- Posts: 5
Incoperate Go-poential to Martini
- jsbach
-
Topic Author
- Offline
- Fresh Boarder
Less
More
9 years 11 months ago #3816
by jsbach
Incoperate Go-poential to Martini was created by jsbach
Dear Martini users:
Our group are working on incorperate Go-like potential to Martini so we can simulate protein folding use Martini force field. To achieve this, we have made following modifications: 1) Move BB from COM to Ca position (In fact we could keep it in COM in this stage, however it will convience for us in future because we may use some specific REMD method later which requires the bead carry Go potential locate at Ca position); 2) As BB position moved, the BB-BB bond and BB-BB-SC1 parameter need to be modified correspondingly. We get these values by analysising a high resolusion PDB data base.
However, at the second step we got a little lost as we not sure whether we should modify the BB-SC1-SC* angle and the diherdral involved BB (as used in Trp). Based on our statistic, there is little change of BB-SC1-SC* angles in PDB after move BB from COM to Ca, however we are not sure whether we can keep the reference value of these angles. It seems when BB are placed in COM, the BB-SC1-SC* reference angle are quite different from our statistic result. We want to ask if anyone knows are the BB-SC1-SC* angles are directly set to an ideal value from statistic or they are empirical value which have best performence? It seems to be very dangerous to touch these value until we understand the logic behin it. Thank you very much!
Yours
Guang
Our group are working on incorperate Go-like potential to Martini so we can simulate protein folding use Martini force field. To achieve this, we have made following modifications: 1) Move BB from COM to Ca position (In fact we could keep it in COM in this stage, however it will convience for us in future because we may use some specific REMD method later which requires the bead carry Go potential locate at Ca position); 2) As BB position moved, the BB-BB bond and BB-BB-SC1 parameter need to be modified correspondingly. We get these values by analysising a high resolusion PDB data base.
However, at the second step we got a little lost as we not sure whether we should modify the BB-SC1-SC* angle and the diherdral involved BB (as used in Trp). Based on our statistic, there is little change of BB-SC1-SC* angles in PDB after move BB from COM to Ca, however we are not sure whether we can keep the reference value of these angles. It seems when BB are placed in COM, the BB-SC1-SC* reference angle are quite different from our statistic result. We want to ask if anyone knows are the BB-SC1-SC* angles are directly set to an ideal value from statistic or they are empirical value which have best performence? It seems to be very dangerous to touch these value until we understand the logic behin it. Thank you very much!
Yours
Guang
Please Log in or Create an account to join the conversation.
- Clement
-
- Offline
- Admin
Less
More
- Posts: 211
9 years 11 months ago - 9 years 11 months ago #3818
by Clement
Replied by Clement on topic Incoperate Go-poential to Martini
There is already a Martini protein model where backbone beads are moved from COM to Cα positions: the ElNeDyn description, adding an elastic network between backbone beads of proteins to stabilize them. You don't need this latter for your specific goal of course, but at least you could check in there how the bonded parameters are modified from the "standard" Martini. Another solution is to dive into the code of the martinize.py script which generates Martini protein topologies to check for any specific cases.
X. Periole, M. Cavalli, S.J. Marrink, M. Ceruso
Combining an elastic network with a coarse-grained molecular force field: structure, dynamics and intermolecular recognition
J. Chem. Th. Comp., 5:2531-2543, 2009
Hope this helps!
X. Periole, M. Cavalli, S.J. Marrink, M. Ceruso
Combining an elastic network with a coarse-grained molecular force field: structure, dynamics and intermolecular recognition
J. Chem. Th. Comp., 5:2531-2543, 2009
Hope this helps!
Last edit: 9 years 11 months ago by Clement.
Please Log in or Create an account to join the conversation.
- jsbach
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 5
9 years 11 months ago #3823
by jsbach
Replied by jsbach on topic Incoperate Go-poential to Martini
Dear Clement:
Thank you for your quick reply very much! In fact we also got a little puzzle by the elnedy force field. The BB-BB-SC1 angle change from100 degree to 110 degree, the BB-BB bond change from 0.35 to 0.38 nm and BB-SC1 distace also changed afer move from COM to CA based on our statistic, however,as far as we aware, the reference bond and angle values of elnedy and Martini force field are same.
