- Posts: 23
Fitting angle parameters
- P_Banerjee
-
Topic Author
- Offline
- Junior Boarder
Less
More
8 years 7 months ago #4838
by P_Banerjee
Fitting angle parameters was created by P_Banerjee
Hello users,
I have been trying to reproduce CG angle distributions from the atomistic ones. I have two angles defined in my CG structure, A-B-A and B-A-B. Despite many trials, the angle distributions do not average around the provided equilibrium value. I was wondering if the number of exclusions has a role to play. A-B-A is off by -10 degrees while B-A-B is off by +10 degrees, while the radius of gyration of the CG polymer chain is close to that of the atomistic.
Looking forward to valuable suggestions!
Thank you.
I have been trying to reproduce CG angle distributions from the atomistic ones. I have two angles defined in my CG structure, A-B-A and B-A-B. Despite many trials, the angle distributions do not average around the provided equilibrium value. I was wondering if the number of exclusions has a role to play. A-B-A is off by -10 degrees while B-A-B is off by +10 degrees, while the radius of gyration of the CG polymer chain is close to that of the atomistic.
Looking forward to valuable suggestions!
Thank you.
Please Log in or Create an account to join the conversation.
- Pim
-
- Offline
- Expert Boarder
Less
More
- Posts: 105
8 years 7 months ago #4843
by Pim
Replied by Pim on topic Fitting angle parameters
Hello,
In principle, the number of exclusions is indeed a factor that plays a role in all MD simulations. It is, however, not a parameter we often tamper with. For polymers in Martini, we've typically used nr_excl = 1, only excluding the nearest neighbours. If for some reason in your polymer the B-beads attract each other strongly through non-bonded interactions, this may lead to an angle that is somewhat too small. In that case, perhaps setting the nr_excl to 2 may be useful, but be careful and check if all properties of your system are still accurately represented.
On another note: Can you tell what is preventing your beads from reaching the correct angle? It is a little bit unusual you cannot reach the equilibrium value unless there are steric interactions that prevent the system from reaching it. Additionally, I hope you have double-checked the placement of your beads (are they on the center of mass for example, and do you also calculate you atomistic angles from the center of mass?). Also, a 10 degree deviation is really not too bad in terms of agreement for some systems.
In principle, the number of exclusions is indeed a factor that plays a role in all MD simulations. It is, however, not a parameter we often tamper with. For polymers in Martini, we've typically used nr_excl = 1, only excluding the nearest neighbours. If for some reason in your polymer the B-beads attract each other strongly through non-bonded interactions, this may lead to an angle that is somewhat too small. In that case, perhaps setting the nr_excl to 2 may be useful, but be careful and check if all properties of your system are still accurately represented.
On another note: Can you tell what is preventing your beads from reaching the correct angle? It is a little bit unusual you cannot reach the equilibrium value unless there are steric interactions that prevent the system from reaching it. Additionally, I hope you have double-checked the placement of your beads (are they on the center of mass for example, and do you also calculate you atomistic angles from the center of mass?). Also, a 10 degree deviation is really not too bad in terms of agreement for some systems.
Please Log in or Create an account to join the conversation.
- P_Banerjee
-
Topic Author
- Offline
- Junior Boarder
Less
More
- Posts: 23
8 years 7 months ago #4845
by P_Banerjee
Replied by P_Banerjee on topic Fitting angle parameters
Hey,
I'll give you a more detailed description of system. My system is coarse-grained Polyacrylamide in water. The B beads are of type SC1 and represent the backbone while the A beads are of type P5 and represent the pendant amide groups. I have been using number of exclusions as 2; so this should mean the A beads of angle A-B-A do not have a non-bonded interaction between them. Despite that, the angle A-B-A is smaller by 10 degrees from the equilibrium angle. I am in a tight spot now.
Yes, I have made sure that I pick the centers of mass of the groups of atoms to represent the beads. Do you think fitting a different angle potential, let's say, quartic angle potential would help? I ask because my angle distributions from the atomistic are bimodal.
Thank you for your time. :)
I'll give you a more detailed description of system. My system is coarse-grained Polyacrylamide in water. The B beads are of type SC1 and represent the backbone while the A beads are of type P5 and represent the pendant amide groups. I have been using number of exclusions as 2; so this should mean the A beads of angle A-B-A do not have a non-bonded interaction between them. Despite that, the angle A-B-A is smaller by 10 degrees from the equilibrium angle. I am in a tight spot now.
