I am trying to dock a small molecule with PDB NMR structure, So it has 10 model structure, in that case, How should I move further, I mean which model structure has to choose and on what criteria??
Normally, all models within an NMR pdb file equally fulfil the experimentally determined distance constraints.If the models diverge, this means that there was no information available which would constrain that part of the model to a single conformation.
The reason why I asked, cause I did docking for the first model and I found the most probable site, which is the one which shows more constraint in the remaining NMR model structure.
Hope you got the point and please guide further to assist the problem.
I think it would be logical to move with the structure that has higher resolution. Then prepare the protein structures and then look for probable sites by various methods available. Then based on the prior knowledge and site scores decide that which is the best. Moreover do some grid validation study like in situ redocking and ROC using a decoy set. See which is more suitable grid and go ahead with the one with higher conformity.
RCSB database ususally reports 10/20 NMR models for each structure. The structure website has actually this listed as: "Conformers calcualted" and "Conformers submitted". It also always has listed the method, how these 10/20 "best" conformations are selected - usually it's by lowest energy (molecular mechanics / force field). Bear in mind that this is a very simple method which may give wrong results. Also, in many cases the energy differences between different conformers may be very small (these are not listed by RCSB), making it somewhat difficult to claim that these 10/20 conformations are really representative od the structure in question.
So few ideas how to deal with it.
1. Calculate RMSD (or a similar measure) for the entire set of 10 models (each model to each model to get a RMSD matrix). If it's small, it means that the structure is well-defined and you'll have it easier. It would by anyhow good to dock the molecule to all 10 models to account for protein flexibility. If you can't afford it docking to no 1 only (if best by energy) is somewhat justified. Docking to a model which has the lowest mean RMSD to all oter 9 models would be, however, also a good idea.
2. If the RMSD is large, you're in trouble. Definitely dock to all 10 models, since the docking results will be quite different. In such case it may be, however, worth performing a longer MD of the protein and then dock to as many selected snapshots (selected by low energy, frequency of occuring or something similar) as you can afford.