which leads me to believe it is something to do with their location within the system (ie on the edge of the protein).


I am not sure where and what could go wrong! I might simply be that your topologies are not perfect! Which atomistic forcefield are you using?

DavidSBCB wrote: I have also looked at changing the bond energy cutoff ( -fc flag in the g_dihfix option) to no avail and have also looked at using the method on individual proteins out of a complex and havent had any luck.

I am not sure what you were trying to do here!

DavidSBCB wrote: Has anyone had any problem like this before, not necessarily using the CG2FG method but in general, as I have absolutely no idea what is going on or where else to look to try and fix this.

Does anyone have any ideas?

Many thanks for reading.

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

  • DavidSBCB
  • DavidSBCB's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
13 years 1 month ago #513 by DavidSBCB
Many thanks for the response Xavier,

I thought it might be something to do with the topologies also, as they are the only consistent thing within all simulations, however i've been experimenting with molecules of various sizes and have found that proteins with less than around 150 aa's when run alone turn out fine, but once I go over that number,(or by having multiple proteins present) I start to loose atoms!

There is most likely something wrong with my protocol so I will dissect it to attempt to get more specifics, however if anyone has any ideas! (im pretty stumped at the moment)

Thanks

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

More
13 years 1 month ago #514 by xavier

DavidSBCB wrote: Many thanks for the response Xavier,

I thought it might be something to do with the topologies also, as they are the only consistent thing within all simulations, however i've been experimenting with molecules of various sizes and have found that proteins with less than around 150 aa's when run alone turn out fine, but once I go over that number,(or by having multiple proteins present) I start to loose atoms!

That is good news that you see a protein size effect. The number of residues might not be the problem bu the number of atoms is likely to be. Could it be that everything turns bad when you have more than 100 atoms?
I do not think the code would be the problem. We have done larger systems. But the topology is more likely to be the problem. How do you build your topologies?

Compare carefully a working topology to a non-working one might be useful ... something funky might somewhere.

Which atomistic and CG forcefields are you using?

DavidSBCB wrote: There is most likely something wrong with my protocol so I will dissect it to attempt to get more specifics, however if anyone has any ideas! (im pretty stumped at the moment)

Thanks

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

  • DavidSBCB
  • DavidSBCB's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
13 years 1 month ago #515 by DavidSBCB
That was a thought of mine too, someone I know reported a similar effect when the whitespace between columns was removed due to atom numbers, I have checked this and it doesnt seem to be the problem, so far the effect has occurred on proteins about around 150 residues, I will attempt to isolate the number further and will start comparing working topologies with non-working ones. The atomistic topology is built using pdb2gmx and g_dihfix and the coarsegrain is made by hand incorporating a .itp file

I am using the martini_v2.1 as a coarsegrain forcefield and have tried the GROMOS96 43a2 force field (improved alkane dihedrals) with MAPPING and the GROMOS96 53a6 force field with MAPPING as atomistic forcefields.

Thanks for the help.

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

More
13 years 1 month ago #516 by xavier

DavidSBCB wrote: That was a thought of mine too, someone I know reported a similar effect when the whitespace between columns was removed due to atom numbers, I have checked this and it doesnt seem to be the problem, so far the effect has occurred on proteins about around 150 residues, I will attempt to isolate the number further and will start comparing working topologies with non-working ones.

And lets us know what you find out and if we need to help. This issue should be solved.

The atomistic topology is built using pdb2gmx and g_

    dihfix and the coarsegrain is made by hand incorporating a .itp file

    I am using the martini_v2.1 as a coarsegrain forcefield and have tried the GROMOS96 43a2 force field (improved alkane dihedrals) with MAPPING and the GROMOS96 53a6 force field with MAPPING as atomistic forcefields.

Everything is standard ... that what was important to know.

Thanks for the help.

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

  • DavidSBCB
  • DavidSBCB's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
13 years 1 month ago #517 by DavidSBCB
Problem solved!

After playing around with and checking topologies for a few hours it seems that when I reference the protein topology through a .itp in the atomistic topology (eg #include "protein.itp") it moves the atoms around, however when the topology is explicitly written in the atomistic .top file - all proteins one after the other, the problem appears to go away (so far tested with 3 proteins different complexes).

Many thanks for all the help Xavier!

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

More
13 years 1 month ago #518 by xavier

DavidSBCB wrote: Problem solved!

After playing around with and checking topologies for a few hours it seems that when I reference the protein topology through a .itp in the atomistic topology (eg #include "protein.itp") it moves the atoms around, however when the topology is explicitly written in the atomistic .top file - all proteins one after the other, the problem appears to go away (so far tested with 3 proteins different complexes).

Many thanks for all the help Xavier!


That is really good you can do your transformation, However what you describe should not happen at all! The include statement only copies the file so it should make no difference!

There may be another problem that could be the definition of other things relate to the proteins that are misplaced in the itp files are not included at the right location.

XAvier.

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

  • DavidSBCB
  • DavidSBCB's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
13 years 1 month ago #519 by DavidSBCB
Yes, this what I assume is happening, there must be a subtle difference in the topology files when I explicitly include the protein details, however I haven't been able to see what is different! If I work it out i'll be sure to post so you know exactly what was going on!

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

More
13 years 1 month ago #520 by xavier

DavidSBCB wrote: Yes, this what I assume is happening, there must be a subtle difference in the topology files when I explicitly include the protein details, however I haven't been able to see what is different! If I work it out i'll be sure to post so you know exactly what was going on!


cool.

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

Time to create page: 0.101 seconds