I use RosettaDock to find the binding site of the proteins. The structure is solved by NMR.
I use the default option:
in:files:s vegf_vegf.pdb -docking:randomize1 -out:nstruct 1000 -out:file:o vegfdimer
I did twice with the same option and the same pdb file, but the result is different.
The first one is right. The second is wrong. I don’t know what is the reason?
It seem that it is not reproducible.
I search the forum to fine some answers.
1. NO other data: lowest score = best result
2. Experimental hints: Using those to figure out which model is most likely to be accurate
Why I have the different result with the same initial condition?
How can I confirm that the protein really search the whole surface of the other protein?
How the program move one of the protein to the surface?
Dose it make a latitude and longitude lines on the protein?
Thanks for reading this.
"Why I have the different result with the same initial condition?"
Rosetta uses a Monte Carlo search protocol. You had a different random number seed and thus got different results. Combining the flags -constant_seed and -jran will allow you to manually set the RNG seed and make runs produce the same results.
"How can I confirm that the protein really search the whole surface of the other protein?"
With these flags, it should search the entire surface of the first one, but not search across the surface of the second. If you ran nstruct 1000, you should have 1000 results, which should be pretty widely distributed.
"How the program move one of the protein to the surface?"
I'm not sure what you're asking. It samples the rigid-body degree of freedom between the two substructures, which includes changing the distance between their centers of mass...
"Dose it make a latitude and longitude lines on the protein?"
It is not doing gridded docking by default, nor is there a gridding mode (that I'm aware of).
1000 decoys is far too few for a global search of one partner. 10^4 - 10^6 is more reasonable, depending on the proteins sizes. Please see the 2003 JMB paper and the CAPRI papers for a variety of examples.