You are here

Allow region of protein to move in a certain direction

4 posts / 0 new
Last post
Allow region of protein to move in a certain direction


I have a protein structure. I want to move a specified region (say residues x to y) of this protein in a any direction (more like a dynamic simulation). Is there a protocol that allows this in Rosetta? I have tried the backrub, relax with constraints.

Post Situation: 
Wed, 2017-08-23 17:07

I don't quite understand what you're after. How do you want the protein region to move? What protocol are you attempting to use? What's the overall goal of the modeling run?

There's a number of different ways of limiting changes to a particular region of the protein in Rosetta. Constraints are one, and using resfiles and movemaps are another. Which to choose depends on what protocol you're using (different protocols take different inputs), and what you're hoping to achieve.

Thu, 2017-08-24 08:21

Thanks for the reply. I am specifically looking for a protocol that can perturb the structure from the left to create the structure similar to that on the right in the attached picture. I tried the movemap with relax protocol and backrub with resfile. But, I could not get them to perturb the structure in the manner that I am looking for.

File attachments: 
Thu, 2017-08-24 09:48

I know of at least one person in the Rosetta developer community who is working on a protocol tailored to do something very similar to this -- however, the protocol is still in the development stage and hasn't yet been published.


In the meantime, the approach I probably would suggest would be to do relax with constraints, in an interative fashion. Start with an unconstrained relax, allowing the structure to move where it wants. Now take a look at the relaxed structures, looking for where the greatest deviations are between the outputs and where you want it to be. (I would suggest using either atom pair distances or dihedral angle values, and would recommend against using coordinates.)  Add a few constraints for those residues with the largest deviation to what you want, bringing them more toward the value that you desire. Then re-run the relax protocol (on the output structures), but with the constraints turned on. Repeat this, adding constraints each time until you get structures of the type you want.  You will likely need to play around with things in order to determine the best constraints to add, and how strongly you turn them on.

I'd probably suggest against a Movemap, as that fixes the structure of locations which you turn off, but doesn't necessarily apply any bias toward the structure you want.

I'd also say that a pure backrub protocol is unlikely to be useful, as the goal of backrub is to make local backbone perturbations without major rearrangements. The sort of change you're looking for is much larger than what backrub is likely to give you. (That's not to say that you can't use backrub with other sampling methods, but I wouldn't use it by itself.)


But I'd also suggest stepping back a minute to ask why you want to make the perturbation. Is your only goal to get Protein A in a structure that's similar to Protein B? If you're just concerned about the endpoint (and not any intermediates), I'd suggest you look instead to RosettaCM homology modeling. This will allow you to take a particular protein sequence (Protein A) and model its structure using at template that should be structually similar (Protein B). 


Thu, 2017-08-24 11:00