- Posts: 10
Elastic network to just one domain of the protein
- Prunotto
-
Topic Author
- Offline
- Fresh Boarder
Less
More
8 years 8 months ago #4762
by Prunotto
Elastic network to just one domain of the protein was created by Prunotto
Dear all,
I have one question that might be trivial, but I cannot find a solution.
I have a protein that can be divided into 3 domains: A, B and C. For what concerns the domains A and C, I am quite sure about their conformation, while the conformation of the domain B is still uncertain.
My question is: is it possible to apply one of the elastic networks (either the classical elastic network or ElneDyn, is the same) to the domains A and C, but let the domain B free to fluctuate?
Looking forward to receiving your reply
All the best,
Alessio
I have one question that might be trivial, but I cannot find a solution.
I have a protein that can be divided into 3 domains: A, B and C. For what concerns the domains A and C, I am quite sure about their conformation, while the conformation of the domain B is still uncertain.
My question is: is it possible to apply one of the elastic networks (either the classical elastic network or ElneDyn, is the same) to the domains A and C, but let the domain B free to fluctuate?
Looking forward to receiving your reply
All the best,
Alessio
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
8 years 8 months ago - 8 years 8 months ago #4763
by mnmelo
Replied by mnmelo on topic Elastic network to just one domain of the protein
Hello Alessio,
Two points on your problem:
1. You want to leave domain B free to fluctuate because you're not sure about its conformation.
Well, this is problematic, because, as you know, Martini can not spontaneously reproduce secondary structure. That's why you must enforce known secondary structure using either elastic networks or bond/angle/dihedral restrictions.
Leaving B to fluctuate (which really means to assign to it the least restrictive set of Martini bonded potentials for proteins, designed for random coil segments) can lead to meaningless results. You're better off using atomistic simulations to get an idea of the structural propensities of B, and then enforce those in Martini.
2. The two types of secondary structure enforcement (elastic networks and bonded restrictions) can be combined. At least in theory. In practice this will involve creating two topologies, one with elastic networks and the other with bonded restrictions, and then manually cutting and pasting a lot of potentials, and pruning unwanted elastic bonds. Given that you probably want to do this for the wrong reasons (see 1.) I'd say, don't.
Good luck,
Manel
Two points on your problem:
1. You want to leave domain B free to fluctuate because you're not sure about its conformation.
Well, this is problematic, because, as you know, Martini can not spontaneously reproduce secondary structure. That's why you must enforce known secondary structure using either elastic networks or bond/angle/dihedral restrictions.
Leaving B to fluctuate (which really means to assign to it the least restrictive set of Martini bonded potentials for proteins, designed for random coil segments) can lead to meaningless results. You're better off using atomistic simulations to get an idea of the structural propensities of B, and then enforce those in Martini.
2. The two types of secondary structure enforcement (elastic networks and bonded restrictions) can be combined. At least in theory. In practice this will involve creating two topologies, one with elastic networks and the other with bonded restrictions, and then manually cutting and pasting a lot of potentials, and pruning unwanted elastic bonds. Given that you probably want to do this for the wrong reasons (see 1.) I'd say, don't.
Good luck,
Manel
Last edit: 8 years 8 months ago by mnmelo.
Please Log in or Create an account to join the conversation.
- Prunotto
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 10
8 years 8 months ago #4766
by Prunotto
Replied by Prunotto on topic Elastic network to just one domain of the protein
Hi Manel,
thank you very much for your reply.
I probably explained it wrong, but what happens in my system is that I HAVE the secondary structure also for what concerns the domain B.
However, such domain should in theory be rotated of about 90 degrees with respect to the model that I obtained. So, basically it is the tertiary structure to be wrong, not the secondary structure.
As a matter of fact, I thought that the secondary structure could be imposed with DSSP, while the elastic networks were meant to maintain the tertiary structure. Isn't it like that?
Looking forward for your reply
Best,
Alessio
thank you very much for your reply.
I probably explained it wrong, but what happens in my system is that I HAVE the secondary structure also for what concerns the domain B.
