Having a very good starting model before starting any investigation of protein chemistry is essential. The starting model should be solvated, as many faults as possible should be corrected, all the parts needed should be present - the protein, the substrate, any water or other small molecules, etc., and the amount of ionization should be correct. This last requirement depends on the chemistry of the protein, on the pH of the solvent, and on other factors. Because of these factors, the problem of determining which sites are ionized might seem very hard to solve. In practice, however, the process of determining which sites are ionized is really quite straightforward.
By default, when hydrogen atoms are added to a PDB file or other data set, all valencies are satisfied by the addition of hydrogen atoms, i.e., no charged sites are generated. This option was deliberately chosen so that an unambiguous starting structure would be produced as a starting point for further manipulations. Of course, charged sites exist in real proteins, and the objective at this point is to selectively ionize sites in a protein in order to generate a more realistic model.
Although obvious, it bears repeating: know as much as possible about the system being studied, in particular about the ionized sites in it. If there is an ionized site, then add or remove hydrogen atoms as needed in order to have the correct ionization. This can be done using a GUI, or by using keyword SITE.
When the positions of the hydrogen atoms are optimized some salt bridges might form spontaneously, but in many cases there is a barrier to proton migration and a salt bridge that should exist does not in fact form. The easiest way to make a specific salt bridge is to use SITE to ionize both residues that form the salt bridge. Quite often it is not easy to know if a salt bridge should exist, however there is an easy test to answer this question. If a salt bridge lowers the heat of formation, then the salt bridge should exist, and vice versa. To test whether a salt bridge lowers the heat of formation, run two geometry optimization jobs - one with the salt bridge and one without the salt bridge. Of course, in order to allow heats of formation to be compared, the set of keywords used should be the same, in particular if EPS=78.4 is used in one job it should be used in the other.