Yes, I have made sure that I pick the centers of mass of the groups of atoms to represent the beads. Do you think fitting a different angle potential, let's say, quartic angle potential would help? I ask because my angle distributions from the atomistic are bimodal.
Thank you for your time. :)
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
8 years 7 months ago #4846
by mnmelo
Replied by mnmelo on topic Fitting angle parameters
Hi,
Some notes:
First of all, you're definitely overmapping the polyacrylamide's 5 backbone to two martini beads. This might get you in trouble down the line since Martini bead behavior is parameterized for mappings closer to 4-to-1, with consequent longer inter-bead bonds. A better bet might be to have a 5-to-1 single backbone bead. If directionality is important you may try playing with bonded parameters so that backbone beads align in a zig-zag pattern.
Regarding the problems you mention with angle potentials: since you're using very tight mappings it's likely that backbone bead i lies in the repulsive distance of bead i+2. This, of course, makes the fine tuning of the angle i,i+1,i+2 difficult since you must increase the angle potential to compensate the particle repulsion. Instead of the drastic measure of excluding all second neighbors GROMACS also lets you exclude specific pairs, and that's probably the best if you carry on with the fine mapping.
Good luck.
Some notes:
First of all, you're definitely overmapping the polyacrylamide's 5 backbone to two martini beads. This might get you in trouble down the line since Martini bead behavior is parameterized for mappings closer to 4-to-1, with consequent longer inter-bead bonds. A better bet might be to have a 5-to-1 single backbone bead. If directionality is important you may try playing with bonded parameters so that backbone beads align in a zig-zag pattern.
Regarding the problems you mention with angle potentials: since you're using very tight mappings it's likely that backbone bead i lies in the repulsive distance of bead i+2. This, of course, makes the fine tuning of the angle i,i+1,i+2 difficult since you must increase the angle potential to compensate the particle repulsion. Instead of the drastic measure of excluding all second neighbors GROMACS also lets you exclude specific pairs, and that's probably the best if you carry on with the fine mapping.
Good luck.
Please Log in or Create an account to join the conversation.
- P_Banerjee
-
Topic Author
- Offline
- Junior Boarder
Less
More
- Posts: 23
8 years 7 months ago #4847
by P_Banerjee
Replied by P_Banerjee on topic Fitting angle parameters
Hey!
So the mapping I follow is thus:
The backbone bead is formed from the center of mass of three carbons from the chain: the carbon immediately preceding the monomer + the two backbone carbons of the monomer itself. This way it is a 3:1 mapping for the backbone bead. My atomistic force-field is GROMOS, so there are no hydrogens on the backbone carbons.
If the above description fails to put forth my point, I would request you to have a look at this paper pubs.rsc.org/en/Content/ArticleLanding/2...M00481B#!divAbstract . I have followed the A-mapping procedure for my system.
Suggestions are welcome, rather I am in dire need of them .
Thank you for having taken the pain to read and respond.
So the mapping I follow is thus:
The backbone bead is formed from the center of mass of three carbons from the chain: the carbon immediately preceding the monomer + the two backbone carbons of the monomer itself. This way it is a 3:1 mapping for the backbone bead. My atomistic force-field is GROMOS, so there are no hydrogens on the backbone carbons.
If the above description fails to put forth my point, I would request you to have a look at this paper pubs.rsc.org/en/Content/ArticleLanding/2...M00481B#!divAbstract . I have followed the A-mapping procedure for my system.
Suggestions are welcome, rather I am in dire need of them .
Thank you for having taken the pain to read and respond.
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
8 years 7 months ago #4848
by mnmelo
Replied by mnmelo on topic Fitting angle parameters
Well, it doesn't change the fact that you end up with a 5(heavy atoms)-to-2-beads mapping. If you have no specific reason to want that much detail, really, don't do it. See how when such a fine mapping was used
for PEG
a new particle type had to be created to account for partition imbalances. This is probably what awaits you.
Look at the Martini protein backbone for alanine. It's a 5-to-1 mapping right there. I'd definitely start with something like this and only go for finer detail if 1) I know I'll need it for some reason; or 2) it ends up not behaving as I expect in my application.
