I have a scenario where a protein is composed of two domains (its actually two chains, but I can make it a single chain) for Rosetta's purposes.
There is a binding pocket at the interface of these domains, and it is known that the domains "spread open" slightly upon ligand binding. I am trying to figure out how to compose a ligand docking simulation that (a) allows domain translational and rotational flexibilty, and (b) backbone perterbations and side chain optimizations, when the ligand is bound.
I have been able to do (b) with the attchached RosettaScripts XML, however, I can't figure out how to allow a domain to have "free" translation annd rotation flexibility to allow for the simulation to separate the two domains upon binding the ligand.
Does anyone have any pointers on how to simulate ligand docking while allowign two domains to move/rotate as a domain entity?
Unfortunaltey, there are no known ligand bound crystal structures, otherwise I would start with that PDB pose.
Thanks!
Brett
| Attachment | Size |
|---|---|
 rosettascripts_docking.txt | 1.84 KB |
Doing "induced fit" style docking with Rosetta is still an unsolved problem. (It's a general unsolved problem for most ligand docking programs.)
The best you can get is probably a "conformational selection" approach. That is, pregenerate an ensemble of models which show a range of flexibility in the relevant backbone structures. You then do separate "lock and key" style docking across the ensemble. (You may be able to add in flexible backbone repacking minimization once the ligand is bound to further optimize things.) You can then (hopefully) pick out the backbone conformations which result in best binding of the ligand.
Regarding the ensemble generation, the techinque you use will really depend on what sort of movement/differences you expect to see in your induced fit docking. In your case I'd probably go with a protein-protein docking approach. That is, do a regular Rosetta protein-protein docking run (in the abscence of the ligand). But instead of picking a single "best" conformation to bring forward to ligand docking, you should take an ensemble of produced structures. - You can filter out the structures with really bad scores and use clustering to remove practical duplicates, but you should be aiming to take several dozen structures into ligand docking. (Potentially more, depending on how flexible the backbone is.) Since you know a little bit about how things should change (the pocket opening up), you can incorporate that information into your filtering step. Only take forward into docking those structures which have an "opened" binding pocket (however you choose to measure it).