I am trying to model Opa3 which is a protein in the mitochondria. There is evidence showing that when this protein is mutated there is a build up of 3-methylglutaconic aciduria, type 3 (MGA-3). We are trying to find a ligand that is in the pathway of production/breakdown of this compound so we can mutate the amino acids the ligand would bind to. Because of the unique conditions in the mitochondria I don't really know where to start in predicting a stable model and a ligand binding site. Any suggestions or advice on what tools or programs to use?
I'm not sure what question you are asking - you want to "find a ligand" - you don't know what ligand it is, or you don't know where it binds?
I'm vague on what's different about the mitochondria - I assume just pH? There is some "pH mode" code in Rosetta but I don't know its condition or if it's released.
I'm also unsure of what difference you mean regarding the "unique conditions in the mitochondria" - but Rosetta isn't really fine-tuned to any particular conditions. "Physiological" is about as specific we get as to what conditions Rosetta can model. Unless you have a specific situation you think is critical for modeling success, I'd probably recommend trying regular Rosetta protocols out, and seeing if the results are sensible.
Regarding what you're trying to accomplish, there are two or three general types of ligand docking runs. The first is having a given protein, and then docking a large number of compounds against it to see which ones binds (usually termed "virtual screening"). The first-and-a-half is the opposite, having a given compound, and then docking against multiple proteins to see which proteins it binds to (also called virtual screening). Rosetta isn't very good at this, mainly because each protein/ligand pair takes too long to run. It's workable with a few dozen/hundred compounds/proteins, but when you get into the thousands to hundreds of thousands compounds virtual screening is typically used for, it becomes onerous. If that's your use case, you'd probably be better off with a program made for virtual screening. (Something like Autodock.)
The second/third type of docking experiment is where you know which ligand and which protein you're docking, but you want to know where and how the ligand is bound. For that Rosetta does work. Take a look at the ligand docking application http://www.rosettacommons.org/manuals/archive/rosetta3.4_user_guide/d4/d... (and the related literature references) for a description of how to do it.
I second what Rocco says on conditions - just try the regular scorefunction first and see if it works before trying something harder.
Thanks for your help. sorry for confusing everyone, my situation is that I have a protein that I need to first predict the structure, and second try to dock a select few ligands from a pathway to see if any of these ligands bind and where. I was first asking if there was a special way i had to predict a protein structure because of the pH difference in the mitochondria. Then I want to try to find a target site for a select few number of ligands so we can have an idea of where the ligands would bind. Then with this knowledge we will see what happens when we mutate the amino acids that the ligand/s bind to. I know that pH can play a part in changing an amino acid's pI, which can also change the way a protein folds so I was wondering if I had to do anything differently when modeling my protein located in the mitochondria.
You're way out on the bleeding edge here - I would say a project like that would be tough for a programmer/developer writing a mode specifically for it, much less someone just using the existing code.
Go ahead and try it with default Rosetta and see if what you get is consistent with your experimental knowledge of the system - presumably you know something which can be used to constrain and filter the results.
I've queried the author of the pH code for comment but I don't know that the code is officially released.
Here is what the pH author said:
"The basic protonation states are available for use in Rosetta3.4. I later fine-tuned my protocol for pKa predictions. I'll check in the refinements sometime in the next couple of weeks but the calibrations might not be very useful as he is trying to model folding at a different pH (which I didn't aim for).
To turn on variable protonation states, you can use the flag -pH:pH_mode and then specify the pH using -pH:value_pH 7.2 (for example). This will add extra protonation states every time the packer is called. The score function passed over to the packer need to have a non zero weight on the e_pH score term (which is the pH energy term). The required params files in /residue_types/protonation_states/ need to included.
I haven't tried to calibrate the score function for folding, so I can't comment on the score function. However, the standard or score12 weights with additional e_pH and hack_elec terms (weights of 1.0 each) would be a good place to start.
Please ask him to contact me directly if he has any more questions.