normal Bonds shifting while going from atomistic to CG

  • aishwaryts
  • aishwaryts's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
7 years 1 week ago #7179 by aishwaryts
Hello all,

I am developing parameters for certain molecule.

I got the atomistic trajectory and converted to pseudo atomistic trajectory and got the bonds and angles.

When I run a CG simulation with the same parameters the distributions are shifting.
For example:

[ moleculetype]
; Name nrexcl
GLMN 1

[ atoms ]
; nr type resnr residue atom cgnr charge mass
1 N0 1 GLMN N0 1 0 42.0400
2 P1 2 GLMN P1 2 0 60.0600
3 P1 3 GLMN P1 3 0 43.0540
4 P5 4 GLMN P5 4 0 58.0600
5 P2 5 GLMN P2 5 0 43.0540
6 N0 6 GLMN N0 6 0 59.0600
7 P1 7 GLMN P1 7 0 59.0640

[ bonds ]
; ai aj funct length fc
1 2 1 0.236 12000
1 3 1 0.192 35000
3 4 1 0.318 12000
1 5 1 0.474 15000
5 6 1 0.245 30000
5 7 1 0.223 30000
[angles]
2 1 3 1 144 200
1 3 4 1 133 180
2 1 5 1 96 180
3 1 5 1 115.5 240
1 5 6 1 77.8 300
1 5 7 1 81 300
[dihedrals]
2 1 3 4 1 -559 5.73 1
2 1 5 7 1 0 7 1
2 1 5 6 1 94.5 9 1
6 5 1 3 1 229 8 1

I have attached the pictures for the distributions. There is quite a lot shift in distributions.

distributions

What could be the reason?

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

More
7 years 1 week ago #7183 by bart
The non-bonded interactions.

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

  • aishwaryts
  • aishwaryts's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
7 years 1 week ago #7189 by aishwaryts
Replied by aishwaryts on topic Bonds shifting while going from atomistic to CG
Even after playing with bead types the distributions remain the same.

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

More
7 years 6 days ago #7191 by bart
What I tried to say in a way too short manner, is that the bonded interactions alone are not completely responsible for the distributions you get from any MD molecule (CG or AA). The non-bonded parameters will also influence it. This is because the bonded interactions (even within the same molecule) can interact with each other through the non-bonded interactions.
For the non-bonded interactions to be roughly correct a table was published in the original martini paper giving you a relatively good insight in what beads you should choose. Then you create a pseudo CG trajectroy from you AA simulation and use that to define your intial CG bonded interactions. Then you simulate your CG molecule and track the CG bonded-interactions distribution. Since this will most likely not be perfect you can alter the CG bonded interactions and run again. Keep doing this until you are satisfied with the results. You could try and alter the bead type, but this could potentially ruin the correct partitioning or any other general non-bonded interaction. Therefore I would advice to only do that if you have good reasons to do so. I hope this is more clear than my previous answer.

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

More
7 years 5 days ago #7195 by yogi@martini
Replied by yogi@martini on topic Bonds shifting while going from atomistic to CG
Thanks Bart for such nice answer. I just need small clarification. If we get shift in the bond distribution as mentioned by Aishwaryts, changing the bead type is the only available option? Are there any guidelines in choosing the bead? for example if you get shift on the left side you take more polar molecule or something like that, or is it completely random choice? Please enlighten me!

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

More
7 years 5 days ago #7196 by bart
Again I would like to stress, that diverging from the original partition bases bead assignment is probably not going to solve your problem and bears great risk of ruing the behavior of the molecule.

I missed the picture of the graph up to now, but if your axes is in nm you are overmapping your structure big time. The distance you would like to achieve is less than half a CG bead's diameter (0.47 nm). I guess you should review your mapping.

I would like to stress that you should take a good look at the orginal paper and the mapping/parametrization tutorials on our web page.

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

  • aishwaryts
  • aishwaryts's Avatar Topic Author
  • Offline
  • Fresh Boarder
More
7 years 5 days ago #7197 by aishwaryts
Replied by aishwaryts on topic Bonds shifting while going from atomistic to CG
Thank you! Bart, for the reply.

I am trying to parametrise the N-acetylglucosamine sugar attached other sugar molecules by different glycosidic linkages. I have mapped like this , following the paper "Martini Coarse-Grained Force Field: Extension to Carbohydrates (J. Chem. Theory Comput. 2009, 5, 3195–3210)".

The lowest bond length used in this paper (for disaccharides involving glucose) is 0.222 nm, which I am also observing. Only the bond between SP2 and SP1 is coming around 0.190 nm which is where the bonds and angle distributions are shifting.

1)I tried increasing the polarity of beads by 1 level so that increased attraction will give shorter the bond lengths but not much difference was observed.

2)I also tried manually tweaking the bond lengths which was also ineffective.

3)Lastly yesterday I tried nrexcl = 2 and now all the distributions match exactly to the pseudo atomistic distributions. (But is it okay to do something like this? as we are increasing inclusions to 2, this will lower interactions between beads).

After reading the last answer, I feel that I should include N atom in the SP1 so that I get a bigger bond length.

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

More
7 years 4 days ago #7198 by bart
Once again I would like to stress to not change the bead types because of this reason just don't. The fact that it helps when you set nrexcl to 2 means that indeed overlap of beads which are interacting through non-bonded interactions are causing the beads to not come so close. Even if you would change the bead types (What you should not do!) there is no interacting in Martini so short ranged that it would allow for 0.2 nm. I don't know what will happen in total if you set nrexcl for your molecule, maybe it works perfectly fine and this is your solution. However, I still wouldn't advice of putting beads so close together. This however is up to you and if you really need to and there is no other way maybe you should. Just be careful for you are pushing the model into areas for which it was not designed.

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

More
7 years 2 days ago #7203 by peterkroon
Replied by peterkroon on topic Bonds shifting while going from atomistic to CG
Be sure to also validate the partition free energy of your new molecule, preferably against experimental data. Particularly for (overmapped) rings the bonds are usually made longer than they're supposed to be (compared to AA) to make the partition free energy match.
Constraints instead of bonds might also be applicable; but do note that either way with bonds this short your simulation is going to be unstable. Instead of nrexcl=2 you can maybe also put one specific exclusion between the troublesome beads.

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

Time to create page: 0.112 seconds