Good luck
Look at the Martini protein backbone for alanine. It's a 5-to-1 mapping right there. I'd definitely start with something like this and only go for finer detail if 1) I know I'll need it for some reason; or 2) it ends up not behaving as I expect in my application.
Good luck
Please Log in or Create an account to join the conversation.
- Pim
-
- Offline
- Expert Boarder
Less
More
- Posts: 105
8 years 7 months ago #4850
by Pim
Replied by Pim on topic Fitting angle parameters
If your system is polyacrylamide -[CH2-CH(CONH2)]- and you have a separate backbone bead and a side chain bead, you should not even have an ABA and BAB angle?! You only have backbone-backbone-sidechain (BBA) angles, right?
The problems with the big side chain beads and the small backbone beads will remain however and I agree with Manel that a coarser mapping seems more intuitive.
The problems with the big side chain beads and the small backbone beads will remain however and I agree with Manel that a coarser mapping seems more intuitive.
Please Log in or Create an account to join the conversation.
- P_Banerjee
-
Topic Author
- Offline
- Junior Boarder
Less
More
- Posts: 23
8 years 7 months ago #4853
by P_Banerjee
Replied by P_Banerjee on topic Fitting angle parameters
I totally understand the confusion regarding the mapping scheme. It took me a while to comprehend the A-mapping procedure followed by this paper.
pubs.rsc.org/en/Content/ArticleLanding/2...M00481B#!divAbstract
The supporting information for this paper shows the mapping scheme very clearly. Please do have a look. www.sciencedirect.com/science/article/pii/S0021979712007837
I have followed the very same procedure as theirs.
Now, I was thinking, since I obtain bimodal distributions for the angles, would it be fine to ignore the smaller peak of the distribution and perform a fitting?
Thank you for your kind replies.
The supporting information for this paper shows the mapping scheme very clearly. Please do have a look. www.sciencedirect.com/science/article/pii/S0021979712007837
I have followed the very same procedure as theirs.
Now, I was thinking, since I obtain bimodal distributions for the angles, would it be fine to ignore the smaller peak of the distribution and perform a fitting?
Thank you for your kind replies.
Please Log in or Create an account to join the conversation.
- Pim
-
- Offline
- Expert Boarder
Less
More
- Posts: 105
8 years 7 months ago #4854
by Pim
Replied by Pim on topic Fitting angle parameters
I see now, sorry for not reading the paper the first time.
Indeed, just like Rossi et al and Wang et al in their papers you've linked to, they often have bimodal angle distributions, but choose to represent these by a unimodal CG potential. Note however that they do not IGNORE the 2nd peak, they fit their potential to the whole distribution (unless your 2nd mode is really, really small). This gives a broad potential energy well, so that all the conformations that are observed atomistically can in principle be sampled in the CG simulation, although the minima will not be in the right place.
If you really can't reproduce the results from Wang et al, you may want to ask if they can share their itp file with you.
Indeed, just like Rossi et al and Wang et al in their papers you've linked to, they often have bimodal angle distributions, but choose to represent these by a unimodal CG potential. Note however that they do not IGNORE the 2nd peak, they fit their potential to the whole distribution (unless your 2nd mode is really, really small). This gives a broad potential energy well, so that all the conformations that are observed atomistically can in principle be sampled in the CG simulation, although the minima will not be in the right place.
If you really can't reproduce the results from Wang et al, you may want to ask if they can share their itp file with you.
Please Log in or Create an account to join the conversation.
- P_Banerjee
-
Topic Author
- Offline
- Junior Boarder
Less
More
- Posts: 23
8 years 7 months ago #4856
by P_Banerjee
Replied by P_Banerjee on topic Fitting angle parameters
Yeah, I'll do that.
One last question. Wang et al have used P5 to represent the amide(-CONH2) bead which has 3 heavy atoms. Wouldn't SP5 be a better choice, or rather SNda since the amide participates in hydrogen bonding with oxygen as the acceptor and H in N-H as the donor?
One last question. Wang et al have used P5 to represent the amide(-CONH2) bead which has 3 heavy atoms. Wouldn't SP5 be a better choice, or rather SNda since the amide participates in hydrogen bonding with oxygen as the acceptor and H in N-H as the donor?
Please Log in or Create an account to join the conversation.
Time to create page: 0.117 seconds