Our difficulty is in elnedy elastic bond are used, however for Go model we need to make two residues have nature contact could not "see" each other when the distance between them are too far (So we use pair LJ potential to represents such nature contact).
Thank you for your quick reply very much! In fact we also got a little puzzle by the elnedy force field. The BB-BB-SC1 angle change from100 degree to 110 degree, the BB-BB bond change from 0.35 to 0.38 nm and BB-SC1 distace also changed afer move from COM to CA based on our statistic, however,as far as we aware, the reference bond and angle values of elnedy and Martini force field are same.
Our difficulty is in elnedy elastic bond are used, however for Go model we need to make two residues have nature contact could not "see" each other when the distance between them are too far (So we use pair LJ potential to represents such nature contact).
Please Log in or Create an account to join the conversation.
- Clement
-
- Offline
- Admin
Less
More
- Posts: 211
9 years 11 months ago #3824
by Clement
Replied by Clement on topic Incoperate Go-poential to Martini
Alright. Now I get your question. Sorry, bit slow today. The standard Martini angles parameters were kept and transferred to the ElNeDyn approach. Which moved backbone beads from COM to Cα. Which works great. How come, is there any reason behind that?
The easiest way to get bonded parameters for any Martini model is to extract them from atomistic simulations of the same molecule. As far as I know, amino-acids aren't an exception to this. But I don't know if such specific study was (re)done when ElNeDyn was developed... But as you mentioned, based on your statistics "there is little change of BB-SC1-SC* angles after moving backbone beads from COM to Cα". The same conclusion was probably drawn at this point, and the "standard" angles parameters were kept.
I invited more people in the conversation; hopefully some will come up with less evasive comments...
The easiest way to get bonded parameters for any Martini model is to extract them from atomistic simulations of the same molecule. As far as I know, amino-acids aren't an exception to this. But I don't know if such specific study was (re)done when ElNeDyn was developed... But as you mentioned, based on your statistics "there is little change of BB-SC1-SC* angles after moving backbone beads from COM to Cα". The same conclusion was probably drawn at this point, and the "standard" angles parameters were kept.
I invited more people in the conversation; hopefully some will come up with less evasive comments...
Please Log in or Create an account to join the conversation.
- jsbach
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 5
9 years 11 months ago #3825
by jsbach
Replied by jsbach on topic Incoperate Go-poential to Martini
Thank you very much!
The BB-SC1-SC* angle statisticted from PDB, even for origin Martini force field, is different from the reference angle in Martini (in this case, both for the statistic here and the force field, the BB are placed in COM). This make me very reluctant to continue move the BB to Ca position as I think it is possible the angle potential and the other potential (e.g. LJ potential) works together to hold the protein structure correct (in another way, I think some of these difference between reference value and PDB everage value were conteracted). If I simply "correct" one value, I'll bring more change than I expected and ruin everything.....
The BB-SC1-SC* angle statisticted from PDB, even for origin Martini force field, is different from the reference angle in Martini (in this case, both for the statistic here and the force field, the BB are placed in COM). This make me very reluctant to continue move the BB to Ca position as I think it is possible the angle potential and the other potential (e.g. LJ potential) works together to hold the protein structure correct (in another way, I think some of these difference between reference value and PDB everage value were conteracted). If I simply "correct" one value, I'll bring more change than I expected and ruin everything.....
Please Log in or Create an account to join the conversation.
- djurre
-
- Offline
- Admin
Less
More
- Posts: 272
9 years 11 months ago #3826
by djurre
Replied by djurre on topic Incoperate Go-poential to Martini
I'm not sure I understand all your questions but let me try to contribute my 2 cents.
You are right that both Martini (BB at COM) and Elnedyn (BB at CA) angles and dihedrals have been tuned with the respect to distributions obtained from PDB structures. The value for the bond equilibrium might not always be the same as the final outcome because, as you mentioned, bonded and non-bonded terms might interfere. The angles are different between Martini and Elnedyn because the distributions will change (a bit) if you change the BB-position.
The angles (and dihedrals in Martini) are responsible for keeping secondary structure stable. So if you want to use the Go-potential to fold the protein, you might want to get rid of those all together.