However, such domain should in theory be rotated of about 90 degrees with respect to the model that I obtained. So, basically it is the tertiary structure to be wrong, not the secondary structure.
As a matter of fact, I thought that the secondary structure could be imposed with DSSP, while the elastic networks were meant to maintain the tertiary structure. Isn't it like that?
Looking forward for your reply
Best,
Alessio
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
8 years 8 months ago - 8 years 8 months ago #4767
by mnmelo
Replied by mnmelo on topic Elastic network to just one domain of the protein
Ok, so you're in a better situation than I understood.
Just to clear things up, since what you say about secondary and tertiary Martini structures isn't accurate:
Martini can't reproduce secondary structure -- it depends on too-fine backbone-backbone interactions -- and that's what you have to restrain. This can be done with an elastic network or what you call "DSSP", which is in fact a set of bonded potentials intended to keep the secondary structure that DSSP identifies. The latter option is best suited for smaller, alpha-helical proteins, as it doesn't do so good a job with beta-sheet elements -- I therefore recommend you use an elastic network for your multi-domain protein. In any case, the goal here is to restrain the secondary structure.
Tertiary structure involves more sidechain-sidechain interactions, and Martini is better at getting those right. In practice, however, correct tertiary structure can also be difficult to precisely achieve. You can help Martini by keeping an elastic network also between secondary structure elements that shouldn't move apart (say, if your domain A never falls apart it's ok to keep it internally elastically networked). This is something that only elastic networks can help with. Bonded restraints will do very little to bias the tertiary structure.
So, what should you do with your system?
I'd definitely use an elastic network, and start by finding an appropriate elastic network cutoff. At the very least you'll want your alpha-helices and beta-sheets to be internally restrained.
You can then increase the cutoff to make sure each of your domains is internally secure. If the interaction between A and C is also very well-defined and stiff, you may further increase the cutoff to have A and C held at their correct relative positions (I'd be wary here, and wouldn't do this. I'd resort to elastically binding domains to each other only if the dynamics show they drift apart in an unrealistic way).
After you have an elastic network that binds all that needs to be bound, you must prune it by hand to remove links between elements that should be free. Remove any elastic network links between B and the other domains.
Also, it is important to keep the linker segments between domains unbiased, and therefore unlinked.
There is no easy scripted way to do this pruning, I'm afraid. An iterative, targeted method has been published to refine an elastic network so as to reproduce proper structural fluctuation ( Globisch, C. T., Krishnamani, V., Deserno, M., & Peter, C. (2013). Optimization of an elastic network augmented coarse grained model to study CCMV capsid deformation. ), but I'm not sure how applicable it is to your case (while possibly being quite an overkill).
Finally, you say that the model you obtain has domain B 90º off. I don't quite understand you here:
-Is it that B drifts 90º from where it should when you simulate it? Then that means you do know how its tertiary structure should behave. If you find Martini doesn't get it right you have some rationale to elastically network B to some other part of the protein.
-Or is it that your initial configuration has B in a wrong position? In this case you can either try the pruned elastic network approach above, and hope sidechain-sidechain interactions get it right, or maybe force that initial configuration to better match what you expect.
(Both these points raise, however, the question of how you got that initial conformation, and how you know what the correct position of B is.)
Once more, good luck!
Manel
Just to clear things up, since what you say about secondary and tertiary Martini structures isn't accurate:
Martini can't reproduce secondary structure -- it depends on too-fine backbone-backbone interactions -- and that's what you have to restrain. This can be done with an elastic network or what you call "DSSP", which is in fact a set of bonded potentials intended to keep the secondary structure that DSSP identifies. The latter option is best suited for smaller, alpha-helical proteins, as it doesn't do so good a job with beta-sheet elements -- I therefore recommend you use an elastic network for your multi-domain protein. In any case, the goal here is to restrain the secondary structure.
Tertiary structure involves more sidechain-sidechain interactions, and Martini is better at getting those right. In practice, however, correct tertiary structure can also be difficult to precisely achieve. You can help Martini by keeping an elastic network also between secondary structure elements that shouldn't move apart (say, if your domain A never falls apart it's ok to keep it internally elastically networked). This is something that only elastic networks can help with. Bonded restraints will do very little to bias the tertiary structure.
