What is the reason for adding hydrogen and removing unnecessary water molecules from the protein structure before protein-ligand docking? FYI, the tools I used for docking is GOLD.
My answer is mainly based on molecular dynamics,but in this context it shouldn't make much of a difference.
The main reason has to do with the
water model used during the simulation. Examples of water models are
TIP4P; same molecule (water) but different topology.
pdb usually comes from different sources and it can contain molecules that are not standard and therefore not recognised by the model of your simulation. Usually, once the structure is loaded, your system is
re-built so that it conforms to the model topology. There are exception though, for example if your water is part of a functional active site. In that case you can keep it, but its structures (topology of the water molecule) has to be recognised by the model.
Hydrogens are usually added to "saturate" and stabilise the bonds.
Most crystallographic structures are X-ray and the hydrogens are absent. In the case of neutron diffraction models, there may be many missing high b-factor ones (example in PDB:5MOP). As a consequence models are reprotonated. Also missing carbons and missing loop residues are added for some in silico applications too (not docking).
However, it is always good to manually tweak the protonation. In the case of enzymes with protanatable side chains in the active site, some reading is required to get the charges right: by default aspartates will be unprotonated, but they may be protonated (glutamic acid) in the active site and it is important to get it right (cf. this post about it). Some times its the ligand parameterisation may give the physiologically incorrect protonation (cf. this post).
Crystallographic waters can be important catalytically and structurally, but can be pushed out and actually it is a good thing: you have a hit that take its place. Polar bonds are stronger than hydrophobic ones and most types of π bonding.
Waters in most docking software is
implicit and is modelled as a parameter in the force-fields —think of it as a ghostly watery haze. So a left behind water, not only will get in the way of the docked small molecule, it may for some programs cause issues as they will try to dock it. A notable exception is WaterDock mod for AutoDock, but the jury is out on whether it is better. Water is polarisable in reality in a very complex way (protonated (OH3+) or deprotonated (OH-) with very different charges) and most water models do not cut it and instead do not account for this.
The way implicit waters are computed differs in some applications, which is also something to look into.