You are here

conformational sampling of multi domain protein

5 posts / 0 new
Last post
conformational sampling of multi domain protein

Hi all

       I have a protein with two doamin connected with two linker. the crystal structure shows only closed conformation. I don't know the open conformation and want to sample it. so I want to only change the conformation of linker region treating the two domain as rigid. is there any protocol available on rosetta that I can use for this purpose? Sorry if I haven't put my question in a very clear way. I am new to rosetta and not well aware of most of the terminology used here.


Thank you


Post Situation: 
Tue, 2017-07-04 11:20


Poke through the documentation and demo and let us know if you have further questions afterwards.


Your exact problem isn't clear - you say you have two domains with two linkers, which doesn't make a ton of sense - do you mean both termini are in one domain?  In that case we don't really have a protocol for what you want to do exactly; I'd probably start with unconstrained relax ensembles, but it won't be very efficient.

Tue, 2017-07-04 11:22

thank you

Yes. both termini are in one domain. 



Tue, 2017-07-04 11:30

Well.  We don't have a protocol that addresses this that I am aware of.  It's essentially a domain insertion problem, but I don't think any of the domain insertion tools work very well in Rosetta.  Think of it as domain B being inserted into pre-existing domain A, where A is the one with both termini.  Your difference is only that it already exists and you want to sample it.

How "open" is your expected open state?  Do you know anything about it?  Do you know how far it is from the closed state?  Do you know any relative residue positions that can be used to bias sampling?

One option is to try some combination of relax and loop modeling, both in as unconstrained a state as possible.  If the loops are fairly long you might be able to get wide sampling this way.  You'd want to do loop modeling, select widely varied loops, and relax on those to see if it opens up the domains.

Another option is to use an undocumented mode of AnchoredDesign.  You can set the problem up with the "target protein" being way out in space, and have your entire domain B be the "anchor", and domain A be the "scaffold".  In thise case, AnchoredDesign will treat the entire loop1 - domain B - loop2 region as one loop, except holding the internal coordinates of domain B fixed so that it won't unfold.  It will then sample the loop1-loop2 region via loop modeling.  I've done this before (I wrote AnchoredDesign) and can help you set it up if you want to try it.


I suspect Remodel can do domain insertion remodeling - Remodel can do just about anything, if you know how to use it.  I do not.  (,


The Topology Broker almost certainly has a way to do your problem as well

Tue, 2017-07-04 12:20

Yes, the Broker can definitely do something like this. As often happens, there are some questions about exactly what you want Rosetta to do.

As Steven alluded, with the Broker, you can definitely do something like this. For example, you could hold the two domains A and B fixed relative to themselves, and sample the linker with a FragmentMoverCM. Because the insertion of a fragment will cause all the downstream amino acids to move in cartesian space relative to the upstream amino acids, this approach will probably wind up fully sampling the rigid body degrees of freedom between your two domains as the downstream domain swings around wildly with every fragment insertion. Is this what you want? If you think that the domains are more like "beads on a string" in might very well be. If not, you'll need to be more specific about how you want the rigid body degrees of freedom to be sampled.

If you choose to go this way, you'll want to define one RigidChunkCMs for each of your two domains, then make a FragmentCM to insert fragments across the whole protein (you can use the Robetta server to get fragments). The RigidChunkCMs will prevent fragment insertion within the domains (or in whatever areas you choose) leaving only the linker to actually see changes in torsional angles.

There are a bunch of examples in the Broker paper: as well as in the docs that Steven linked to: 

Feel free to post with further questions. :)

Wed, 2017-07-05 06:30