I have performed protein-protein docking and have got 30 different models from ClusPro. I do not know whether they are accurate. Can anyone help?
I haven't used ClusPro myself. As you can see in the link below, cluspro is comparable to ZDOCK, with which I had lots of experience and good fortune. Most protein-protein docking looks at clustering of docking models in graphs depicting RMS vs predicted energy of interaction. The lower the RMS and predicted energy of the cluster the better. Furthermore it is important to asses where the predicted interaction sites are and how frequently they are repeated or overlapping. Do the intermolecular forces within the predicted binding sites make sense?
The real question is how good are the structures you have. Are they models or crystal structures. Are you able to restrict the likely interaction site based on in vitro data? If so are there unstructured loops involved? You may need multiple conformations of both proteins representing a pre- and post- docking set of structures due to an induced fit scenario.
Hey John, Thank you for your answer. We have crystal structure of proteins. Although we are speculating their binding sites, but we want to do a blind docking as we are not very sure of the interacting regions. we would also like to compare the blind docking results with docking at our speculated sites to see if our preduction matches with that of the software. There are no unstructured loops involved.
One problem is that ClusPro is generating many docked structures and i dont know which one to choose.
Another thing is that i want to see the interacting residues and the type of bonds these residues are forming
Deblina,
There may not be a "right" answer given your approach. However you may begin to address your hypothesis from a statistical vantage point by asking what are the most prominent predicted sites of interaction. Given that ClusPro gives a large number of docked poses you may wish to choose the top 10 or even 30 and examine them for gross similarities such as amino positions involved for each protein.
You might want to decide qualitatively where your cut off is based on the drop off in cluster size (under view model scores see the number of members represented in a cluster). Remember the "top" model is the docked complex with the lowest RMS for each cluster. I performed a test dock today and my largest cluster was 84 some mid range clusters were around 40 and I would consider the lowest to have been 20 or less. Given the fact that I only had 29 models I might only pick the top 15 given this last complex had a cluster size of 24. However the top ten might also be good in this scenario. Remember this is the best of 70,000 possible interactions.
Analysis of the most frequently predicted interacting amino acids as well as some "representative" docked models may provide you with some ideas of what kinds of intermolecular forces are at play. However it is always important to remember that further in vitro experimentation is required to confirm computer predictions.
As for visualizing your results I suggest Pymol (http://www.pymol.org/). This a relatively inexpensive though competent molecular viewer. However if you you want freeware autodock tools (MGLtools) is also very good (http://mgltools.scripps.edu/) though you may have to separate the proteins manually in each complex. In this case the ClusPro help page can give you more information (http://cluspro.bu.edu/help.php). Furthermore, each of these molecular views have faqs and guides to help users (including analysis of intermolecular forces).
Good luck!
-John
John,
Thank you for your suggestions. I will analyze the first 15 models with higher cluster numbers and yes in vitro experiments would be necessary.
Regarding visualization, I am already using Pymol to see the models and can also label the residues but cannot see what kind of interactions they are forming (hydrogen bonds or ionic bonds).
To find out interactions in MGL tools, they need .dlg files i.e docking log files but ClusPro doesnt give such files. Earlier i used Autodock for protein-ligand docking and it generated .dlg files which could then be visualized on ADT to find the interactions.
Deblina,
There are a number of places that go over the details of using pymol for analysis. A group from UCLA goes over more than you probably need/want to know (http://www.doe-mbi.ucla.edu/CHEM125/pymol_tutorial_060418.pdf). This pymol wiki page delineates the process of visualizing molecular forces in pymol succinctly (http://www.pymolwiki.org/index.php/Displaying_Biochemical_Properties#Contact_Potential).
For polar interactions it is relatively easy, "Using the actions [A] button for an object or selection you can display Hydrogen bonds and Polar Contacts. [A]->find->polar contacts->" I have found that "to any atoms" works best.
- Badges
- Science topic
- Similar topics
- Filing