Hi all,
I've recently started to use the Rosetta suite for some basic modeling of an orphan GPCR. Our research group has discovered several compounds that interact with the membrane protein, either directly or indirectly. To see if any of these compounds could act as ligand, I turend to the ligand-dockng protocol. I went through the tutorial (with some issues) however the results were succesful. Now that I am trying to do it with my orphan GPCR, i've run into some issues:
It seems that the tutorial uses a crystal_complex file of the crystallized structure of the receptor and antagonist to score the models, as well to align the antagonist to the binding pocket. My orphan receptor has no crystallized strucure, however I have a decent predicted model. I do not know where the ligands could possibly dock to so my questions are:
(1) Can I still use this protocol to predict potential protein-ligand interaction without a known crystallized strucure, and if so;
(2) How would I proceed to try and model these protein-ligand interactions?
I'm very new to this with just basic terminal/python knowledge, so I apologize if my questions are too basic.
Thanks in advance,
Marcus
The crystallized receptor shouldn't be necessary for the ligand docking. You can omit it (remove it from the tag) without any harm to the protocol. (You won't get an rmsd to native in the scorefile, but that's not necessary.)
You *will* need an idea about where the ligand will bind, though. Rosetta does not currently have binding site detection, so you'll have to provide the starting points yourself. However, you can have *multiple* binding sites, and dock into all of them. So what I'd recommend is to make something like a grid of potential binding sites - that is, find a region(s) of the protein you think the ligand will bind to, and then place starting points regularly, every 3-5 Ang or so. You can then specify in the Transform mover to do an initial perturb of 5 Ang to sample around that starting position. (There was an issue with this randomization that was actually fixed in the past week or so - I'd recommend downloading the most recent weekly release 2016.40 or later.)
There's two approaches in sampling the multiple starting locations. One way is to set up multiple independent runs, each with a different starting location. This will work and is very controllable, but requires a fair amount of setup. The other option is to use the PDB subtag of the StartFrom mover: https://www.rosettacommons.org/docs/latest/scripting_documentation/RosettaScripts/Movers/movers_pages/StartFromMover (This was actually added to scan differnt binding sites of GPCRs.) This takes a name of a PDB file with atoms (typically I use water oxygens) at each of the possible starting positions. When you run the protocol, the StartFromMover will pick a random starting position, move the ligand to that position, and then run the docking. (You need to have the ligand present in the input file, but it doesn't need to be in a particular location.) Make sure when you do this to increase the total number of structures (-nstruct) according to the number of binding sites, such that each starting site gets (on average) the standard number of output structures.
Analysis of a multi-site docking run is the same as a single site dock - just combine all the results from the different sites into one pool, and analyze as you would normally.