I've tried to run a simulation like you propose a while a go, just to see if I could do it. I switched of the angles and dihedrals and excluded the interactions between the protein beads. In stead the CA-beads felt each other by a Go-like-potential. The interactions with the outside world (e.g. water) were the normal Martini interactions.
It worked in principle, but there were some problems:
--- The Go potential assumes that there is no water. Since the was in my system I would have had to correct the Mi-potentials for that somehow. If I remember correctly water would sometimes also get trapped inside the protein.
--- The sidechains inside the protein would not feel anything and would sometimes start spinning etc.
I hope it helps. If you want to discuss more about how I'd implemented things you can also contact me off list.
You are right that both Martini (BB at COM) and Elnedyn (BB at CA) angles and dihedrals have been tuned with the respect to distributions obtained from PDB structures. The value for the bond equilibrium might not always be the same as the final outcome because, as you mentioned, bonded and non-bonded terms might interfere. The angles are different between Martini and Elnedyn because the distributions will change (a bit) if you change the BB-position.
The angles (and dihedrals in Martini) are responsible for keeping secondary structure stable. So if you want to use the Go-potential to fold the protein, you might want to get rid of those all together.
I've tried to run a simulation like you propose a while a go, just to see if I could do it. I switched of the angles and dihedrals and excluded the interactions between the protein beads. In stead the CA-beads felt each other by a Go-like-potential. The interactions with the outside world (e.g. water) were the normal Martini interactions.
It worked in principle, but there were some problems:
--- The Go potential assumes that there is no water. Since the was in my system I would have had to correct the Mi-potentials for that somehow. If I remember correctly water would sometimes also get trapped inside the protein.
--- The sidechains inside the protein would not feel anything and would sometimes start spinning etc.
I hope it helps. If you want to discuss more about how I'd implemented things you can also contact me off list.
Please Log in or Create an account to join the conversation.
- jsbach
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 5
9 years 11 months ago - 9 years 11 months ago #3827
by jsbach
Replied by jsbach on topic Incoperate Go-poential to Martini
Dear djurre:
Thank you very much! Yes, we removed all the secondary-related terms and use a unbiased BB atom type determined based on the result of a series of PMF simulations between different tyes of Martini particles.
For the angle, the Gromacs code is modified and a multi-well angle potential is added (shape like
\111_11_111/ )
1\_/1\_/1\_/
The well bottoms correspond to ideal angles of different seondary elements..............
The Go-like potential we used require waters in the system (If fact, we just need Martini + pair potential represents nature contact in currest stage, although it would be great if the BB can be put in Ca position and Secondary biases can be removed which is necessary for next stage )
Till now, all the effert above could hold helix stable, but beta structure is lost...
However, based on disscussion above, it seems it is fine to keep the BB-SC1-SC* untouched.
By the way, could you tell me the logic behind the BB-SC1-SC2 and BB-SC1-SC3 angle of Trp?
I have no idea why use the 210 for BB-SC1-SC2, should this equal to 150 degree? However, based on statistic from PDB, it seems this angle should be 130...
Thank you very much! Yes, we removed all the secondary-related terms and use a unbiased BB atom type determined based on the result of a series of PMF simulations between different tyes of Martini particles.
For the angle, the Gromacs code is modified and a multi-well angle potential is added (shape like
\111_11_111/ )
1\_/1\_/1\_/
The well bottoms correspond to ideal angles of different seondary elements..............
The Go-like potential we used require waters in the system (If fact, we just need Martini + pair potential represents nature contact in currest stage, although it would be great if the BB can be put in Ca position and Secondary biases can be removed which is necessary for next stage )
Till now, all the effert above could hold helix stable, but beta structure is lost...
However, based on disscussion above, it seems it is fine to keep the BB-SC1-SC* untouched.
By the way, could you tell me the logic behind the BB-SC1-SC2 and BB-SC1-SC3 angle of Trp?
I have no idea why use the 210 for BB-SC1-SC2, should this equal to 150 degree? However, based on statistic from PDB, it seems this angle should be 130...
Last edit: 9 years 11 months ago by jsbach.
Please Log in or Create an account to join the conversation.
Time to create page: 0.108 seconds