So, what should you do with your system?
I'd definitely use an elastic network, and start by finding an appropriate elastic network cutoff. At the very least you'll want your alpha-helices and beta-sheets to be internally restrained.
You can then increase the cutoff to make sure each of your domains is internally secure. If the interaction between A and C is also very well-defined and stiff, you may further increase the cutoff to have A and C held at their correct relative positions (I'd be wary here, and wouldn't do this. I'd resort to elastically binding domains to each other only if the dynamics show they drift apart in an unrealistic way).
After you have an elastic network that binds all that needs to be bound, you must prune it by hand to remove links between elements that should be free. Remove any elastic network links between B and the other domains.
Also, it is important to keep the linker segments between domains unbiased, and therefore unlinked.
There is no easy scripted way to do this pruning, I'm afraid. An iterative, targeted method has been published to refine an elastic network so as to reproduce proper structural fluctuation ( Globisch, C. T., Krishnamani, V., Deserno, M., & Peter, C. (2013). Optimization of an elastic network augmented coarse grained model to study CCMV capsid deformation. ), but I'm not sure how applicable it is to your case (while possibly being quite an overkill).
Finally, you say that the model you obtain has domain B 90º off. I don't quite understand you here:
-Is it that B drifts 90º from where it should when you simulate it? Then that means you do know how its tertiary structure should behave. If you find Martini doesn't get it right you have some rationale to elastically network B to some other part of the protein.
-Or is it that your initial configuration has B in a wrong position? In this case you can either try the pruned elastic network approach above, and hope sidechain-sidechain interactions get it right, or maybe force that initial configuration to better match what you expect.
(Both these points raise, however, the question of how you got that initial conformation, and how you know what the correct position of B is.)
Once more, good luck!
Manel
Last edit: 8 years 8 months ago by mnmelo. Reason: Fix URL
Please Log in or Create an account to join the conversation.
- Prunotto
-
Topic Author
- Offline
- Fresh Boarder
Less
More
- Posts: 10
8 years 8 months ago #4768
by Prunotto
Replied by Prunotto on topic Elastic network to just one domain of the protein
Thank you so much for your reply, it was really very complete and helpful!
Yes, actually I am in the second situation, so I will definitely try the solution that you suggested!
Just one last question to make sure that I understood correctly: so, if I use an elastic network, the secondary structure will be maintained even without specifying the secondary structure itself with DSSP?
Thank you so much!
Alessio
Yes, actually I am in the second situation, so I will definitely try the solution that you suggested!
Just one last question to make sure that I understood correctly: so, if I use an elastic network, the secondary structure will be maintained even without specifying the secondary structure itself with DSSP?
Thank you so much!
Alessio
Please Log in or Create an account to join the conversation.
- mnmelo
-
- Offline
- Admin
Less
More
- Posts: 89
8 years 8 months ago #4769
by mnmelo
Replied by mnmelo on topic Elastic network to just one domain of the protein
Yes, the elastic network will take care of keeping the secondary structure on its own. Just take a look at the resulting structure+network (with any viewer that allows many bonds per atom -- VMD is notoriously annoying in that respect) and you'll see the system can't really depart from that secondary structure.
HOWEVER, this does not mean you can do away with DSSP: you will still need to feed structure information into the martinization process so that bead types get properly assigned. As you'll know, from having read the Martini protein paper, different secondary structures require different backbone bead types, and the only way that can be incorporated into the topology is if structural information is known.
And that's it. Have fun!
Manel
HOWEVER, this does not mean you can do away with DSSP: you will still need to feed structure information into the martinization process so that bead types get properly assigned. As you'll know, from having read the Martini protein paper, different secondary structures require different backbone bead types, and the only way that can be incorporated into the topology is if structural information is known.
And that's it. Have fun!
Manel
Please Log in or Create an account to join the conversation.
Time to create page: 0.222 seconds