"3) And the target should not have missing residues; [how can I know about this criteria?]"
Compare the sequence downloaded from the pdb (SEQRES records) to the sequence extracted from the coordinates.
"How can I select the best one for docking?"
Do a 3D alignment of the potential template structures and look at it carefully - it gives you an idea of which parts of the structure are flexible.
If there is a lot of flexibility in the vicinity of the putative ligand binding site, you might consider using several different structures representing different conformational states.
By comparing free to liganded structures, you can assess whether there are conformational changes associated with ligand binding. If so, a liganded structure (with the ligand removed) is better for docking.
From my point of view, in some cases, NMR structures are better than crystal structures for molecular docking. Because crystal structure is just a static structure, but proteins usually have different conformations with dynamic motions which may be important for the binding to its target. NMR structures are ensemble with different conformations which will be a good starting point for docking.
Also, HADDOCK is recommended if you have any information about the interactions between these two (or more) molecules.
Pengfei Ding , Mario E. Valdés-Tresanco , Annemarie Honegger
thank you.
since i need to use automation (for tens of targets and hundreds of legends) in a network, the practical criteria is the resolution; but i will try to count missing residues.